Skip to content

The fastest, litest AI Gateway. Rust core with Python SDK. Call 100+ LLM APIs in OpenAI (or native) format with cost tracking, guardrails, load balancing, and logging [Bedrock, Azure, OpenAI, Anthropic, OpenAI, VertexAI, vLLM, Nvidia NIM]

fix(llms): join api_base and route with one shared helper (#47535)

Mateo Wang66c8f5b1a9ce
diff --git a/.github/workflows/test-unit.yml b/.github/workflows/test-unit.yml
index c593acbe8fb..2276efc4e34 100644
--- a/.github/workflows/test-unit.yml
+++ b/.github/workflows/test-unit.yml
@@ -216,10 +216,12 @@ jobs:
- shard: llms-providers
test-path: |-
+ tests/unit/llms/test_api_base_trailing_slash_routes.py
tests/unit/llms/test_cache_control_and_reasoning.py
tests/unit/llms/test_custom_llm.py
tests/unit/llms/test_file_content_block.py
tests/unit/llms/test_file_search_responses.py
+ tests/unit/llms/test_get_complete_url_trailing_slash.py
tests/unit/llms/test_lifecycle_fix.py
tests/unit/llms/test_oss_decision.py
tests/unit/llms/test_polling_url_origin_match.py
diff --git a/litellm/litellm_core_utils/url_utils.py b/litellm/litellm_core_utils/url_utils.py
index 43e16599bf6..4a1eaa87d3a 100644
--- a/litellm/litellm_core_utils/url_utils.py
+++ b/litellm/litellm_core_utils/url_utils.py
@@ -23,7 +23,7 @@ import asyncio
import socket
from ipaddress import ip_address, ip_network
from typing import Any, Final, Protocol
-from urllib.parse import quote, urlparse, urlunparse
+from urllib.parse import quote, urlparse, urlsplit, urlunparse
import httpx
from typing_extensions import ReadOnly, TypedDict
@@ -129,6 +129,28 @@ def encode_url_path_segments(value: object, *, field_name: str = "path") -> str:
return "/".join(encode_url_path_segment(segment, field_name=field_name) for segment in value_str.split("/"))
+def _strip_trailing_slashes_keeping_the_scheme(api_base: str) -> str:
+ stripped: Final = api_base.rstrip("/")
+ parts: Final = urlsplit(stripped)
+ if parts.scheme and not (parts.netloc or parts.path or parts.query or parts.fragment):
+ return api_base
+ if not parts.query or not parts.path.endswith("/"):
+ return stripped
+ url_before_query, _, query = stripped.partition("?")
+ return f"{url_before_query.rstrip('/')}?{query}"
+
+
+def join_api_base(api_base: str, route: str) -> str:
+ return f"{_strip_trailing_slashes_keeping_the_scheme(api_base)}/{route.lstrip('/')}"
+
+
+def ensure_api_route(api_base: str, route: str) -> str:
+ base: Final = _strip_trailing_slashes_keeping_the_scheme(api_base)
+ if urlsplit(base).path.endswith(f"/{route.lstrip('/')}"):
+ return base
+ return join_api_base(base, route)
+
+
def _is_blocked_ip(addr: str) -> bool:
"""Return True for any IP not safe to reach from a user-supplied URL.
diff --git a/litellm/llms/aiml/image_generation/transformation.py b/litellm/llms/aiml/image_generation/transformation.py
index 931038689cc..bb981983e98 100644
--- a/litellm/llms/aiml/image_generation/transformation.py
+++ b/litellm/llms/aiml/image_generation/transformation.py
@@ -4,6 +4,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
from pydantic import ConfigDict, TypeAdapter
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
)
@@ -121,7 +122,7 @@ class AimlImageGenerationConfig(BaseImageGenerationConfig):
complete_url = complete_url.rstrip("/")
# Strip /v1 suffix if present since IMAGE_GENERATION_ENDPOINT already includes v1
complete_url = complete_url.removesuffix("/v1")
- complete_url = f"{complete_url}/{self.IMAGE_GENERATION_ENDPOINT}"
+ complete_url = join_api_base(complete_url, self.IMAGE_GENERATION_ENDPOINT)
return complete_url
def validate_environment(
diff --git a/litellm/llms/aiohttp_openai/chat/transformation.py b/litellm/llms/aiohttp_openai/chat/transformation.py
index dba7a7405ab..74f782b56e7 100644
--- a/litellm/llms/aiohttp_openai/chat/transformation.py
+++ b/litellm/llms/aiohttp_openai/chat/transformation.py
@@ -13,6 +13,7 @@ from typing import TYPE_CHECKING, Any, Final
from aiohttp import ClientResponse
from pydantic import ConfigDict, TypeAdapter
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.openai_like.chat.transformation import OpenAILikeChatConfig
from litellm.types.llms.openai import AllMessageValues
from litellm.types.utils import Choices, ModelResponse
@@ -44,12 +45,7 @@ class AiohttpOpenAIChatConfig(OpenAILikeChatConfig):
Ensure - /v1/chat/completions is at the end of the url
"""
- if api_base is None:
- api_base = "https://api.openai.com"
-
- if not api_base.endswith("/chat/completions"):
- api_base += "/chat/completions"
- return api_base
+ return ensure_api_route("https://api.openai.com" if api_base is None else api_base, "chat/completions")
def validate_environment(
self,
diff --git a/litellm/llms/anthropic/batches/transformation.py b/litellm/llms/anthropic/batches/transformation.py
index b95b734c761..da6c2871289 100644
--- a/litellm/llms/anthropic/batches/transformation.py
+++ b/litellm/llms/anthropic/batches/transformation.py
@@ -7,7 +7,7 @@ import httpx
from httpx import Headers, Response
from typing_extensions import ReadOnly, TypedDict
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, ensure_api_route, join_api_base
from litellm.llms.base_llm.batches.transformation import BaseBatchesConfig
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.types.llms.openai import AllMessageValues, CreateBatchRequest
@@ -106,10 +106,8 @@ class AnthropicBatchesConfig(BaseBatchesConfig):
data: CreateBatchRequest,
) -> str:
"""Get the complete URL for batch creation request."""
- api_base = api_base or self.anthropic_model_info.get_api_base(api_base)
- if not api_base.endswith("/v1/messages/batches"):
- api_base = f"{api_base.rstrip('/')}/v1/messages/batches"
- return api_base
+ api_base = self.anthropic_model_info.resolve_anthropic_api_base(api_base)
+ return ensure_api_route(api_base, "v1/messages/batches")
def transform_create_batch_request(
self,
@@ -158,9 +156,9 @@ class AnthropicBatchesConfig(BaseBatchesConfig):
Returns:
Complete URL for Anthropic batch retrieval: {api_base}/v1/messages/batches/{batch_id}
"""
- api_base = api_base or self.anthropic_model_info.get_api_base(api_base)
+ api_base = self.anthropic_model_info.resolve_anthropic_api_base(api_base)
encoded_batch_id: Final = encode_url_path_segment(batch_id, field_name="batch_id")
- return f"{api_base.rstrip('/')}/v1/messages/batches/{encoded_batch_id}"
+ return join_api_base(api_base, f"v1/messages/batches/{encoded_batch_id}")
def transform_retrieve_batch_request(
self,
diff --git a/litellm/llms/anthropic/common_utils.py b/litellm/llms/anthropic/common_utils.py
index 6f08d6b0d61..1f5e4e7de39 100644
--- a/litellm/llms/anthropic/common_utils.py
+++ b/litellm/llms/anthropic/common_utils.py
@@ -35,6 +35,7 @@ from litellm.litellm_core_utils.prompt_templates.mid_conversation_system import
parts_of,
)
from litellm.litellm_core_utils.reasoning_effort_utils import non_numeric_thinking_budget_error
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.anthropic.wif import (
aget_anthropic_wif_token,
anthropic_base_without_chat_suffix,
@@ -468,7 +469,7 @@ def _fetch_anthropic_models_page(
# which a read-only one cannot support
query: Final = f"?after_id={quote(after_id)}" if after_id else ""
response: Final = litellm.module_level_client.get(
- url=f"{api_base}/v1/models{query}",
+ url=f"{join_api_base(api_base, 'v1/models')}{query}",
headers=headers,
follow_redirects=False,
)
@@ -1406,7 +1407,7 @@ class AnthropicModelInfo(BaseLLMModelInfo):
return {**caller_headers, **anthropic_headers}
@staticmethod
- def get_api_base(api_base: str | None = None) -> str | None:
+ def resolve_anthropic_api_base(api_base: str | None = None) -> str:
from litellm.secret_managers.main import get_secret_str
return (
@@ -1416,6 +1417,10 @@ class AnthropicModelInfo(BaseLLMModelInfo):
or "https://api.anthropic.com"
)
+ @staticmethod
+ def get_api_base(api_base: str | None = None) -> str | None:
+ return AnthropicModelInfo.resolve_anthropic_api_base(api_base)
+
@staticmethod
def get_api_key(api_key: str | None = None) -> str | None:
"""An empty or whitespace-only key counts as unset: it can never authenticate anything, and
@@ -1533,14 +1538,14 @@ class AnthropicModelInfo(BaseLLMModelInfo):
api_base: str | None,
litellm_params: Mapping[str, object] | None,
) -> list[str]: # mutable-ok: matches get_models' list[str] contract shared by every provider override
- resolved_api_base: Final = AnthropicModelInfo.get_api_base(api_base)
+ resolved_api_base: Final = AnthropicModelInfo.resolve_anthropic_api_base(api_base)
auth_header: Final = AnthropicModelInfo.get_auth_header(
api_key,
resolved_api_base,
litellm_params=litellm_params,
allow_workload_identity=config_allows_workload_identity(self),
)
- if resolved_api_base is None or auth_header is None:
+ if auth_header is None:
raise ValueError(
"ANTHROPIC_API_BASE/ANTHROPIC_BASE_URL or ANTHROPIC_API_KEY/ANTHROPIC_AUTH_TOKEN (or workload "
"identity federation via ANTHROPIC_FEDERATION_RULE_ID/ANTHROPIC_ORGANIZATION_ID/"
diff --git a/litellm/llms/anthropic/count_tokens/transformation.py b/litellm/llms/anthropic/count_tokens/transformation.py
index 592b5fb0f00..c2d87c641be 100644
--- a/litellm/llms/anthropic/count_tokens/transformation.py
+++ b/litellm/llms/anthropic/count_tokens/transformation.py
@@ -25,6 +25,7 @@ from litellm.litellm_core_utils.prompt_templates.mid_conversation_system import
anthropic_system_blocks,
split_leading_system_run,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.anthropic.common_utils import merge_anthropic_beta_headers
from litellm.llms.anthropic.wif import resolve_anthropic_base
from litellm.types.llms.openai import AllMessageValues, ChatCompletionImageObject, ValidUserMessageContentTypes
@@ -317,7 +318,7 @@ class AnthropicCountTokensConfig:
Returns:
The endpoint URL for the CountTokens API
"""
- return resolve_anthropic_base(api_base) + "/v1/messages/count_tokens"
+ return join_api_base(resolve_anthropic_base(api_base), "v1/messages/count_tokens")
def transform_request_to_count_tokens(
self,
diff --git a/litellm/llms/anthropic/files/handler.py b/litellm/llms/anthropic/files/handler.py
index f1f9c204b58..1a793e32317 100644
--- a/litellm/llms/anthropic/files/handler.py
+++ b/litellm/llms/anthropic/files/handler.py
@@ -10,7 +10,7 @@ import litellm
from litellm._logging import verbose_logger
from litellm._uuid import uuid
from litellm.litellm_core_utils.litellm_logging import Logging
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.custom_httpx.http_handler import get_async_httpx_client
from litellm.types.llms.openai import (
FileContentRequest,
@@ -74,7 +74,7 @@ class AnthropicFilesHandler:
batch_id = file_id
# Get Anthropic API credentials
- api_base = self.anthropic_model_info.get_api_base(api_base)
+ api_base = self.anthropic_model_info.resolve_anthropic_api_base(api_base)
auth_header: Final = await self.anthropic_model_info.aget_auth_header(
api_key, api_base, litellm_params=litellm_params, allow_workload_identity=True
)
@@ -84,7 +84,7 @@ class AnthropicFilesHandler:
# Construct the Anthropic batch results URL
encoded_batch_id: Final = encode_url_path_segment(batch_id, field_name="batch_id")
- results_url: Final = f"{api_base.rstrip('/')}/v1/messages/batches/{encoded_batch_id}/results"
+ results_url: Final = join_api_base(api_base, f"v1/messages/batches/{encoded_batch_id}/results")
# Prepare headers
headers: Final = {
diff --git a/litellm/llms/anthropic/files/transformation.py b/litellm/llms/anthropic/files/transformation.py
index 8433235dde4..db14d3366ff 100644
--- a/litellm/llms/anthropic/files/transformation.py
+++ b/litellm/llms/anthropic/files/transformation.py
@@ -22,7 +22,7 @@ from openai.types.file_deleted import FileDeleted
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.prompt_templates.common_utils import extract_file_data
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.files.transformation import (
BaseFilesConfig,
@@ -80,7 +80,7 @@ class AnthropicFilesConfig(BaseFilesConfig):
stream: bool | None = None,
) -> str:
api_base = AnthropicModelInfo.get_api_base(api_base) or ANTHROPIC_FILES_API_BASE
- return f"{api_base.rstrip('/')}/v1/files"
+ return join_api_base(api_base, "v1/files")
def get_error_class(
self,
@@ -226,7 +226,7 @@ class AnthropicFilesConfig(BaseFilesConfig):
) -> tuple[str, dict]:
api_base: Final = AnthropicModelInfo.get_api_base(litellm_params.get("api_base")) or ANTHROPIC_FILES_API_BASE
encoded_file_id: Final = encode_url_path_segment(file_id, field_name="file_id")
- return f"{api_base.rstrip('/')}/v1/files/{encoded_file_id}", {}
+ return join_api_base(api_base, f"v1/files/{encoded_file_id}"), {}
def transform_retrieve_file_response(
self,
@@ -245,7 +245,7 @@ class AnthropicFilesConfig(BaseFilesConfig):
) -> tuple[str, dict]:
api_base: Final = AnthropicModelInfo.get_api_base(litellm_params.get("api_base")) or ANTHROPIC_FILES_API_BASE
encoded_file_id: Final = encode_url_path_segment(file_id, field_name="file_id")
- return f"{api_base.rstrip('/')}/v1/files/{encoded_file_id}", {}
+ return join_api_base(api_base, f"v1/files/{encoded_file_id}"), {}
def transform_delete_file_response(
self,
@@ -264,7 +264,7 @@ class AnthropicFilesConfig(BaseFilesConfig):
litellm_params: dict,
) -> tuple[str, dict]:
api_base: Final = AnthropicModelInfo.get_api_base(litellm_params.get("api_base")) or ANTHROPIC_FILES_API_BASE
- url: Final = f"{api_base.rstrip('/')}/v1/files"
+ url: Final = join_api_base(api_base, "v1/files")
params: Final[dict[str, str]] = {}
if purpose:
params["purpose"] = purpose
@@ -299,8 +299,8 @@ class AnthropicFilesConfig(BaseFilesConfig):
api_base: Final = AnthropicModelInfo.get_api_base(litellm_params.get("api_base")) or ANTHROPIC_FILES_API_BASE
encoded_file_id: Final = encode_url_path_segment(file_id, field_name="file_id")
if file_id.startswith(ANTHROPIC_MESSAGE_BATCH_ID_PREFIX):
- return f"{api_base.rstrip('/')}/v1/messages/batches/{encoded_file_id}/results", {}
- return f"{api_base.rstrip('/')}/v1/files/{encoded_file_id}/content", {}
+ return join_api_base(api_base, f"v1/messages/batches/{encoded_file_id}/results"), {}
+ return join_api_base(api_base, f"v1/files/{encoded_file_id}/content"), {}
def transform_file_content_response(
self,
diff --git a/litellm/llms/anthropic/pass_through/messages/transformation.py b/litellm/llms/anthropic/pass_through/messages/transformation.py
index a71cd72b768..9f3aa0130d1 100644
--- a/litellm/llms/anthropic/pass_through/messages/transformation.py
+++ b/litellm/llms/anthropic/pass_through/messages/transformation.py
@@ -9,6 +9,7 @@ from litellm.anthropic_beta_headers_manager import get_unsupported_headers, is_b
from litellm.exceptions import AuthenticationError
from litellm.litellm_core_utils.core_helpers import drops_unsupported_params
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.anthropic.pass_through.convertible_blocks import drop_non_string_tool_result_texts
from litellm.llms.anthropic.pass_through.safeguards import proxy_serves_safeguards
from litellm.llms.base_llm.anthropic_messages.transformation import (
@@ -287,10 +288,7 @@ class AnthropicMessagesConfig(BaseAnthropicMessagesConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- api_base = AnthropicModelInfo.get_api_base(api_base) or "https://api.anthropic.com"
- if not api_base.endswith("/v1/messages"):
- api_base = f"{api_base}/v1/messages"
- return api_base
+ return ensure_api_route(AnthropicModelInfo.get_api_base(api_base) or "https://api.anthropic.com", "v1/messages")
def validate_anthropic_messages_environment(
self,
diff --git a/litellm/llms/anthropic/skills/transformation.py b/litellm/llms/anthropic/skills/transformation.py
index 6d419a56889..8ea9d192c24 100644
--- a/litellm/llms/anthropic/skills/transformation.py
+++ b/litellm/llms/anthropic/skills/transformation.py
@@ -9,7 +9,7 @@ import httpx
from pydantic import TypeAdapter
from litellm._logging import verbose_logger
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.skills.transformation import (
BaseSkillsAPIConfig,
LiteLLMLoggingObj,
@@ -75,13 +75,11 @@ class AnthropicSkillsConfig(BaseSkillsAPIConfig):
"""Get complete URL for Anthropic Skills API"""
from litellm.llms.anthropic.common_utils import AnthropicModelInfo
- if api_base is None:
- api_base = AnthropicModelInfo.get_api_base()
-
+ resolved_api_base: Final = AnthropicModelInfo.get_api_base(api_base) or "https://api.anthropic.com"
if skill_id:
encoded_skill_id: Final = encode_url_path_segment(skill_id, field_name="skill_id")
- return f"{api_base}/v1/skills/{encoded_skill_id}"
- return f"{api_base}/v1/{endpoint}"
+ return join_api_base(resolved_api_base, f"v1/skills/{encoded_skill_id}")
+ return join_api_base(resolved_api_base, f"v1/{endpoint}")
def transform_create_skill_request(
self,
diff --git a/litellm/llms/anthropic/wif.py b/litellm/llms/anthropic/wif.py
index b47c4f28eb2..ccb81c1244f 100644
--- a/litellm/llms/anthropic/wif.py
+++ b/litellm/llms/anthropic/wif.py
@@ -15,6 +15,7 @@ from typing_extensions import assert_never
import litellm
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.auth.client_credentials import keycloak_assertion_source
from litellm.llms.base_llm.auth.identity_source import (
AnthropicIdentitySourceKind,
@@ -324,7 +325,7 @@ def anthropic_internal_issuer_jwks(credential_values: Mapping[str, object]) -> A
def build_anthropic_wif_spec(params: AnthropicWifParams, api_base: str) -> TokenExchangeSpec:
return TokenExchangeSpec(
- token_url=api_base.rstrip("/") + ANTHROPIC_TOKEN_EXCHANGE_PATH,
+ token_url=join_api_base(api_base, ANTHROPIC_TOKEN_EXCHANGE_PATH),
assertion_ref=params.assertion_ref,
assertion_field="assertion",
static_body=MappingProxyType(
diff --git a/litellm/llms/apiserpent/search/transformation.py b/litellm/llms/apiserpent/search/transformation.py
index bb8abff7272..faf1903ffe8 100644
--- a/litellm/llms/apiserpent/search/transformation.py
+++ b/litellm/llms/apiserpent/search/transformation.py
@@ -14,6 +14,7 @@ from urllib.parse import urlencode
import httpx
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.apiserpent.search.defaults import (
DEEP_SEARCH_PATH,
NUM_MAX,
@@ -77,13 +78,11 @@ class APISerpentSearchConfig(BaseSearchConfig):
Build the search URL. APISerpent uses GET, so the transformed request is
serialized into the query string. The endpoint path (quick vs deep) is
always applied; an ``api_base`` / ``APISERPENT_API_BASE`` override only
- changes the host. The ``endswith`` guard keeps this idempotent, since the
+ changes the host. ``ensure_api_route`` keeps this idempotent, since the
handler re-invokes this method with the already-resolved URL as api_base.
"""
- base = (api_base or get_secret_str("APISERPENT_API_BASE") or APISERPENT_BASE).rstrip("/")
path: Final = DEEP_SEARCH_PATH if self._is_deep_search(optional_params) else QUICK_SEARCH_PATH
- if not base.endswith(path):
- base = f"{base}{path}"
+ base: Final = ensure_api_route(api_base or get_secret_str("APISERPENT_API_BASE") or APISERPENT_BASE, path)
if data and isinstance(data, dict) and APISERPENT_PARAMS_KEY in data:
query_string: Final = urlencode(data[APISERPENT_PARAMS_KEY], doseq=True)
diff --git a/litellm/llms/azure/audio_transcription/transformation.py b/litellm/llms/azure/audio_transcription/transformation.py
index 858f122f65a..a5fde7f7158 100644
--- a/litellm/llms/azure/audio_transcription/transformation.py
+++ b/litellm/llms/azure/audio_transcription/transformation.py
@@ -13,6 +13,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.audio_utils.utils import process_audio_file
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.audio_transcription.transformation import (
AudioTranscriptionRequestData,
BaseAudioTranscriptionConfig,
@@ -110,7 +111,7 @@ class AzureSpeechAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
"language": optional_params.get("language", self.DEFAULT_LANGUAGE),
"format": self._get_azure_response_format(optional_params.get("response_format")),
}
- return f"{base_url}{self.STT_ENDPOINT_PATH}?{urlencode(query_params)}"
+ return f"{join_api_base(base_url, self.STT_ENDPOINT_PATH)}?{urlencode(query_params)}"
def transform_audio_transcription_request(
self,
@@ -153,7 +154,6 @@ class AzureSpeechAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
)
def _resolve_stt_base_url(self, api_base: str) -> str:
- api_base = api_base.rstrip("/")
parsed_url: Final = urlparse(api_base)
hostname: Final = parsed_url.hostname or ""
diff --git a/litellm/llms/azure/common_utils.py b/litellm/llms/azure/common_utils.py
index 52e3cdf380e..23c5aa35f4b 100644
--- a/litellm/llms/azure/common_utils.py
+++ b/litellm/llms/azure/common_utils.py
@@ -18,6 +18,7 @@ import litellm
from litellm._logging import verbose_logger
from litellm.caching.caching import DualCache
from litellm.constants import DEFAULT_MAX_RETRIES
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.custom_httpx.http_handler import HTTPHandler
from litellm.llms.openai.common_utils import BaseOpenAILLM, OpenAIAsyncHTTPClient, OpenAIHTTPClient
@@ -616,7 +617,7 @@ class BaseAzureLLM(BaseOpenAILLM):
v1_params: Final[_AzureV1ClientParams] = {
"api_key": v1_api_key,
- "base_url": f"{api_base}/openai/v1/",
+ "base_url": f"{ensure_api_route(api_base or '', 'openai/v1')}/",
}
if "timeout" in azure_client_params:
v1_params["timeout"] = azure_client_params["timeout"]
diff --git a/litellm/llms/azure/realtime/handler.py b/litellm/llms/azure/realtime/handler.py
index 75f96bfdf63..3d4576fd7ac 100644
--- a/litellm/llms/azure/realtime/handler.py
+++ b/litellm/llms/azure/realtime/handler.py
@@ -10,6 +10,7 @@ from typing import Final, Protocol, cast
from litellm._logging import verbose_proxy_logger
from litellm.constants import REALTIME_WEBSOCKET_MAX_MESSAGE_SIZE_BYTES
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.types.realtime import RealtimeQueryParams
from ....litellm_core_utils.litellm_logging import Logging as LiteLLMLogging
@@ -136,7 +137,8 @@ class AzureOpenAIRealtime(AzureChatCompletion):
query_parts.append(urlencode({"intent": intent}))
qs: Final = "&".join(query_parts)
- return f"{api_base}{path}?{qs}" if qs else f"{api_base}{path}"
+ endpoint_url: Final = join_api_base(api_base, path)
+ return f"{endpoint_url}?{qs}" if qs else endpoint_url
def construct_url(
self,
diff --git a/litellm/llms/azure/realtime/http_transformation.py b/litellm/llms/azure/realtime/http_transformation.py
index 542a6a8c02f..545c3348182 100644
--- a/litellm/llms/azure/realtime/http_transformation.py
+++ b/litellm/llms/azure/realtime/http_transformation.py
@@ -3,6 +3,7 @@
from typing import Final
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.realtime.http_transformation import BaseRealtimeHTTPConfig
from litellm.secret_managers.main import get_secret_str
@@ -39,9 +40,9 @@ class AzureRealtimeHTTPConfig(BaseRealtimeHTTPConfig):
return _azure_realtime_v1_url(self.get_api_base(api_base), "calls")
def get_transcription_session_url(self, api_base: str | None, model: str, api_version: str | None = None) -> str:
- base: Final = self.get_api_base(api_base).rstrip("/")
+ base: Final = self.get_api_base(api_base)
version: Final = api_version or get_secret_str("AZURE_API_VERSION") or "2024-12-17"
- return f"{base}/openai/realtime/transcription_sessions?api-version={version}"
+ return f"{join_api_base(base, 'openai/realtime/transcription_sessions')}?api-version={version}"
def get_realtime_calls_headers(self, ephemeral_key: str) -> dict[str, str]:
return {
diff --git a/litellm/llms/azure/search/transformation.py b/litellm/llms/azure/search/transformation.py
index f2bb2ecc68c..f7cdf88ade0 100644
--- a/litellm/llms/azure/search/transformation.py
+++ b/litellm/llms/azure/search/transformation.py
@@ -35,6 +35,7 @@ from typing import TYPE_CHECKING, Final, Literal
import httpx
from pydantic import ConfigDict, ValidationError
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.azure.common_utils import cached_azure_ad_token_refresh_provider
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.search.transformation import (
@@ -340,10 +341,7 @@ class BingGroundingSearchConfig(BaseSearchConfig):
f"{PROJECT_ENDPOINT_ENV} is not set. Set it to your Microsoft Foundry project "
f"endpoint, e.g. https://<account>.services.ai.azure.com/api/projects/<project>."
)
- trimmed: Final = resolved_base.rstrip("/")
- if trimmed.endswith(_RESPONSES_PATH):
- return trimmed
- return f"{trimmed}{_RESPONSES_PATH}"
+ return ensure_api_route(resolved_base, _RESPONSES_PATH)
def transform_search_request(
self,
diff --git a/litellm/llms/azure/text_to_speech/transformation.py b/litellm/llms/azure/text_to_speech/transformation.py
index eed7a3178ca..acae650ed75 100644
--- a/litellm/llms/azure/text_to_speech/transformation.py
+++ b/litellm/llms/azure/text_to_speech/transformation.py
@@ -11,6 +11,7 @@ from urllib.parse import urlparse
import httpx
import litellm
+from litellm.litellm_core_utils.url_utils import ensure_api_route, join_api_base
from litellm.llms.base_llm.text_to_speech.transformation import (
BaseTextToSpeechConfig,
TextToSpeechRequestData,
@@ -329,8 +330,6 @@ class AzureAVATextToSpeechConfig(BaseTextToSpeechConfig):
f"https://{{region}}.{self.TTS_SPEECH_DOMAIN}"
)
- # Remove trailing slash and parse URL
- api_base = api_base.rstrip("/")
parsed_url: Final = urlparse(api_base)
hostname: Final = parsed_url.hostname or ""
@@ -339,14 +338,10 @@ class AzureAVATextToSpeechConfig(BaseTextToSpeechConfig):
region: Final = self._extract_region_from_hostname(hostname=hostname, domain=self.COGNITIVE_SERVICES_DOMAIN)
return self._build_tts_url(region=region)
- # Check if it's already a TTS endpoint
if self._is_tts_endpoint(hostname=hostname):
- if not api_base.endswith(self.TTS_ENDPOINT_PATH):
- return f"{api_base}{self.TTS_ENDPOINT_PATH}"
- return api_base
+ return ensure_api_route(api_base, self.TTS_ENDPOINT_PATH)
- # Assume it's a custom endpoint, append the path
- return f"{api_base}{self.TTS_ENDPOINT_PATH}"
+ return join_api_base(api_base, self.TTS_ENDPOINT_PATH)
def _is_cognitive_services_endpoint(self, hostname: str) -> bool:
"""Check if hostname is a Cognitive Services endpoint"""
diff --git a/litellm/llms/azure_ai/agents/handler.py b/litellm/llms/azure_ai/agents/handler.py
index 0b0548dd3a2..94f1399f01b 100644
--- a/litellm/llms/azure_ai/agents/handler.py
+++ b/litellm/llms/azure_ai/agents/handler.py
@@ -30,7 +30,7 @@ from typing_extensions import ReadOnly
from litellm._logging import verbose_logger
from litellm.litellm_core_utils.hidden_params import set_hidden_param
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.azure_ai.agents.transformation import (
AzureAIAgentsConfig,
AzureAIAgentsError,
@@ -130,28 +130,29 @@ class AzureAIAgentsHandler:
# See: https://learn.microsoft.com/en-us/azure/ai-foundry/agents/quickstart
# -------------------------------------------------------------------------
def _build_thread_url(self, api_base: str, api_version: str) -> str:
- return f"{api_base}/threads?api-version={api_version}"
+ return f"{join_api_base(api_base, 'threads')}?api-version={api_version}"
def _build_messages_url(self, api_base: str, thread_id: str, api_version: str) -> str:
encoded_thread_id: Final = encode_url_path_segment(thread_id, field_name="thread_id")
- return f"{api_base}/threads/{encoded_thread_id}/messages?api-version={api_version}"
+ return f"{join_api_base(api_base, f'threads/{encoded_thread_id}/messages')}?api-version={api_version}"
def _build_runs_url(self, api_base: str, thread_id: str, api_version: str) -> str:
encoded_thread_id: Final = encode_url_path_segment(thread_id, field_name="thread_id")
- return f"{api_base}/threads/{encoded_thread_id}/runs?api-version={api_version}"
+ return f"{join_api_base(api_base, f'threads/{encoded_thread_id}/runs')}?api-version={api_version}"
def _build_run_status_url(self, api_base: str, thread_id: str, run_id: str, api_version: str) -> str:
encoded_thread_id: Final = encode_url_path_segment(thread_id, field_name="thread_id")
encoded_run_id: Final = encode_url_path_segment(run_id, field_name="run_id")
- return f"{api_base}/threads/{encoded_thread_id}/runs/{encoded_run_id}?api-version={api_version}"
+ run_status_route: Final = f"threads/{encoded_thread_id}/runs/{encoded_run_id}"
+ return f"{join_api_base(api_base, run_status_route)}?api-version={api_version}"
def _build_list_messages_url(self, api_base: str, thread_id: str, api_version: str) -> str:
encoded_thread_id: Final = encode_url_path_segment(thread_id, field_name="thread_id")
- return f"{api_base}/threads/{encoded_thread_id}/messages?api-version={api_version}"
+ return f"{join_api_base(api_base, f'threads/{encoded_thread_id}/messages')}?api-version={api_version}"
def _build_create_thread_and_run_url(self, api_base: str, api_version: str) -> str:
"""URL for the create-thread-and-run endpoint (supports streaming)."""
- return f"{api_base}/threads/runs?api-version={api_version}"
+ return f"{join_api_base(api_base, 'threads/runs')}?api-version={api_version}"
# -------------------------------------------------------------------------
# Response Helpers
@@ -281,8 +282,6 @@ class AzureAIAgentsHandler:
api_version: Final = optional_params.get("api_version", self.config.DEFAULT_API_VERSION)
agent_id: Final = self.config.get_agent_id(model, optional_params)
thread_id: Final = optional_params.get("thread_id")
- api_base = api_base.rstrip("/")
-
verbose_logger.debug("Azure AI Agents completion - api_base: %s, agent_id: %s", api_base, agent_id)
return headers, api_version, agent_id, thread_id, api_base
diff --git a/litellm/llms/azure_ai/anthropic/count_tokens/transformation.py b/litellm/llms/azure_ai/anthropic/count_tokens/transformation.py
index 5946a1f4480..6e3cfce5eb4 100644
--- a/litellm/llms/azure_ai/anthropic/count_tokens/transformation.py
+++ b/litellm/llms/azure_ai/anthropic/count_tokens/transformation.py
@@ -5,8 +5,10 @@ Extends the base Anthropic CountTokens transformation with Azure authentication.
"""
from typing import Final
+from urllib.parse import urlsplit
from litellm.constants import ANTHROPIC_TOKEN_COUNTING_BETA_VERSION
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.anthropic.count_tokens.transformation import (
AnthropicCountTokensConfig,
)
@@ -76,14 +78,8 @@ class AzureAIAnthropicCountTokensConfig(AnthropicCountTokensConfig):
Returns:
The endpoint URL for the CountTokens API
"""
- # Azure AI Anthropic endpoint format:
- # https://<resource>.services.ai.azure.com/anthropic/v1/messages/count_tokens
- api_base = api_base.rstrip("/")
-
- # Ensure the URL has /anthropic path
- if not api_base.endswith("/anthropic"):
- if "/anthropic" not in api_base:
- api_base = f"{api_base}/anthropic"
-
- # Add the count_tokens path
- return f"{api_base}/v1/messages/count_tokens"
+ base: Final = api_base.rstrip("/")
+ if base.endswith("/v1/messages"):
+ return join_api_base(base, "count_tokens")
+ anthropic_base: Final = base if "/anthropic" in urlsplit(base).path else join_api_base(base, "anthropic")
+ return join_api_base(anthropic_base, "v1/messages/count_tokens")
diff --git a/litellm/llms/azure_ai/anthropic/messages_transformation.py b/litellm/llms/azure_ai/anthropic/messages_transformation.py
index 91c75661688..2aacc5771d0 100644
--- a/litellm/llms/azure_ai/anthropic/messages_transformation.py
+++ b/litellm/llms/azure_ai/anthropic/messages_transformation.py
@@ -9,6 +9,7 @@ from litellm.llms.anthropic.pass_through.messages.transformation import (
AnthropicMessagesConfig,
)
from litellm.llms.azure.common_utils import BaseAzureLLM
+from litellm.llms.azure_ai.anthropic.transformation import azure_anthropic_messages_url
from litellm.types.router import GenericLiteLLMParams
@@ -96,25 +97,7 @@ class AzureAnthropicMessagesConfig(AnthropicMessagesConfig):
"Expected format: https://<resource-name>.services.ai.azure.com/anthropic"
)
- # Ensure the URL ends with /v1/messages
- api_base = api_base.rstrip("/")
- if api_base.endswith("/v1/messages") or api_base.endswith("/anthropic/v1/messages"):
- # Already correct
- pass
- else:
- # Check if /anthropic is already in the path
- if "/anthropic" in api_base:
- # /anthropic exists, ensure we end with /anthropic/v1/messages
- # Extract the base URL up to and including /anthropic
- parts: Final = api_base.split("/anthropic", 1)
- api_base = parts[0] + "/anthropic"
- else:
- # /anthropic not in path, add it
- api_base = api_base + "/anthropic"
- # Add /v1/messages
- api_base = api_base + "/v1/messages"
-
- return api_base
+ return azure_anthropic_messages_url(api_base)
def _remove_scope_from_cache_control(self, anthropic_messages_request: dict) -> None:
"""
diff --git a/litellm/llms/azure_ai/anthropic/transformation.py b/litellm/llms/azure_ai/anthropic/transformation.py
index 18a4eeb7c0a..c75213c3e2c 100644
--- a/litellm/llms/azure_ai/anthropic/transformation.py
+++ b/litellm/llms/azure_ai/anthropic/transformation.py
@@ -4,12 +4,23 @@ Azure Anthropic transformation config - extends AnthropicConfig with Azure authe
from typing import Final
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.anthropic.chat.transformation import AnthropicConfig
from litellm.llms.azure.common_utils import BaseAzureLLM
from litellm.types.llms.openai import AllMessageValues
from litellm.types.router import GenericLiteLLMParams
+def azure_anthropic_messages_url(api_base: str) -> str:
+ base: Final = api_base.rstrip("/")
+ if base.endswith("/v1/messages"):
+ return base
+ path_start: Final = base.find("/", base.find("://") + len("://")) if "://" in base else 0
+ anthropic_start: Final = base.find("/anthropic", path_start) if path_start >= 0 else -1
+ anthropic_root: Final = base[:anthropic_start] if anthropic_start >= 0 else base
+ return join_api_base(anthropic_root, "anthropic/v1/messages")
+
+
def _promote_extra_body_to_optional_params(optional_params: dict) -> None:
"""Promote anthropic-native passthrough keys out of ``extra_body``.
diff --git a/litellm/llms/azure_ai/embed/handler.py b/litellm/llms/azure_ai/embed/handler.py
index 57e6ca51ade..dc3bc6bdc5e 100644
--- a/litellm/llms/azure_ai/embed/handler.py
+++ b/litellm/llms/azure_ai/embed/handler.py
@@ -5,6 +5,7 @@ from urllib.parse import urlsplit, urlunsplit
from openai import OpenAI
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.azure_ai.common_utils import is_foundry_model_inference_base
from litellm.llms.custom_httpx.http_handler import (
AsyncHTTPHandler,
@@ -74,13 +75,18 @@ class AzureAIEmbedding(OpenAIChatCompletion):
api_base: str | None,
client: HTTPHandler | AsyncHTTPHandler | None = None,
) -> EmbeddingResponse:
+ if api_base is None:
+ raise ValueError(
+ "api_base is None. Please set AZURE_AI_API_BASE or dynamically via `api_base` param, "
+ "to make the request."
+ )
if client is None or not isinstance(client, AsyncHTTPHandler):
client = get_async_httpx_client(
llm_provider=litellm.LlmProviders.AZURE_AI,
params={"timeout": timeout},
)
- url: Final = f"{api_base}/images/embeddings"
+ url: Final = join_api_base(api_base, "images/embeddings")
response: Final = await client.post(
url=url,
@@ -123,7 +129,7 @@ class AzureAIEmbedding(OpenAIChatCompletion):
if client is None or not isinstance(client, HTTPHandler):
client = HTTPHandler(timeout=timeout, concurrent_limit=1)
- url: Final = f"{api_base}/images/embeddings"
+ url: Final = join_api_base(api_base, "images/embeddings")
response: Final = client.post(
url=url,
diff --git a/litellm/llms/azure_ai/image_generation/flux_transformation.py b/litellm/llms/azure_ai/image_generation/flux_transformation.py
index 8b4fa78cdf4..6a7bebd9f6b 100644
--- a/litellm/llms/azure_ai/image_generation/flux_transformation.py
+++ b/litellm/llms/azure_ai/image_generation/flux_transformation.py
@@ -3,6 +3,7 @@ from types import MappingProxyType
from typing import Final
from litellm.exceptions import BadRequestError, UnsupportedParamsError
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.image_generation import GPTImageGenerationConfig
from litellm.types.llms.openai import OpenAIImageGenerationOptionalParams
@@ -52,7 +53,8 @@ class AzureFoundryFluxImageGenerationConfig(GPTImageGenerationConfig):
return f"{api_base}?api-version={api_version}"
provider_model_path: Final = AzureFoundryFluxImageGenerationConfig.get_flux2_provider_model_path(model)
- return f"{api_base}/providers/blackforestlabs/v1/{provider_model_path}?api-version={api_version}"
+ flux_route: Final = f"providers/blackforestlabs/v1/{provider_model_path}"
+ return f"{join_api_base(api_base, flux_route)}?api-version={api_version}"
@staticmethod
def is_flux2_model(model: str) -> bool:
diff --git a/litellm/llms/azure_ai/image_generation/mai_transformation.py b/litellm/llms/azure_ai/image_generation/mai_transformation.py
index 18b4b6f456a..8bb5d3e6b8d 100644
--- a/litellm/llms/azure_ai/image_generation/mai_transformation.py
+++ b/litellm/llms/azure_ai/image_generation/mai_transformation.py
@@ -3,6 +3,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
from litellm.exceptions import UnsupportedParamsError
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
)
@@ -41,9 +42,9 @@ class AzureFoundryMAIImageGenerationConfig(BaseImageGenerationConfig):
if "/mai/" in path:
prefix, _, _ = path.partition("/images/")
- path = f"{prefix}/images/generations"
+ path = join_api_base(prefix, "images/generations")
else:
- path = f"{path}/mai/v1/images/generations"
+ path = join_api_base(path, "mai/v1/images/generations")
if separator:
return f"{path}?{query}"
@@ -63,9 +64,9 @@ class AzureFoundryMAIImageGenerationConfig(BaseImageGenerationConfig):
if "/mai/" in path:
prefix, _, _ = path.partition("/images/")
- path = f"{prefix}/images/edits"
+ path = join_api_base(prefix, "images/edits")
else:
- path = f"{path}/mai/v1/images/edits"
+ path = join_api_base(path, "mai/v1/images/edits")
if separator:
return f"{path}?{query}"
diff --git a/litellm/llms/azure_ai/vector_stores/transformation.py b/litellm/llms/azure_ai/vector_stores/transformation.py
index 7638229bc32..a4777be146f 100644
--- a/litellm/llms/azure_ai/vector_stores/transformation.py
+++ b/litellm/llms/azure_ai/vector_stores/transformation.py
@@ -5,6 +5,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.azure.common_utils import BaseAzureLLM
from litellm.llms.base_llm.vector_store.transformation import (
BaseQueryEmbeddingVectorStoreConfig,
@@ -172,7 +173,8 @@ class AzureAIVectorStoreConfig(BaseQueryEmbeddingVectorStoreConfig, BaseAzureLLM
litellm_logging_obj.model_call_details["input"] = query_text
litellm_logging_obj.model_call_details["embedding_model"] = litellm_params.get("litellm_embedding_model")
litellm_logging_obj.model_call_details["top_k"] = top_k
- return f"{api_base}/indexes/{vector_store_id}/docs/search?api-version=2024-07-01", {
+ search_route: Final = f"indexes/{vector_store_id}/docs/search"
+ return f"{join_api_base(api_base, search_route)}?api-version=2024-07-01", {
"search": "*",
"vectorQueries": [{"vector": query_vector, "fields": vector_field, "kind": "vector", "k": top_k}],
"select": "id,content",
diff --git a/litellm/llms/base_llm/decisions/transformation.py b/litellm/llms/base_llm/decisions/transformation.py
index d74257a4a9d..a122aeb3033 100644
--- a/litellm/llms/base_llm/decisions/transformation.py
+++ b/litellm/llms/base_llm/decisions/transformation.py
@@ -12,6 +12,7 @@ from pydantic_core import ErrorDetails
from typing_extensions import assert_never
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.secret_managers.main import get_secret_str
from litellm.types.decisions import (
@@ -582,7 +583,7 @@ class BaseDecisionsConfig(ABC):
)
def get_complete_url(self, api_base: str, model: str) -> str:
- return f"{api_base.rstrip('/').removesuffix('/v1')}{self.path}"
+ return join_api_base(api_base.rstrip("/").removesuffix("/v1"), self.path)
def signs_with_aws(self, api_base: str) -> bool:
return False
diff --git a/litellm/llms/base_llm/evals/transformation.py b/litellm/llms/base_llm/evals/transformation.py
index 49124e271a6..40093953144 100644
--- a/litellm/llms/base_llm/evals/transformation.py
+++ b/litellm/llms/base_llm/evals/transformation.py
@@ -7,6 +7,7 @@ from typing import TYPE_CHECKING, Any
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.types.llms.openai_evals import (
CancelEvalResponse,
@@ -79,7 +80,7 @@ class BaseEvalsAPIConfig(ABC):
"""
if api_base is None:
raise ValueError("api_base is required")
- return f"{api_base}/v1/{endpoint}"
+ return join_api_base(api_base, f"v1/{endpoint}")
@abstractmethod
def transform_create_eval_request(
diff --git a/litellm/llms/base_llm/realtime/http_transformation.py b/litellm/llms/base_llm/realtime/http_transformation.py
index 43a80edb493..a02054ec8cb 100644
--- a/litellm/llms/base_llm/realtime/http_transformation.py
+++ b/litellm/llms/base_llm/realtime/http_transformation.py
@@ -7,10 +7,11 @@ These are HTTP (not WebSocket) endpoints used by the WebRTC flow:
"""
from abc import ABC, abstractmethod
-from typing import Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
+
class BaseRealtimeHTTPConfig(ABC):
"""
@@ -61,7 +62,7 @@ class BaseRealtimeHTTPConfig(ABC):
"""Return the full URL for POST /realtime/transcription_sessions."""
base = (api_base or "").rstrip("/")
base = base.removesuffix("/v1")
- return f"{base}/v1/realtime/transcription_sessions"
+ return join_api_base(base, "v1/realtime/transcription_sessions")
@abstractmethod
def validate_environment(
@@ -83,8 +84,7 @@ class BaseRealtimeHTTPConfig(ABC):
def get_realtime_calls_url(self, api_base: str | None, model: str, api_version: str | None = None) -> str:
"""Return the full URL for POST /realtime/calls (SDP exchange)."""
- base: Final = (api_base or "").rstrip("/")
- return f"{base}/v1/realtime/calls"
+ return join_api_base(api_base or "", "v1/realtime/calls")
def get_realtime_calls_headers(self, ephemeral_key: str) -> dict:
"""
diff --git a/litellm/llms/base_llm/skills/transformation.py b/litellm/llms/base_llm/skills/transformation.py
index ba202e3db5d..64a8e055832 100644
--- a/litellm/llms/base_llm/skills/transformation.py
+++ b/litellm/llms/base_llm/skills/transformation.py
@@ -7,6 +7,7 @@ from typing import TYPE_CHECKING, Any
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.types.llms.anthropic_skills import (
CreateSkillRequest,
@@ -71,7 +72,7 @@ class BaseSkillsAPIConfig(ABC):
"""
if api_base is None:
raise ValueError("api_base is required")
- return f"{api_base}/v1/{endpoint}"
+ return join_api_base(api_base, f"v1/{endpoint}")
@abstractmethod
def transform_create_skill_request(
diff --git a/litellm/llms/bedrock/chat/agentcore/transformation.py b/litellm/llms/bedrock/chat/agentcore/transformation.py
index 48b3795e274..bb5bb246a93 100644
--- a/litellm/llms/bedrock/chat/agentcore/transformation.py
+++ b/litellm/llms/bedrock/chat/agentcore/transformation.py
@@ -18,6 +18,7 @@ from litellm.litellm_core_utils.prompt_templates.common_utils import (
convert_content_list_to_str,
)
from litellm.litellm_core_utils.streaming_handler import CustomStreamWrapper
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.a2a.common_utils import extract_text_from_a2a_response
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM
@@ -98,14 +99,14 @@ class AmazonAgentCoreConfig(BaseConfig, BaseAWSLLM):
# Build the base endpoint URL for AgentCore
# Note: We don't use get_runtime_endpoint as AgentCore has its own endpoint structure
if aws_bedrock_runtime_endpoint:
- base_url = aws_bedrock_runtime_endpoint
+ base_url = f"{aws_bedrock_runtime_endpoint}"
else:
base_url = f"https://bedrock-agentcore.{region}.{get_aws_dns_suffix(region)}"
# Based on boto3 client.invoke_agent_runtime, the path is:
# /runtimes/{URL-ENCODED-ARN}/invocations?qualifier=<value>
encoded_arn: Final = quote(agent_runtime_arn, safe="")
- endpoint_url = f"{base_url}/runtimes/{encoded_arn}/invocations"
+ endpoint_url = join_api_base(base_url, f"runtimes/{encoded_arn}/invocations")
# Add qualifier as query parameter if provided
if "qualifier" in optional_params:
diff --git a/litellm/llms/bedrock/chat/chat_completions/transformation.py b/litellm/llms/bedrock/chat/chat_completions/transformation.py
index 907f8e50ac7..123e4a9e6b4 100644
--- a/litellm/llms/bedrock/chat/chat_completions/transformation.py
+++ b/litellm/llms/bedrock/chat/chat_completions/transformation.py
@@ -33,6 +33,7 @@ from litellm.litellm_core_utils.prompt_templates.image_handling import (
inline_remote_image_urls,
inline_remote_media,
)
+from litellm.litellm_core_utils.url_utils import ensure_api_route, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM, bedrock_bearer_token
from litellm.llms.bedrock.common_utils import (
@@ -371,12 +372,9 @@ class AmazonBedrockRuntimeChatCompletionsConfig(OpenAILikeChatConfig):
),
aws_region_name=aws_region_name,
)
- base: Final = proxy_endpoint_url.rstrip("/")
- if base.endswith("/openai/v1/chat/completions"):
- return base
- if base.endswith("/openai/v1"):
- return f"{base}/chat/completions"
- return f"{base}/openai/v1/chat/completions"
+ if proxy_endpoint_url.rstrip("/").endswith("/openai/v1"):
+ return join_api_base(proxy_endpoint_url, "chat/completions")
+ return ensure_api_route(proxy_endpoint_url, "openai/v1/chat/completions")
def _params_with_region_from_path(
self, optional_params: dict, model: str | None
diff --git a/litellm/llms/bedrock/chat/converse_handler.py b/litellm/llms/bedrock/chat/converse_handler.py
index 7d02b9ef230..a6f9192339f 100644
--- a/litellm/llms/bedrock/chat/converse_handler.py
+++ b/litellm/llms/bedrock/chat/converse_handler.py
@@ -10,6 +10,7 @@ from litellm.anthropic_beta_headers_manager import (
)
from litellm.litellm_core_utils.get_litellm_params import stored_control_options
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObject
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm import base_model_iterator
from litellm.llms.custom_httpx.http_handler import (
AsyncHTTPHandler,
@@ -363,11 +364,11 @@ class BedrockConverseLLM(BaseAWSLLM):
aws_region_name=aws_region_name,
)
if (stream is not None and stream is True) and not fake_stream:
- endpoint_url = f"{endpoint_url}/model/{modelId}/converse-stream"
- proxy_endpoint_url = f"{proxy_endpoint_url}/model/{modelId}/converse-stream"
+ endpoint_url = join_api_base(endpoint_url, f"model/{modelId}/converse-stream")
+ proxy_endpoint_url = join_api_base(proxy_endpoint_url, f"model/{modelId}/converse-stream")
else:
- endpoint_url = f"{endpoint_url}/model/{modelId}/converse"
- proxy_endpoint_url = f"{proxy_endpoint_url}/model/{modelId}/converse"
+ endpoint_url = join_api_base(endpoint_url, f"model/{modelId}/converse")
+ proxy_endpoint_url = join_api_base(proxy_endpoint_url, f"model/{modelId}/converse")
## COMPLETION CALL
headers = {"Content-Type": "application/json"}
diff --git a/litellm/llms/bedrock/chat/invoke_agent/transformation.py b/litellm/llms/bedrock/chat/invoke_agent/transformation.py
index 6877af74494..baef89912fa 100644
--- a/litellm/llms/bedrock/chat/invoke_agent/transformation.py
+++ b/litellm/llms/bedrock/chat/invoke_agent/transformation.py
@@ -15,7 +15,7 @@ from litellm._uuid import uuid
from litellm.litellm_core_utils.prompt_templates.common_utils import (
convert_content_list_to_str,
)
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM
from litellm.llms.bedrock.common_utils import BedrockError
@@ -101,9 +101,10 @@ class AmazonInvokeAgentConfig(BaseConfig, BaseAWSLLM):
encoded_agent_alias_id: Final = encode_url_path_segment(agent_alias_id, field_name="agent_alias_id")
encoded_session_id: Final = encode_url_path_segment(session_id, field_name="session_id")
- endpoint_url = f"{endpoint_url}/agents/{encoded_agent_id}/agentAliases/{encoded_agent_alias_id}/sessions/{encoded_session_id}/text"
-
- return endpoint_url
+ return join_api_base(
+ endpoint_url,
+ f"agents/{encoded_agent_id}/agentAliases/{encoded_agent_alias_id}/sessions/{encoded_session_id}/text",
+ )
def sign_request(
self,
diff --git a/litellm/llms/bedrock/chat/invoke_transformations/amazon_openai_transformation.py b/litellm/llms/bedrock/chat/invoke_transformations/amazon_openai_transformation.py
index 4bf1a1cba73..f1e1f3e8738 100644
--- a/litellm/llms/bedrock/chat/invoke_transformations/amazon_openai_transformation.py
+++ b/litellm/llms/bedrock/chat/invoke_transformations/amazon_openai_transformation.py
@@ -11,6 +11,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM
from litellm.llms.bedrock.common_utils import BedrockError
from litellm.llms.bedrock.request_metadata import (
@@ -97,13 +98,8 @@ class AmazonBedrockOpenAIConfig(OpenAIGPTConfig, BaseAWSLLM):
# Encode model ID for ARNs (e.g., :imported-model/ -> :imported-model%2F)
model_id = CommonUtils.encode_bedrock_runtime_modelid_arn(model_id)
- # Build the invoke URL
- if stream:
- endpoint_url = f"{endpoint_url}/model/{model_id}/invoke-with-response-stream"
- else:
- endpoint_url = f"{endpoint_url}/model/{model_id}/invoke"
-
- return endpoint_url
+ invoke_route: Final = "invoke-with-response-stream" if stream else "invoke"
+ return join_api_base(endpoint_url, f"model/{model_id}/{invoke_route}")
def sign_request(
self,
diff --git a/litellm/llms/bedrock/chat/invoke_transformations/base_invoke_transformation.py b/litellm/llms/bedrock/chat/invoke_transformations/base_invoke_transformation.py
index 87d4fc0b32e..51b5a99272e 100644
--- a/litellm/llms/bedrock/chat/invoke_transformations/base_invoke_transformation.py
+++ b/litellm/llms/bedrock/chat/invoke_transformations/base_invoke_transformation.py
@@ -18,6 +18,7 @@ from litellm.litellm_core_utils.prompt_templates.factory import (
prompt_factory,
render_custom_prompt_template,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.llms.bedrock.chat.invoke_handler import make_call, make_sync_call
from litellm.llms.bedrock.common_utils import BedrockError
@@ -136,20 +137,13 @@ class AmazonInvokeConfig(BaseConfig, BaseAWSLLM):
aws_bedrock_runtime_endpoint: Final = optional_params.get(
"aws_bedrock_runtime_endpoint", None
) # https://bedrock-runtime.{region_name}.amazonaws.com
- endpoint_url, proxy_endpoint_url = self.get_runtime_endpoint(
+ endpoint_url, _ = self.get_runtime_endpoint(
api_base=api_base,
aws_bedrock_runtime_endpoint=aws_bedrock_runtime_endpoint,
aws_region_name=self._get_aws_region_name(optional_params=optional_params, model=model),
)
-
- if (stream is not None and stream is True) and provider != "ai21":
- endpoint_url = f"{endpoint_url}/model/{modelId}/invoke-with-response-stream"
- proxy_endpoint_url = f"{proxy_endpoint_url}/model/{modelId}/invoke-with-response-stream"
- else:
- endpoint_url = f"{endpoint_url}/model/{modelId}/invoke"
- proxy_endpoint_url = f"{proxy_endpoint_url}/model/{modelId}/invoke"
-
- return endpoint_url
+ invoke_route: Final = "invoke-with-response-stream" if stream is True and provider != "ai21" else "invoke"
+ return join_api_base(endpoint_url, f"model/{modelId}/{invoke_route}")
def sign_request(
self,
diff --git a/litellm/llms/bedrock/claude_platform/common_utils.py b/litellm/llms/bedrock/claude_platform/common_utils.py
index e25e58056f5..77cae177792 100644
--- a/litellm/llms/bedrock/claude_platform/common_utils.py
+++ b/litellm/llms/bedrock/claude_platform/common_utils.py
@@ -5,6 +5,7 @@ import httpx
import litellm
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM
from litellm.llms.bedrock.common_utils import BedrockError
from litellm.secret_managers.main import get_secret_str
@@ -140,9 +141,7 @@ class BedrockClaudePlatformMixin(BaseAWSLLM):
if api_base is None:
aws_region_name: Final = self._get_required_aws_region_name(optional_params)
api_base = f"https://{CLAUDE_PLATFORM_SERVICE_NAME}.{aws_region_name}.api.aws"
- if not api_base.endswith("/v1/messages"):
- api_base = f"{api_base.rstrip('/')}/v1/messages"
- return api_base
+ return ensure_api_route(api_base, "v1/messages")
def sign_request(
self,
diff --git a/litellm/llms/bedrock/common_utils.py b/litellm/llms/bedrock/common_utils.py
index 0cf3e895ad2..4954e646f58 100644
--- a/litellm/llms/bedrock/common_utils.py
+++ b/litellm/llms/bedrock/common_utils.py
@@ -31,6 +31,7 @@ from litellm import verbose_logger
from litellm.litellm_core_utils.aws_partition import get_aws_dns_suffix
from litellm.litellm_core_utils.optional_imports import ensure_optional_import
from litellm.litellm_core_utils.prompt_templates.common_utils import schema_property_count
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.anthropic_messages.transformation import (
BaseAnthropicMessagesConfig,
)
@@ -1305,10 +1306,7 @@ def build_mantle_messages_url(
"""
override: Final = api_base or aws_bedrock_runtime_endpoint or _mantle_api_base_from_env()
if override:
- base: Final = override.rstrip("/")
- if base.endswith(MANTLE_MESSAGES_PATH):
- return base
- return f"{base}{MANTLE_MESSAGES_PATH}"
+ return ensure_api_route(override, MANTLE_MESSAGES_PATH)
return f"https://bedrock-mantle.{region}.api.aws{MANTLE_MESSAGES_PATH}"
diff --git a/litellm/llms/bedrock/count_tokens/mantle_handler.py b/litellm/llms/bedrock/count_tokens/mantle_handler.py
index 14edd39c386..34e622b5e8d 100644
--- a/litellm/llms/bedrock/count_tokens/mantle_handler.py
+++ b/litellm/llms/bedrock/count_tokens/mantle_handler.py
@@ -7,6 +7,7 @@ from typing_extensions import NotRequired, ReadOnly, TypedDict
import litellm
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.anthropic.count_tokens.transformation import COUNT_TOKEN_OPTION_NAMES, AnthropicCountTokensConfig
from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM, run_aws_signing
from litellm.llms.bedrock.common_utils import BedrockError, build_mantle_messages_url
@@ -43,7 +44,7 @@ class BedrockMantleCountTokensHandler(BaseAWSLLM):
messages_url: Final = build_mantle_messages_url(
api_base=None, aws_bedrock_runtime_endpoint=None, region=aws_region_name
)
- return f"{messages_url}{MANTLE_COUNT_TOKENS_SUFFIX}"
+ return join_api_base(messages_url, MANTLE_COUNT_TOKENS_SUFFIX)
async def handle_count_tokens_request(
self,
diff --git a/litellm/llms/bedrock/count_tokens/transformation.py b/litellm/llms/bedrock/count_tokens/transformation.py
index 7cca4f69acc..a9f95c1f9b9 100644
--- a/litellm/llms/bedrock/count_tokens/transformation.py
+++ b/litellm/llms/bedrock/count_tokens/transformation.py
@@ -11,6 +11,7 @@ from typing import Final, Literal
from pydantic import JsonValue
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM
from litellm.llms.bedrock.common_utils import get_bedrock_base_model
@@ -236,9 +237,7 @@ class BedrockCountTokensConfig(BaseAWSLLM):
aws_bedrock_runtime_endpoint=aws_bedrock_runtime_endpoint,
aws_region_name=aws_region_name,
)
- endpoint: Final = f"{base_url}/model/{encoded_model_id}/count-tokens"
-
- return endpoint
+ return join_api_base(base_url, f"model/{encoded_model_id}/count-tokens")
def transform_bedrock_response_to_anthropic(
self, bedrock_response: Mapping[str, JsonValue]
diff --git a/litellm/llms/bedrock/embed/embedding.py b/litellm/llms/bedrock/embed/embedding.py
index 58cfb6cb95e..4d27d94664f 100644
--- a/litellm/llms/bedrock/embed/embedding.py
+++ b/litellm/llms/bedrock/embed/embedding.py
@@ -12,6 +12,7 @@ import httpx
import litellm
from litellm.constants import BEDROCK_EMBEDDING_PROVIDERS_LITERAL
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.cohere.embed.handler import embedding as cohere_embedding
from litellm.llms.custom_httpx.http_handler import (
AsyncHTTPHandler,
@@ -512,9 +513,9 @@ class BedrockEmbedding(BaseAWSLLM):
aws_region_name=aws_region_name,
)
if has_async_invoke:
- endpoint_url = f"{endpoint_url}/async-invoke"
+ endpoint_url = join_api_base(endpoint_url, "async-invoke")
else:
- endpoint_url = f"{endpoint_url}/model/{modelId}/invoke"
+ endpoint_url = join_api_base(endpoint_url, f"model/{modelId}/invoke")
if batch_data is not None:
if aembedding:
@@ -611,7 +612,7 @@ class BedrockEmbedding(BaseAWSLLM):
# Encode the ARN for use in URL path
encoded_arn: Final = quote(invocation_arn, safe="")
- status_url: Final = f"{endpoint_url.rstrip('/')}/async-invoke/{encoded_arn}"
+ status_url: Final = join_api_base(endpoint_url, f"async-invoke/{encoded_arn}")
# Prepare headers for GET request
headers: Final = {"Content-Type": "application/json"}
diff --git a/litellm/llms/bedrock/image_edit/handler.py b/litellm/llms/bedrock/image_edit/handler.py
index 53bb332f511..466993de9d1 100644
--- a/litellm/llms/bedrock/image_edit/handler.py
+++ b/litellm/llms/bedrock/image_edit/handler.py
@@ -14,6 +14,7 @@ import httpx
import litellm
from litellm._logging import verbose_logger
from litellm.litellm_core_utils.litellm_logging import Logging as LitellmLogging
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.bedrock.image_edit.amazon_nova_canvas_image_edit_transformation import (
BedrockAmazonNovaCanvasImageEditConfig,
)
@@ -226,7 +227,7 @@ class BedrockImageEdit(BaseAWSLLM):
aws_bedrock_runtime_endpoint=boto3_credentials_info.aws_bedrock_runtime_endpoint,
aws_region_name=boto3_credentials_info.aws_region_name,
)
- proxy_endpoint_url = f"{proxy_endpoint_url}/model/{modelId}/invoke"
+ proxy_endpoint_url = join_api_base(proxy_endpoint_url, f"model/{modelId}/invoke")
data: Final = self._get_request_body(
model=model,
image=image,
diff --git a/litellm/llms/bedrock/image_generation/image_handler.py b/litellm/llms/bedrock/image_generation/image_handler.py
index e81dfe76942..abd6d40f9e9 100644
--- a/litellm/llms/bedrock/image_generation/image_handler.py
+++ b/litellm/llms/bedrock/image_generation/image_handler.py
@@ -8,6 +8,7 @@ import httpx
import litellm
from litellm._logging import verbose_logger
from litellm.litellm_core_utils.litellm_logging import Logging as LitellmLogging
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.bedrock.image_generation.amazon_nova_canvas_transformation import (
AmazonNovaCanvasConfig,
)
@@ -248,7 +249,7 @@ class BedrockImageGeneration(BaseAWSLLM):
aws_bedrock_runtime_endpoint=boto3_credentials_info.aws_bedrock_runtime_endpoint,
aws_region_name=boto3_credentials_info.aws_region_name,
)
- proxy_endpoint_url = f"{proxy_endpoint_url}/model/{modelId}/invoke"
+ proxy_endpoint_url = join_api_base(proxy_endpoint_url, f"model/{modelId}/invoke")
data: Final = self._get_request_body(
model=model,
prompt=prompt,
diff --git a/litellm/llms/bedrock/rerank/handler.py b/litellm/llms/bedrock/rerank/handler.py
index 01ef5af2b67..0634275135b 100644
--- a/litellm/llms/bedrock/rerank/handler.py
+++ b/litellm/llms/bedrock/rerank/handler.py
@@ -6,6 +6,7 @@ from pydantic import ConfigDict, TypeAdapter
import litellm
from litellm.litellm_core_utils.litellm_logging import Logging as LitellmLogging
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.custom_httpx.http_handler import (
AsyncHTTPHandler,
HTTPHandler,
@@ -160,7 +161,7 @@ class BedrockRerankHandler(BaseAWSLLM):
aws_region_name=boto3_credentials_info.aws_region_name,
)
proxy_endpoint_url = proxy_endpoint_url.replace("bedrock-runtime", "bedrock-agent-runtime")
- proxy_endpoint_url = f"{proxy_endpoint_url}/rerank"
+ proxy_endpoint_url = join_api_base(proxy_endpoint_url, "rerank")
body: Final = json.dumps(data).encode("utf-8")
headers = {"Content-Type": "application/json"}
diff --git a/litellm/llms/bedrock/responses/transformation.py b/litellm/llms/bedrock/responses/transformation.py
index 086d1211835..7cdd7b08f99 100644
--- a/litellm/llms/bedrock/responses/transformation.py
+++ b/litellm/llms/bedrock/responses/transformation.py
@@ -46,6 +46,7 @@ from litellm.litellm_core_utils.prompt_templates.image_handling import (
async_convert_url_to_base64,
convert_url_to_base64,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.responses.codex_compat import drop_unsupported_tools, normalize_codex_input_items
from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM
@@ -223,9 +224,10 @@ class BedrockOpenAIResponsesConfig(BaseAWSLLM, OpenAIResponsesAPIConfig):
litellm_params: dict, # mutable-ok: signature fixed by the BaseResponsesAPIConfig override contract
) -> str:
region: Final = self._get_aws_region_name(optional_params=litellm_params, model=None)
+ configured_endpoint: Final = litellm_params.get("aws_bedrock_runtime_endpoint")
override: Final = (
api_base
- or litellm_params.get("aws_bedrock_runtime_endpoint")
+ or (f"{configured_endpoint}" if configured_endpoint else None)
or get_secret_str("AWS_BEDROCK_RUNTIME_ENDPOINT")
)
# Partition-aware: bedrock-runtime is amazonaws.com.cn in China, and other
@@ -237,7 +239,7 @@ class BedrockOpenAIResponsesConfig(BaseAWSLLM, OpenAIResponsesAPIConfig):
(host[: -len(suffix)] for suffix in BEDROCK_RUNTIME_OPENAI_BASE_SUFFIXES if host.endswith(suffix)),
host,
)
- return f"{base}{BEDROCK_RUNTIME_OPENAI_RESPONSES_PATH}"
+ return join_api_base(base, BEDROCK_RUNTIME_OPENAI_RESPONSES_PATH)
def supports_native_file_search(self) -> bool:
return False
diff --git a/litellm/llms/bedrock/vector_stores/transformation.py b/litellm/llms/bedrock/vector_stores/transformation.py
index ba8ce7e5625..96d1da19bcc 100644
--- a/litellm/llms/bedrock/vector_stores/transformation.py
+++ b/litellm/llms/bedrock/vector_stores/transformation.py
@@ -6,7 +6,7 @@ from urllib.parse import urlparse
import httpx
from litellm._logging import verbose_logger
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.vector_store.transformation import BaseVectorStoreConfig
from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM
from litellm.llms.bedrock.common_utils import BedrockError
@@ -196,7 +196,7 @@ class BedrockVectorStoreConfig(BaseVectorStoreConfig, BaseAWSLLM):
aws_region_name=self.get_aws_region_name_for_non_llm_api_calls(aws_region_name=aws_region_name),
endpoint_type="agent",
)
- return f"{endpoint_url}/knowledgebases"
+ return join_api_base(endpoint_url, "knowledgebases")
def transform_search_vector_store_request(
self,
@@ -212,7 +212,7 @@ class BedrockVectorStoreConfig(BaseVectorStoreConfig, BaseAWSLLM):
query = " ".join(query)
encoded_vector_store_id: Final = encode_url_path_segment(vector_store_id, field_name="vector_store_id")
- url: Final = f"{api_base}/{encoded_vector_store_id}/retrieve"
+ url: Final = join_api_base(api_base, f"{encoded_vector_store_id}/retrieve")
request_body: Final[dict[str, object]] = {
"retrievalQuery": BedrockKBRetrievalQuery(text=query),
diff --git a/litellm/llms/bedrock_mantle/messages/transformation.py b/litellm/llms/bedrock_mantle/messages/transformation.py
index 78881153e91..aa6510f0491 100644
--- a/litellm/llms/bedrock_mantle/messages/transformation.py
+++ b/litellm/llms/bedrock_mantle/messages/transformation.py
@@ -2,6 +2,7 @@ from collections.abc import Mapping
from types import MappingProxyType
from typing import Final
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM
from litellm.llms.bedrock.common_utils import MANTLE_MESSAGES_PATH
from litellm.llms.bedrock.messages.mantle_transformation import AmazonMantleMessagesConfig
@@ -33,7 +34,7 @@ def build_mantle_native_messages_url(api_base: str | None, litellm_params: Mappi
configured,
)
host: Final = f"https://bedrock-mantle.{region}.api.aws" if MANTLE_HOST_RE.match(stripped) else stripped
- return f"{host}{MANTLE_MESSAGES_PATH}"
+ return join_api_base(host, MANTLE_MESSAGES_PATH)
class BedrockMantleAnthropicMessagesConfig(BedrockMantleAuthMixin, AmazonMantleMessagesConfig):
diff --git a/litellm/llms/bedrock_mantle/responses/transformation.py b/litellm/llms/bedrock_mantle/responses/transformation.py
index 8a0ce58c8b2..3532cfbd963 100644
--- a/litellm/llms/bedrock_mantle/responses/transformation.py
+++ b/litellm/llms/bedrock_mantle/responses/transformation.py
@@ -22,6 +22,7 @@ import httpx
import litellm
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.responses.codex_compat import drop_unsupported_tools, normalize_codex_input_items
from litellm.llms.base_llm.responses.transformation import ReasoningSummaryLevel
@@ -100,7 +101,7 @@ class BedrockMantleResponsesAPIConfig(BedrockMantleAuthMixin, OpenAIResponsesAPI
if MANTLE_HOST_RE.match(base):
base = f"https://bedrock-mantle.{region}.api.aws"
path: Final = "/openai/v1/responses" if self.use_openai_path else "/v1/responses"
- return f"{base}{path}"
+ return join_api_base(base, path)
def validate_environment(self, headers: dict, model: str, litellm_params: GenericLiteLLMParams | None) -> dict:
litellm_params = litellm_params or GenericLiteLLMParams()
diff --git a/litellm/llms/black_forest_labs/image_edit/transformation.py b/litellm/llms/black_forest_labs/image_edit/transformation.py
index 66eacc97fba..7cfb0daf334 100644
--- a/litellm/llms/black_forest_labs/image_edit/transformation.py
+++ b/litellm/llms/black_forest_labs/image_edit/transformation.py
@@ -19,7 +19,7 @@ from typing_extensions import NotRequired, ReadOnly, TypedDict
import litellm
from litellm.constants import DEFAULT_MAX_RECURSE_DEPTH
-from litellm.litellm_core_utils.url_utils import async_safe_get, safe_get
+from litellm.litellm_core_utils.url_utils import async_safe_get, join_api_base, safe_get
from litellm.llms.base_llm.image_edit.transformation import BaseImageEditConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.images.main import ImageEditOptionalRequestParams
@@ -189,11 +189,10 @@ class BlackForestLabsImageEditConfig(BaseImageEditConfig):
"""
Get the complete URL for the Black Forest Labs API request.
"""
- base_url: str = api_base or get_secret_str("BFL_API_BASE") or DEFAULT_API_BASE
- base_url = base_url.rstrip("/")
+ base_url: Final = api_base or get_secret_str("BFL_API_BASE") or DEFAULT_API_BASE
endpoint: Final = self._get_model_endpoint(model)
- return f"{base_url}{endpoint}"
+ return join_api_base(base_url, endpoint)
def _read_image_bytes(
self,
diff --git a/litellm/llms/black_forest_labs/image_generation/transformation.py b/litellm/llms/black_forest_labs/image_generation/transformation.py
index 9d586fba26a..928af314506 100644
--- a/litellm/llms/black_forest_labs/image_generation/transformation.py
+++ b/litellm/llms/black_forest_labs/image_generation/transformation.py
@@ -14,6 +14,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
from pydantic import ConfigDict, TypeAdapter
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
)
@@ -202,11 +203,10 @@ class BlackForestLabsImageGenerationConfig(BaseImageGenerationConfig):
"""
Get the complete URL for the Black Forest Labs API request.
"""
- base_url: str = api_base or get_secret_str("BFL_API_BASE") or DEFAULT_API_BASE
- base_url = base_url.rstrip("/")
+ base_url: Final = api_base or get_secret_str("BFL_API_BASE") or DEFAULT_API_BASE
endpoint: Final = self._get_model_endpoint(model)
- return f"{base_url}{endpoint}"
+ return join_api_base(base_url, endpoint)
def transform_image_generation_request(
self,
diff --git a/litellm/llms/chatgpt/responses/transformation.py b/litellm/llms/chatgpt/responses/transformation.py
index 3e7db2981ef..34953a79c60 100644
--- a/litellm/llms/chatgpt/responses/transformation.py
+++ b/litellm/llms/chatgpt/responses/transformation.py
@@ -9,6 +9,7 @@ from litellm.litellm_core_utils.hidden_params import HIDDEN_PARAMS_ATTR, set_hid
from litellm.litellm_core_utils.llm_response_utils.convert_dict_to_response import (
safe_convert_created_field,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.common_utils import OpenAIError
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.responses.sse_output_recovery import (
@@ -253,8 +254,7 @@ class ChatGPTResponsesAPIConfig(OpenAIResponsesAPIConfig):
litellm_params: dict,
) -> str:
api_base = api_base or self.authenticator.get_api_base() or CHATGPT_API_BASE
- api_base = api_base.rstrip("/")
- return f"{api_base}/responses"
+ return join_api_base(api_base, "responses")
def supports_native_websocket(self) -> bool:
"""ChatGPT does not support native WebSocket for Responses API"""
diff --git a/litellm/llms/cloudflare/chat/transformation.py b/litellm/llms/cloudflare/chat/transformation.py
index ef11e51a5eb..0a76588f767 100644
--- a/litellm/llms/cloudflare/chat/transformation.py
+++ b/litellm/llms/cloudflare/chat/transformation.py
@@ -3,6 +3,7 @@ from typing import Final
import httpx
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.openai.chat.gpt_transformation import OpenAIGPTConfig
from litellm.secret_managers.main import (
@@ -59,7 +60,7 @@ class CloudflareChatConfig(OpenAIGPTConfig):
verbose_logger.warning(
"Cloudflare api_base ending in '/ai/run' is the legacy Workers AI path and no longer serves OpenAI-compatible requests; rewriting to the '/ai/v1' endpoint"
)
- return f"{trimmed[: -len('/ai/run')]}/ai/v1"
+ return join_api_base(trimmed[: -len("/ai/run")], "ai/v1")
return api_base
def validate_environment(
diff --git a/litellm/llms/cloudflare/common_utils.py b/litellm/llms/cloudflare/common_utils.py
index 9f1293cb9c4..02855a879a2 100644
--- a/litellm/llms/cloudflare/common_utils.py
+++ b/litellm/llms/cloudflare/common_utils.py
@@ -3,6 +3,7 @@ from typing import Final
from pydantic import TypeAdapter, ValidationError
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.secret_managers.main import (
get_secret_str,
normalize_nonempty_secret_str,
@@ -34,10 +35,10 @@ def workers_ai_request_model(model: str) -> str:
def workers_ai_run_url(api_base: str, model: str) -> str:
normalized_api_base: Final = api_base.rstrip("/")
if normalized_api_base.endswith("/ai/v1"):
- return f"{normalized_api_base.removesuffix('/ai/v1')}/ai/run/{model}"
+ return join_api_base(normalized_api_base.removesuffix("/ai/v1"), f"ai/run/{model}")
if normalized_api_base.endswith("/ai/run"):
return f"{normalized_api_base}/{model}"
- return f"{normalized_api_base}/ai/run/{model}"
+ return join_api_base(normalized_api_base, f"ai/run/{model}")
def unwrap_workers_ai_result(payload: object) -> object:
diff --git a/litellm/llms/cohere/rerank/transformation.py b/litellm/llms/cohere/rerank/transformation.py
index b6d94d4853a..8b641c24e9b 100644
--- a/litellm/llms/cohere/rerank/transformation.py
+++ b/litellm/llms/cohere/rerank/transformation.py
@@ -6,6 +6,7 @@ from pydantic import ConfigDict, TypeAdapter
import litellm
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.rerank.transformation import BaseRerankConfig
from litellm.secret_managers.main import get_secret_str
@@ -31,11 +32,7 @@ class CohereRerankConfig(BaseRerankConfig):
optional_params: dict | None = None,
) -> str:
if api_base:
- # Remove trailing slashes and ensure clean base URL
- api_base = api_base.rstrip("/")
- if not api_base.endswith("/v1/rerank"):
- api_base = f"{api_base}/v1/rerank"
- return api_base
+ return ensure_api_route(api_base, "v1/rerank")
return "https://api.cohere.ai/v1/rerank"
def get_supported_cohere_rerank_params(self, model: str) -> list:
diff --git a/litellm/llms/cohere/rerank_v2/transformation.py b/litellm/llms/cohere/rerank_v2/transformation.py
index e79357bf52f..009a220534c 100644
--- a/litellm/llms/cohere/rerank_v2/transformation.py
+++ b/litellm/llms/cohere/rerank_v2/transformation.py
@@ -1,5 +1,6 @@
from typing import Any, Final
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.cohere.rerank.transformation import CohereRerankConfig
from litellm.types.rerank import OptionalRerankParams, RerankRequest
@@ -19,11 +20,7 @@ class CohereRerankV2Config(CohereRerankConfig):
optional_params: dict | None = None,
) -> str:
if api_base:
- # Remove trailing slashes and ensure clean base URL
- api_base = api_base.rstrip("/")
- if not api_base.endswith("/v2/rerank"):
- api_base = f"{api_base}/v2/rerank"
- return api_base
+ return ensure_api_route(api_base, "v2/rerank")
return "https://api.cohere.ai/v2/rerank"
def get_supported_cohere_rerank_params(self, model: str) -> list:
diff --git a/litellm/llms/cometapi/chat/transformation.py b/litellm/llms/cometapi/chat/transformation.py
index 26050700efc..2bf119a0a4a 100644
--- a/litellm/llms/cometapi/chat/transformation.py
+++ b/litellm/llms/cometapi/chat/transformation.py
@@ -10,6 +10,7 @@ from typing import Any, Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.base_model_iterator import BaseModelResponseIterator
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.types.llms.openai import AllMessageValues, ChatCompletionToolParam
@@ -112,16 +113,14 @@ class CometAPIConfig(OpenAIGPTConfig):
# Ensure we include /v1 prefix when missing
if api_base.endswith("/v1"):
- return f"{api_base}/{endpoint}"
- if api_base.endswith("/v1/"):
- return f"{api_base}{endpoint}"
+ return join_api_base(api_base, endpoint)
# If user provided https://api.cometapi.com, add /v1
if api_base == "https://api.cometapi.com":
- return f"{api_base}/v1/{endpoint}"
+ return join_api_base(api_base, f"v1/{endpoint}")
# Generic fallback: if '/v1' not in path, add it
if "/v1" not in api_base.split("//", 1)[-1]:
- return f"{api_base}/v1/{endpoint}"
- return f"{api_base}/{endpoint}"
+ return join_api_base(api_base, f"v1/{endpoint}")
+ return join_api_base(api_base, endpoint)
def get_error_class(self, error_message: str, status_code: int, headers: dict | httpx.Headers) -> BaseLLMException:
"""
diff --git a/litellm/llms/cometapi/embed/transformation.py b/litellm/llms/cometapi/embed/transformation.py
index 60f781be261..e3b5c722664 100644
--- a/litellm/llms/cometapi/embed/transformation.py
+++ b/litellm/llms/cometapi/embed/transformation.py
@@ -9,6 +9,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.secret_managers.main import get_secret_str
@@ -44,9 +45,7 @@ class CometAPIEmbeddingConfig(BaseEmbeddingConfig):
"""
Get the complete URL for the CometAPI embedding endpoint.
"""
- api_base = "https://api.cometapi.com/v1" if api_base is None else api_base.rstrip("/")
- complete_url: Final = f"{api_base}/embeddings"
- return complete_url
+ return join_api_base("https://api.cometapi.com/v1" if api_base is None else api_base, "embeddings")
def validate_environment(
self,
diff --git a/litellm/llms/cometapi/image_generation/transformation.py b/litellm/llms/cometapi/image_generation/transformation.py
index 567dbfd308e..ffaabbb28cd 100644
--- a/litellm/llms/cometapi/image_generation/transformation.py
+++ b/litellm/llms/cometapi/image_generation/transformation.py
@@ -4,6 +4,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
from pydantic import ConfigDict, TypeAdapter
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
)
@@ -77,16 +78,13 @@ class CometAPIImageGenerationConfig(BaseImageGenerationConfig):
"""
Get the complete url for the request
"""
- complete_url: str = (
+ base_url: Final = (
api_base
or get_secret_str("COMETAPI_BASE_URL")
or get_secret_str("COMETAPI_API_BASE")
or self.DEFAULT_BASE_URL
)
-
- complete_url = complete_url.rstrip("/")
- complete_url = f"{complete_url}/{self.IMAGE_GENERATION_ENDPOINT}"
- return complete_url
+ return join_api_base(base_url, self.IMAGE_GENERATION_ENDPOINT)
def validate_environment(
self,
diff --git a/litellm/llms/custom_httpx/llm_http_handler.py b/litellm/llms/custom_httpx/llm_http_handler.py
index ac2ac1fe31b..a7ca0d5f49e 100644
--- a/litellm/llms/custom_httpx/llm_http_handler.py
+++ b/litellm/llms/custom_httpx/llm_http_handler.py
@@ -59,7 +59,7 @@ from litellm.litellm_core_utils.realtime_errors import (
websocket_close_reason,
)
from litellm.litellm_core_utils.realtime_streaming import RealTimeStreaming
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.anthropic_messages.transformation import (
BaseAnthropicMessagesConfig,
)
@@ -6937,7 +6937,7 @@ class BaseLLMHTTPHandler:
headers={}, model=model or "", api_key=api_key
)
else:
- url = f"{api_base.rstrip('/')}/v1/realtime/{endpoint}"
+ url = join_api_base(api_base, f"v1/realtime/{endpoint}")
headers = {
"Authorization": f"Bearer {api_key}",
"Content-Type": "application/json",
@@ -7006,7 +7006,7 @@ class BaseLLMHTTPHandler:
url = provider_config.get_realtime_calls_url(api_base=api_base, model=model or "", api_version=api_version)
headers: dict[str, object] = provider_config.get_realtime_calls_headers(ephemeral_key=openai_ephemeral_key)
else:
- url = f"{api_base.rstrip('/')}/v1/realtime/calls"
+ url = join_api_base(api_base, "v1/realtime/calls")
headers = {
"Authorization": f"Bearer {openai_ephemeral_key}",
}
diff --git a/litellm/llms/dashscope/chat/transformation.py b/litellm/llms/dashscope/chat/transformation.py
index 9a1b381a2ba..966d1a01b8d 100644
--- a/litellm/llms/dashscope/chat/transformation.py
+++ b/litellm/llms/dashscope/chat/transformation.py
@@ -5,6 +5,7 @@ Translates from OpenAI's `/v1/chat/completions` to DashScope's `/v1/chat/complet
from collections.abc import Coroutine
from typing import Final, Literal, overload
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import AllMessageValues, ChatCompletionToolParam
@@ -82,7 +83,4 @@ class DashScopeChatConfig(OpenAIGPTConfig):
"""
If api_base is not provided, use the default DashScope /chat/completions endpoint.
"""
- resolved_api_base: Final = self._resolve_chat_api_base(api_base).rstrip("/")
- if resolved_api_base.endswith("/chat/completions"):
- return resolved_api_base
- return f"{resolved_api_base}/chat/completions"
+ return ensure_api_route(self._resolve_chat_api_base(api_base), "chat/completions")
diff --git a/litellm/llms/dashscope/embed/transformation.py b/litellm/llms/dashscope/embed/transformation.py
index e5a987ca1ad..d0c2b4dff0c 100644
--- a/litellm/llms/dashscope/embed/transformation.py
+++ b/litellm/llms/dashscope/embed/transformation.py
@@ -16,6 +16,7 @@ from typing import Final
import httpx
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.secret_managers.main import get_secret_str
@@ -98,10 +99,7 @@ class DashScopeEmbeddingConfig(BaseEmbeddingConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- base: Final = self._resolve_embedding_api_base(api_base).rstrip("/")
- if base.endswith("/embeddings"):
- return base
- return f"{base}/embeddings"
+ return ensure_api_route(self._resolve_embedding_api_base(api_base), "embeddings")
def transform_embedding_request(
self,
diff --git a/litellm/llms/dashscope/rerank/transformation.py b/litellm/llms/dashscope/rerank/transformation.py
index 4ec7d697f02..48ce4b5a07b 100644
--- a/litellm/llms/dashscope/rerank/transformation.py
+++ b/litellm/llms/dashscope/rerank/transformation.py
@@ -34,6 +34,7 @@ from pydantic import ConfigDict, TypeAdapter
from litellm._uuid import uuid
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.rerank.transformation import BaseRerankConfig
from litellm.secret_managers.main import get_secret_str
@@ -94,10 +95,10 @@ class DashScopeRerankConfig(BaseRerankConfig):
return cleaned
if cleaned.endswith("/v1"):
- return f"{cleaned}/reranks"
+ return join_api_base(cleaned, "reranks")
# Unknown base: append /reranks rather than silently ignoring the caller's api_base.
- return f"{cleaned}/reranks"
+ return join_api_base(cleaned, "reranks")
def validate_environment(
self,
diff --git a/litellm/llms/dashscope/videos/transformation.py b/litellm/llms/dashscope/videos/transformation.py
index 1db91cbb127..5c738423a5a 100644
--- a/litellm/llms/dashscope/videos/transformation.py
+++ b/litellm/llms/dashscope/videos/transformation.py
@@ -16,7 +16,7 @@ from pydantic import BaseModel, Field, StrictBool, StrictInt, StrictStr, TypeAda
import litellm
from litellm.exceptions import UnsupportedParamsError
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.videos.async_task_utils import (
input_reference_url,
@@ -708,7 +708,7 @@ class DashScopeVideoConfig(BaseVideoConfig):
if parameters:
request_data["parameters"] = parameters
- return request_data, (), f"{api_base}{DASHSCOPE_VIDEO_SYNTHESIS_PATH}"
+ return request_data, (), join_api_base(api_base, DASHSCOPE_VIDEO_SYNTHESIS_PATH)
@staticmethod
def _reference_input(
@@ -773,7 +773,7 @@ class DashScopeVideoConfig(BaseVideoConfig):
def _task_url(video_id: str, api_base: str) -> str:
original_task_id: Final = extract_original_video_id(video_id)
encoded_task_id: Final = encode_url_path_segment(original_task_id, field_name="video_id")
- return f"{api_base}{DASHSCOPE_TASKS_PATH}/{encoded_task_id}"
+ return join_api_base(api_base, f"{DASHSCOPE_TASKS_PATH}/{encoded_task_id}")
def transform_video_status_retrieve_response(
self,
diff --git a/litellm/llms/databricks/chat/transformation.py b/litellm/llms/databricks/chat/transformation.py
index d46d04578cc..61b41c2a173 100644
--- a/litellm/llms/databricks/chat/transformation.py
+++ b/litellm/llms/databricks/chat/transformation.py
@@ -22,6 +22,7 @@ from litellm.litellm_core_utils.prompt_templates.common_utils import (
strip_name_from_message,
strip_prefill_prefix_field,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.base_model_iterator import BaseModelResponseIterator
from litellm.llms.base_llm.base_utils import (
type_to_response_format_param, # pyright: ignore[reportUnknownVariableType] # base_utils helper returns an untyped dict
@@ -339,10 +340,7 @@ class DatabricksConfig(DatabricksBase, OpenAILikeChatConfig, AnthropicConfig):
stream: bool | None = None,
) -> str:
use_ai_gateway: Final = self.uses_ai_gateway(model)
- api_base = self._get_api_base(api_base, use_ai_gateway=use_ai_gateway)
- url_base: Final = api_base.rstrip("/") if use_ai_gateway else api_base
- complete_url: Final = f"{url_base}/chat/completions"
- return complete_url
+ return join_api_base(self._get_api_base(api_base, use_ai_gateway=use_ai_gateway), "chat/completions")
def get_supported_openai_params(self, model: str | None = None) -> list:
return [
diff --git a/litellm/llms/databricks/common_utils.py b/litellm/llms/databricks/common_utils.py
index 3112b2458b4..cd9432ac36f 100644
--- a/litellm/llms/databricks/common_utils.py
+++ b/litellm/llms/databricks/common_utils.py
@@ -20,6 +20,7 @@ from urllib.parse import urlsplit, urlunsplit
from pydantic import BaseModel, ConfigDict
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.secret_managers.main import get_secret_str
@@ -278,10 +279,10 @@ class DatabricksBase:
if normalized_api_base.endswith("/ai-gateway/mlflow/v1"):
return normalized_api_base
if normalized_api_base.endswith("/serving-endpoints"):
- return f"{normalized_api_base.removesuffix('/serving-endpoints')}/ai-gateway/mlflow/v1"
+ return join_api_base(normalized_api_base.removesuffix("/serving-endpoints"), "ai-gateway/mlflow/v1")
api_base_parts: Final = urlsplit(normalized_api_base)
if api_base_parts.path in ("", "/"):
- return f"{normalized_api_base}/ai-gateway/mlflow/v1"
+ return join_api_base(normalized_api_base, "ai-gateway/mlflow/v1")
return api_base
def get_oauth_m2m_token(
@@ -319,7 +320,7 @@ class DatabricksBase:
api_base_parts: Final = urlsplit(api_base)
workspace_url: Final = urlunsplit((api_base_parts.scheme, api_base_parts.netloc, "", "", ""))
- token_url: Final = f"{workspace_url}/oidc/v1/token"
+ token_url: Final = join_api_base(workspace_url, "oidc/v1/token")
try:
response: Final = requests.post(
@@ -477,5 +478,6 @@ class DatabricksBase:
# Debug logging with redaction (never log actual tokens)
verbose_logger.debug("Databricks request headers: %s", self.redact_headers_for_logging(headers))
- endpoint_path: Final = "" if custom_endpoint is True else DATABRICKS_ENDPOINT_PATHS[endpoint_type]
- return f"{api_base}{endpoint_path}", headers
+ if custom_endpoint is True:
+ return api_base, headers
+ return join_api_base(api_base, DATABRICKS_ENDPOINT_PATHS[endpoint_type]), headers
diff --git a/litellm/llms/databricks/decisions/transformation.py b/litellm/llms/databricks/decisions/transformation.py
index c2ac1f220b8..d1698fdf5e2 100644
--- a/litellm/llms/databricks/decisions/transformation.py
+++ b/litellm/llms/databricks/decisions/transformation.py
@@ -6,6 +6,7 @@ from typing import Final
from pydantic import TypeAdapter, ValidationError
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.decisions.transformation import BaseDecisionsConfig
from litellm.types.decisions import DecisionsIRRequest, UnsupportedDecisionsRequest
@@ -84,8 +85,8 @@ class DatabricksDecisionsConfig(BaseDecisionsConfig):
def get_complete_url(self, api_base: str, model: str) -> str:
if is_ai_decide_model(model):
- return f"{databricks_workspace_host(api_base)}{_AI_DECIDE_PATH}"
- return f"{api_base.rstrip('/')}/{validate_databricks_decisions_model(model)}/invocations"
+ return join_api_base(databricks_workspace_host(api_base), _AI_DECIDE_PATH)
+ return join_api_base(api_base, f"{validate_databricks_decisions_model(model)}/invocations")
def transform_decisions_request(
self,
diff --git a/litellm/llms/databricks/responses/transformation.py b/litellm/llms/databricks/responses/transformation.py
index 804c79ea79d..55c3edfefd0 100644
--- a/litellm/llms/databricks/responses/transformation.py
+++ b/litellm/llms/databricks/responses/transformation.py
@@ -10,6 +10,7 @@ Reference: https://docs.databricks.com/aws/en/machine-learning/foundation-model-
import os
from typing import TYPE_CHECKING, Any, Final
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.databricks.common_utils import DatabricksBase
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.types.llms.openai import ResponseInputParam
@@ -70,8 +71,7 @@ class DatabricksResponsesAPIConfig(DatabricksBase, OpenAIResponsesAPIConfig):
) -> str:
api_base = api_base or os.getenv("DATABRICKS_API_BASE")
api_base = self._get_api_base(api_base)
- api_base = api_base.rstrip("/")
- return f"{api_base}/responses"
+ return join_api_base(api_base, "responses")
def transform_responses_api_request(
self,
diff --git a/litellm/llms/deepgram/audio_transcription/transformation.py b/litellm/llms/deepgram/audio_transcription/transformation.py
index a3812dac8ef..a1c3021c9f5 100644
--- a/litellm/llms/deepgram/audio_transcription/transformation.py
+++ b/litellm/llms/deepgram/audio_transcription/transformation.py
@@ -10,6 +10,7 @@ from httpx import Headers, Response
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.audio_utils.utils import process_audio_file
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import (
@@ -179,7 +180,6 @@ class DeepgramAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
) -> str:
if api_base is None:
api_base = get_secret_str("DEEPGRAM_API_BASE") or "https://api.deepgram.com/v1"
- api_base = api_base.rstrip("/") # Remove trailing slash if present
# Build query parameters including the model
all_query_params: Final = {"model": model}
@@ -189,7 +189,7 @@ class DeepgramAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
all_query_params.update(additional_params)
# Construct URL with proper query string encoding
- base_url: Final = f"{api_base}/listen"
+ base_url: Final = join_api_base(api_base, "listen")
query_string: Final = urlencode(all_query_params)
url: Final = f"{base_url}?{query_string}"
diff --git a/litellm/llms/deepgram/common_utils.py b/litellm/llms/deepgram/common_utils.py
index 9b071ac8321..29998895bbc 100644
--- a/litellm/llms/deepgram/common_utils.py
+++ b/litellm/llms/deepgram/common_utils.py
@@ -8,6 +8,7 @@ import httpx
import litellm
from litellm.constants import DEEPGRAM_DEFAULT_API_BASE, DEEPGRAM_LISTEN_DEFAULT_MODEL
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.types.utils import LlmProviders
@@ -50,7 +51,7 @@ def _first_occurrences(query_string: str) -> httpx.QueryParams:
def deepgram_listen_websocket_target(api_base: str | None, query_string: str) -> str:
- listen_url: Final = httpx.URL(f"{(api_base or DEEPGRAM_DEFAULT_API_BASE).rstrip('/')}/listen")
+ listen_url: Final = httpx.URL(join_api_base(api_base or DEEPGRAM_DEFAULT_API_BASE, "listen"))
websocket_url: Final = listen_url.copy_with(scheme=_WEBSOCKET_SCHEMES.get(listen_url.scheme, listen_url.scheme))
params: Final = _first_occurrences(query_string)
query: Final = params if params.get("model") else params.remove("model").add("model", DEEPGRAM_LISTEN_DEFAULT_MODEL)
diff --git a/litellm/llms/deepseek/chat/transformation.py b/litellm/llms/deepseek/chat/transformation.py
index d3319856d29..a9a20101bb1 100644
--- a/litellm/llms/deepseek/chat/transformation.py
+++ b/litellm/llms/deepseek/chat/transformation.py
@@ -10,6 +10,7 @@ from litellm.litellm_core_utils.prompt_templates.common_utils import (
convert_content_list_to_str,
extract_search_results_text,
)
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import AllMessageValues
from litellm.utils import supports_reasoning, supports_vision
@@ -363,9 +364,4 @@ class DeepSeekChatConfig(OpenAIGPTConfig):
"""
If api_base is not provided, use the default DeepSeek /chat/completions endpoint.
"""
- clean_api_base: Final = (api_base or "https://api.deepseek.com/beta").rstrip("/")
-
- if not clean_api_base.endswith("/chat/completions"):
- return f"{clean_api_base}/chat/completions"
-
- return clean_api_base
+ return ensure_api_route(api_base or "https://api.deepseek.com/beta", "chat/completions")
diff --git a/litellm/llms/deepseek/messages/transformation.py b/litellm/llms/deepseek/messages/transformation.py
index 6a6b57131ad..6ddee99aea3 100644
--- a/litellm/llms/deepseek/messages/transformation.py
+++ b/litellm/llms/deepseek/messages/transformation.py
@@ -6,6 +6,7 @@ from collections.abc import Sequence
from typing import Any, Final
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.anthropic.pass_through.messages.transformation import (
AnthropicMessagesConfig,
)
@@ -90,9 +91,9 @@ class DeepSeekAnthropicMessagesConfig(AnthropicMessagesConfig):
base_url = base_url.removesuffix("/beta")
if not base_url.endswith("/anthropic") and "/anthropic/" not in base_url:
- base_url = f"{base_url}/anthropic"
+ base_url = join_api_base(base_url, "anthropic")
- return f"{base_url}/v1/messages"
+ return join_api_base(base_url, "v1/messages")
@staticmethod
def _sanitize_tools_for_deepseek(tools: Any) -> object:
diff --git a/litellm/llms/docker_model_runner/chat/transformation.py b/litellm/llms/docker_model_runner/chat/transformation.py
index adb40311efd..68b63812248 100644
--- a/litellm/llms/docker_model_runner/chat/transformation.py
+++ b/litellm/llms/docker_model_runner/chat/transformation.py
@@ -10,6 +10,7 @@ from typing import Final, Literal, overload
from litellm.litellm_core_utils.prompt_templates.common_utils import (
handle_messages_with_content_list_to_str_conversion,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import AllMessageValues
@@ -104,12 +105,9 @@ class DockerModelRunnerChatConfig(OpenAIGPTConfig):
if not api_base:
api_base = "http://localhost:22088/engines/llama.cpp"
- # Remove trailing slashes from api_base
- api_base = api_base.rstrip("/")
-
# Build the URL: {api_base}/v1/chat/completions
# api_base is expected to already contain the engine path
- complete_url: Final = f"{api_base}/v1/chat/completions"
+ complete_url: Final = join_api_base(api_base, "v1/chat/completions")
return complete_url
diff --git a/litellm/llms/duckduckgo/search/transformation.py b/litellm/llms/duckduckgo/search/transformation.py
index 1816a0d24e4..36893b4bd27 100644
--- a/litellm/llms/duckduckgo/search/transformation.py
+++ b/litellm/llms/duckduckgo/search/transformation.py
@@ -12,6 +12,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
SearchResponse,
@@ -91,7 +92,7 @@ class DuckDuckGoSearchConfig(BaseSearchConfig):
if data and isinstance(data, dict) and "_duckduckgo_params" in data:
params: Final = data["_duckduckgo_params"]
query_string: Final = urlencode(params, doseq=True)
- return f"{api_base}/?{query_string}"
+ return f"{join_api_base(api_base, '')}?{query_string}"
return api_base
diff --git a/litellm/llms/e2b/sandbox/transformation.py b/litellm/llms/e2b/sandbox/transformation.py
index dbf3996248a..b5f42632eef 100644
--- a/litellm/llms/e2b/sandbox/transformation.py
+++ b/litellm/llms/e2b/sandbox/transformation.py
@@ -15,6 +15,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.hidden_params import HIDDEN_PARAMS_ATTR
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.sandbox.transformation import (
SANDBOX_MAX_OUTPUT_BYTES,
BaseSandboxConfig,
@@ -78,7 +79,7 @@ class E2BSandboxConfig(BaseSandboxConfig):
body["metadata"] = metadata
response: Final = await self._http(client).post(
- url=f"{base}/sandboxes",
+ url=join_api_base(base, "sandboxes"),
headers={"X-API-Key": key, "Content-Type": "application/json"},
json=body,
)
@@ -151,10 +152,10 @@ class E2BSandboxConfig(BaseSandboxConfig):
dict[str, object], getattr(handle, HIDDEN_PARAMS_ATTR)
)
key: Final = api_key or hidden_params.get("api_key") or self.validate_environment()
- base: Final = api_base or hidden_params.get("api_base") or E2B_API_BASE
+ base: Final = str(api_base or hidden_params.get("api_base") or E2B_API_BASE)
try:
response: Final = await self._http(client).delete(
- url=f"{base}/sandboxes/{handle.id}",
+ url=join_api_base(base, f"sandboxes/{handle.id}"),
headers={"X-API-Key": key},
)
except httpx.HTTPStatusError as e:
diff --git a/litellm/llms/edenai/chat/transformation.py b/litellm/llms/edenai/chat/transformation.py
index ae5da5d1cfd..480b382076d 100644
--- a/litellm/llms/edenai/chat/transformation.py
+++ b/litellm/llms/edenai/chat/transformation.py
@@ -23,7 +23,7 @@ from litellm.types.llms.base import LiteLLMBaseModel
from litellm.types.llms.openai import AllMessageValues
from litellm.types.utils import ModelResponse, ModelResponseStream, Usage
-from ..common_utils import EdenAIException, reported_cost, resolve_api_base, resolve_api_key
+from ..common_utils import EdenAIException, endpoint_url, reported_cost, resolve_api_base, resolve_api_key
if TYPE_CHECKING:
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
@@ -138,7 +138,7 @@ class EdenAIChatConfig(OpenAIGPTConfig):
def get_models(
self, api_key: str | None = None, api_base: str | None = None
) -> list[str]: # mutable-ok: inherited contract
- response: Final = litellm.module_level_client.get(url=f"{self.get_api_base(api_base)}/models")
+ response: Final = litellm.module_level_client.get(url=endpoint_url(api_base, "models"))
if not response.is_success:
raise EdenAIException(status_code=response.status_code, message=response.text, headers=response.headers)
catalog: Final = _EdenAIModelCatalog.model_validate(response.json())
diff --git a/litellm/llms/edenai/common_utils.py b/litellm/llms/edenai/common_utils.py
index 3216f798fbe..60c83f7ea75 100644
--- a/litellm/llms/edenai/common_utils.py
+++ b/litellm/llms/edenai/common_utils.py
@@ -11,6 +11,7 @@ from pydantic import AliasChoices, Field, ValidationError
import litellm
from litellm.exceptions import AuthenticationError
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.base import LiteLLMBaseModel
@@ -74,7 +75,7 @@ def json_headers(
def endpoint_url(api_base: str | None, path: str) -> str:
- return f"{resolve_api_base(api_base).rstrip('/')}/{path}"
+ return join_api_base(resolve_api_base(api_base), path)
def pick(params: Mapping[str, object], keys: Container[str]) -> Mapping[str, object]:
diff --git a/litellm/llms/elevenlabs/audio_transcription/transformation.py b/litellm/llms/elevenlabs/audio_transcription/transformation.py
index 04d2717d64f..5a673b5ac41 100644
--- a/litellm/llms/elevenlabs/audio_transcription/transformation.py
+++ b/litellm/llms/elevenlabs/audio_transcription/transformation.py
@@ -10,6 +10,7 @@ from pydantic import ConfigDict, TypeAdapter
import litellm
from litellm.litellm_core_utils.audio_utils.utils import process_audio_file
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import (
@@ -165,12 +166,7 @@ class ElevenLabsAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
) -> str:
if api_base is None:
api_base = get_secret_str("ELEVENLABS_API_BASE") or "https://api.elevenlabs.io"
- api_base = api_base.rstrip("/") # Remove trailing slash if present
-
- # ElevenLabs speech-to-text endpoint
- url: Final = f"{api_base}/v1/speech-to-text"
-
- return url
+ return join_api_base(api_base, "v1/speech-to-text")
def validate_environment(
self,
diff --git a/litellm/llms/elevenlabs/text_to_speech/transformation.py b/litellm/llms/elevenlabs/text_to_speech/transformation.py
index eb93543df46..f910ef8c057 100644
--- a/litellm/llms/elevenlabs/text_to_speech/transformation.py
+++ b/litellm/llms/elevenlabs/text_to_speech/transformation.py
@@ -11,7 +11,7 @@ import httpx
from httpx import Headers
import litellm
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.text_to_speech.transformation import (
BaseTextToSpeechConfig,
@@ -297,15 +297,14 @@ class ElevenLabsTextToSpeechConfig(BaseTextToSpeechConfig):
"""
Construct the ElevenLabs endpoint URL, including path voice_id and query params.
"""
- base_url = api_base or get_secret_str("ELEVENLABS_API_BASE") or self.TTS_BASE_URL
- base_url = base_url.rstrip("/")
+ base_url: Final = api_base or get_secret_str("ELEVENLABS_API_BASE") or self.TTS_BASE_URL
voice_id: Final = litellm_params.get(self.ELEVENLABS_VOICE_ID_KEY)
if not isinstance(voice_id, str) or not voice_id.strip():
raise ValueError("ElevenLabs voice_id is required. Pass `voice` when calling `litellm.speech()`.")
encoded_voice_id: Final = encode_url_path_segment(voice_id, field_name="voice_id")
- url = f"{base_url}{self.TTS_ENDPOINT_PATH}/{encoded_voice_id}"
+ url = join_api_base(base_url, f"{self.TTS_ENDPOINT_PATH}/{encoded_voice_id}")
query_params: Final = litellm_params.get(self.ELEVENLABS_QUERY_PARAMS_KEY, {})
if query_params:
diff --git a/litellm/llms/exa_ai/search/transformation.py b/litellm/llms/exa_ai/search/transformation.py
index 89073409599..73e341d0033 100644
--- a/litellm/llms/exa_ai/search/transformation.py
+++ b/litellm/llms/exa_ai/search/transformation.py
@@ -11,6 +11,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
SearchResponse,
@@ -93,11 +94,7 @@ class ExaAISearchConfig(BaseSearchConfig):
"""
Get complete URL for Search endpoint.
"""
- resolved_api_base: Final = api_base or get_secret_str("EXA_API_BASE") or self.EXA_AI_API_BASE
- clean_api_base: Final = resolved_api_base.rstrip("/")
- if clean_api_base.endswith("/search"):
- return clean_api_base
- return f"{clean_api_base}/search"
+ return ensure_api_route(api_base or get_secret_str("EXA_API_BASE") or self.EXA_AI_API_BASE, "search")
def transform_search_request(
self,
diff --git a/litellm/llms/fal_ai/image_edit/flux_lora_depth_transformation.py b/litellm/llms/fal_ai/image_edit/flux_lora_depth_transformation.py
index 6d58caeb384..344d9d3c75b 100644
--- a/litellm/llms/fal_ai/image_edit/flux_lora_depth_transformation.py
+++ b/litellm/llms/fal_ai/image_edit/flux_lora_depth_transformation.py
@@ -4,6 +4,7 @@ from typing import Final
from httpx._types import RequestFiles
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.secret_managers.main import get_secret_str
from litellm.types.images.main import ImageEditOptionalRequestParams
from litellm.types.router import GenericLiteLLMParams
@@ -45,8 +46,8 @@ class FalAIFluxLoraDepthEditConfig(FalAIImageEditConfig):
api_base: str | None,
litellm_params: dict, # mutable-ok: base class contract
) -> str:
- base_url: Final = (api_base or get_secret_str("FAL_AI_API_BASE") or DEFAULT_BASE_URL).rstrip("/")
- return f"{base_url}/{FLUX_LORA_DEPTH_ENDPOINT}"
+ base_url: Final = api_base or get_secret_str("FAL_AI_API_BASE") or DEFAULT_BASE_URL
+ return join_api_base(base_url, FLUX_LORA_DEPTH_ENDPOINT)
def transform_image_edit_request(
self,
diff --git a/litellm/llms/fal_ai/image_generation/transformation.py b/litellm/llms/fal_ai/image_generation/transformation.py
index 0bb5dada071..79913d5c651 100644
--- a/litellm/llms/fal_ai/image_generation/transformation.py
+++ b/litellm/llms/fal_ai/image_generation/transformation.py
@@ -5,6 +5,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from typing_extensions import ReadOnly, TypedDict
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
)
@@ -85,12 +86,10 @@ class FalAIBaseConfig(BaseImageGenerationConfig):
Some providers need `model` in `api_base`
"""
- complete_url: str = api_base or get_secret_str("FAL_AI_API_BASE") or self.DEFAULT_BASE_URL
-
- complete_url = complete_url.rstrip("/")
- if self.IMAGE_GENERATION_ENDPOINT:
- complete_url = f"{complete_url}/{self.IMAGE_GENERATION_ENDPOINT}"
- return complete_url
+ base_url: Final = api_base or get_secret_str("FAL_AI_API_BASE") or self.DEFAULT_BASE_URL
+ if not self.IMAGE_GENERATION_ENDPOINT:
+ return base_url.rstrip("/")
+ return join_api_base(base_url, self.IMAGE_GENERATION_ENDPOINT)
def validate_environment(
self,
diff --git a/litellm/llms/fal_ai/videos/transformation.py b/litellm/llms/fal_ai/videos/transformation.py
index 3136763e1be..86aa6388ff0 100644
--- a/litellm/llms/fal_ai/videos/transformation.py
+++ b/litellm/llms/fal_ai/videos/transformation.py
@@ -10,7 +10,7 @@ import httpx
from httpx._types import FileContent, RequestFiles
from pydantic import TypeAdapter
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.videos.transformation import BaseVideoConfig
from litellm.llms.custom_httpx.http_handler import (
@@ -445,7 +445,7 @@ class FalAIVideoConfig(BaseVideoConfig):
request_id, model_id = self._decode_video_id(video_id)
encoded_request_id: Final[str] = encode_url_path_segment(request_id, field_name="video_id")
return (
- f"{api_base.rstrip('/')}/{_queue_request_base_path(model_id)}/requests/{encoded_request_id}/status",
+ join_api_base(api_base, f"{_queue_request_base_path(model_id)}/requests/{encoded_request_id}/status"),
{},
)
@@ -541,7 +541,7 @@ class FalAIVideoConfig(BaseVideoConfig):
request_id, model_id = self._decode_video_id(video_id)
encoded_request_id: Final[str] = encode_url_path_segment(request_id, field_name="video_id")
return (
- f"{api_base.rstrip('/')}/{_queue_request_base_path(model_id)}/requests/{encoded_request_id}",
+ join_api_base(api_base, f"{_queue_request_base_path(model_id)}/requests/{encoded_request_id}"),
{},
)
diff --git a/litellm/llms/fastcrw/search/transformation.py b/litellm/llms/fastcrw/search/transformation.py
index 8f6f8d87300..27db6a95170 100644
--- a/litellm/llms/fastcrw/search/transformation.py
+++ b/litellm/llms/fastcrw/search/transformation.py
@@ -15,6 +15,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
SearchResponse,
@@ -83,11 +84,7 @@ class FastCRWSearchConfig(BaseSearchConfig):
"""
Get complete URL for Search endpoint.
"""
- resolved_api_base: Final = api_base or get_secret_str("CRW_API_BASE") or self.FASTCRW_API_BASE
- clean_api_base: Final = resolved_api_base.rstrip("/")
- if clean_api_base.endswith("/search"):
- return clean_api_base
- return f"{clean_api_base}/search"
+ return ensure_api_route(api_base or get_secret_str("CRW_API_BASE") or self.FASTCRW_API_BASE, "search")
def transform_search_request(
self,
diff --git a/litellm/llms/firecrawl/search/transformation.py b/litellm/llms/firecrawl/search/transformation.py
index ee99ab24413..6d1e2b22918 100644
--- a/litellm/llms/firecrawl/search/transformation.py
+++ b/litellm/llms/firecrawl/search/transformation.py
@@ -11,6 +11,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
SearchResponse,
@@ -85,11 +86,7 @@ class FirecrawlSearchConfig(BaseSearchConfig):
"""
Get complete URL for Search endpoint.
"""
- resolved_api_base: Final = api_base or get_secret_str("FIRECRAWL_API_BASE") or self.FIRECRAWL_API_BASE
- clean_api_base: Final = resolved_api_base.rstrip("/")
- if clean_api_base.endswith("/search"):
- return clean_api_base
- return f"{clean_api_base}/search"
+ return ensure_api_route(api_base or get_secret_str("FIRECRAWL_API_BASE") or self.FIRECRAWL_API_BASE, "search")
def transform_search_request(
self,
diff --git a/litellm/llms/fireworks_ai/chat/transformation.py b/litellm/llms/fireworks_ai/chat/transformation.py
index 14ae59d4f85..fa5794d71b3 100644
--- a/litellm/llms/fireworks_ai/chat/transformation.py
+++ b/litellm/llms/fireworks_ai/chat/transformation.py
@@ -17,6 +17,7 @@ from litellm.litellm_core_utils.prompt_templates.common_utils import (
strip_litellm_response_only_fields,
unpack_legacy_defs,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import (
AllMessageValues,
@@ -826,7 +827,7 @@ class FireworksAIConfig(FireworksAIMixin, OpenAIGPTConfig):
base = api_base.rstrip("/")
base = base.removesuffix("/v1")
response: Final = litellm.module_level_client.get(
- url=f"{base}/v1/accounts/{account_id}/models",
+ url=join_api_base(base, f"v1/accounts/{account_id}/models"),
headers={"Authorization": f"Bearer {api_key}"},
)
diff --git a/litellm/llms/fireworks_ai/rerank/transformation.py b/litellm/llms/fireworks_ai/rerank/transformation.py
index 4eff64e351f..964f85c821e 100644
--- a/litellm/llms/fireworks_ai/rerank/transformation.py
+++ b/litellm/llms/fireworks_ai/rerank/transformation.py
@@ -12,6 +12,7 @@ from pydantic import ConfigDict, TypeAdapter
from litellm._uuid import uuid
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.rerank.transformation import BaseRerankConfig
from litellm.llms.fireworks_ai.common_utils import FireworksAIMixin
from litellm.types.llms.base import LiteLLMBaseModel
@@ -61,9 +62,9 @@ class FireworksAIRerankConfig(FireworksAIMixin, BaseRerankConfig):
api_base = api_base.rstrip("/")
if not api_base.endswith("/rerank"):
if api_base.endswith("/v1") or api_base.endswith("/inference/v1"):
- api_base = f"{api_base}/rerank"
+ api_base = join_api_base(api_base, "rerank")
else:
- api_base = f"{api_base}/inference/v1/rerank"
+ api_base = join_api_base(api_base, "inference/v1/rerank")
return api_base
return "https://api.fireworks.ai/inference/v1/rerank"
diff --git a/litellm/llms/fireworks_ai/responses/transformation.py b/litellm/llms/fireworks_ai/responses/transformation.py
index 30eaa5a7ca2..db234c09930 100644
--- a/litellm/llms/fireworks_ai/responses/transformation.py
+++ b/litellm/llms/fireworks_ai/responses/transformation.py
@@ -6,6 +6,7 @@ from urllib.parse import unquote
import httpx
from openai.types.responses import EasyInputMessageParam, ResponseInputContentParam, ResponseInputItemParam
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.fireworks_ai.common_utils import (
FIREWORKS_FORWARD_USER_ID_PARAM,
get_fireworks_forwarded_user_id,
@@ -150,8 +151,8 @@ class FireworksAIResponsesAPIConfig(OpenAIResponsesAPIConfig):
return dict(pinned)
def get_complete_url(self, api_base: str | None, litellm_params: Mapping[str, object]) -> str:
- base: Final = (api_base or get_secret_str("FIREWORKS_API_BASE") or FIREWORKS_AI_DEFAULT_API_BASE).rstrip("/")
- return f"{base}/responses"
+ base: Final = api_base or get_secret_str("FIREWORKS_API_BASE") or FIREWORKS_AI_DEFAULT_API_BASE
+ return join_api_base(base, "responses")
def transform_responses_api_request(
self,
diff --git a/litellm/llms/gdc/chat/transformation.py b/litellm/llms/gdc/chat/transformation.py
index 6eac3ac79cd..cd9ffe9e03e 100644
--- a/litellm/llms/gdc/chat/transformation.py
+++ b/litellm/llms/gdc/chat/transformation.py
@@ -13,6 +13,7 @@ from urllib.parse import urlsplit
from typing_extensions import ReadOnly, TypedDict, Unpack
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai_like.chat.transformation import OpenAILikeChatConfig
from litellm.types.llms.openai import AllMessageValues
@@ -140,7 +141,7 @@ class GDCGeminiConfig(OpenAILikeChatConfig):
project = self._validate_path_id(project, "vertex_project", model)
location = self._validate_path_id(location, "vertex_location", model)
- return f"{api_base}/v1/projects/{project}/locations/{location}/chat/completions"
+ return join_api_base(api_base, f"v1/projects/{project}/locations/{location}/chat/completions")
def _read_env_bool(self, val: bool | str | None, env_var: str, default: bool = True) -> bool | str:
def _parse(s: str) -> bool | str:
diff --git a/litellm/llms/gemini/agents/transformation.py b/litellm/llms/gemini/agents/transformation.py
index cfdb55f2048..c19bd9ce87c 100644
--- a/litellm/llms/gemini/agents/transformation.py
+++ b/litellm/llms/gemini/agents/transformation.py
@@ -15,6 +15,7 @@ from typing import Any, Final
import httpx
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.agents.transformation import BaseAgentsAPIConfig
from litellm.llms.gemini.common_utils import GeminiError, GeminiModelInfo
from litellm.types.agents import (
@@ -67,7 +68,7 @@ class GeminiAgentsConfig(BaseAgentsAPIConfig):
return "v1beta"
def _base_url(self, api_base: str | None) -> str:
- return f"{GeminiModelInfo.get_api_base(api_base)}/{self.api_version}"
+ return join_api_base(GeminiModelInfo.get_api_base(api_base), self.api_version)
# ------------------------------------------------------------------ #
# Shared helpers #
@@ -90,7 +91,7 @@ class GeminiAgentsConfig(BaseAgentsAPIConfig):
api_base: str | None,
litellm_params: Mapping[str, object],
) -> str:
- return f"{self._base_url(api_base)}/agents"
+ return join_api_base(self._base_url(api_base), "agents")
def validate_environment(
self,
@@ -177,7 +178,7 @@ class GeminiAgentsConfig(BaseAgentsAPIConfig):
api_base: str | None,
litellm_params: Mapping[str, object],
) -> tuple[str, dict[str, object]]:
- url: Final = f"{self._base_url(api_base)}/agents"
+ url: Final = join_api_base(self._base_url(api_base), "agents")
params: Final[dict[str, object]] = {}
if litellm_params.get("page_size"):
params["pageSize"] = litellm_params["page_size"]
@@ -210,7 +211,7 @@ class GeminiAgentsConfig(BaseAgentsAPIConfig):
api_base: str | None,
litellm_params: Mapping[str, object],
) -> tuple[str, dict[str, object]]:
- url: Final = f"{self._base_url(api_base)}/agents/{name}"
+ url: Final = join_api_base(self._base_url(api_base), f"agents/{name}")
return url, {}
def transform_get_response(
@@ -239,7 +240,7 @@ class GeminiAgentsConfig(BaseAgentsAPIConfig):
api_base: str | None,
litellm_params: Mapping[str, object],
) -> str:
- return f"{self._base_url(api_base)}/agents/{name}"
+ return join_api_base(self._base_url(api_base), f"agents/{name}")
def transform_delete_response(
self,
@@ -265,7 +266,7 @@ class GeminiAgentsConfig(BaseAgentsAPIConfig):
api_base: str | None,
litellm_params: Mapping[str, object],
) -> tuple[str, dict[str, object]]:
- url: Final = f"{self._base_url(api_base)}/agents/{name}/versions"
+ url: Final = join_api_base(self._base_url(api_base), f"agents/{name}/versions")
params: Final[dict[str, object]] = {}
if litellm_params.get("page_size"):
params["pageSize"] = litellm_params["page_size"]
diff --git a/litellm/llms/gemini/audio_transcription/transformation.py b/litellm/llms/gemini/audio_transcription/transformation.py
index 48a345a4940..ea116a7e9de 100644
--- a/litellm/llms/gemini/audio_transcription/transformation.py
+++ b/litellm/llms/gemini/audio_transcription/transformation.py
@@ -9,6 +9,7 @@ from litellm.litellm_core_utils.audio_utils.utils import (
normalize_transcription_language_to_bcp47,
process_audio_file,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.audio_transcription.transformation import (
AudioTranscriptionRequestData,
BaseAudioTranscriptionConfig,
@@ -104,8 +105,7 @@ class GeminiAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
litellm_params: Mapping[str, object],
stream: bool | None = None,
) -> str:
- resolved_api_base: Final = GeminiModelInfo.get_api_base(api_base)
- return f"{resolved_api_base}/v1beta/interactions"
+ return join_api_base(GeminiModelInfo.get_api_base(api_base), "v1beta/interactions")
def transform_audio_transcription_request(
self,
diff --git a/litellm/llms/gemini/common_utils.py b/litellm/llms/gemini/common_utils.py
index 8ee7a463828..13df3921d13 100644
--- a/litellm/llms/gemini/common_utils.py
+++ b/litellm/llms/gemini/common_utils.py
@@ -11,6 +11,7 @@ from pydantic import JsonValue, TypeAdapter
import litellm
from litellm.constants import DEFAULT_MAX_RECURSE_DEPTH
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.base_utils import BaseLLMModelInfo, BaseTokenCounter
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.secret_managers.main import get_secret_str
@@ -377,7 +378,7 @@ class GeminiModelInfo(BaseLLMModelInfo):
return "v1beta"
@staticmethod
- def get_api_base(api_base: str | None = None) -> str | None:
+ def get_api_base(api_base: str | None = None) -> str:
return api_base or get_secret_str("GEMINI_API_BASE") or "https://generativelanguage.googleapis.com"
@staticmethod
@@ -400,13 +401,13 @@ class GeminiModelInfo(BaseLLMModelInfo):
api_base = GeminiModelInfo.get_api_base(api_base)
api_key = GeminiModelInfo.get_api_key(api_key)
endpoint: Final = f"/{self.api_version}/models"
- if api_base is None or api_key is None:
+ if api_key is None:
raise ValueError(
"GEMINI_API_BASE or GEMINI_API_KEY/GOOGLE_API_KEY is not set. Please set the environment variable, to query Gemini's `/models` endpoint."
)
response: Final = litellm.module_level_client.get(
- url=f"{api_base}{endpoint}",
+ url=join_api_base(api_base, endpoint),
headers={"x-goog-api-key": api_key},
)
diff --git a/litellm/llms/gemini/count_tokens/handler.py b/litellm/llms/gemini/count_tokens/handler.py
index 1c9d09330a8..4ed47323d4f 100644
--- a/litellm/llms/gemini/count_tokens/handler.py
+++ b/litellm/llms/gemini/count_tokens/handler.py
@@ -6,6 +6,7 @@ from pydantic import ConfigDict, RootModel, with_config
from typing_extensions import NotRequired, ReadOnly, TypedDict
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.custom_httpx.http_handler import AsyncHTTPHandler, get_async_httpx_client
from litellm.types.utils import LlmProviders
@@ -65,7 +66,7 @@ class GoogleAIStudioTokenCounter:
Construct the URL for the Google Gen AI Studio countTokens endpoint.
"""
base_url: Final = api_base or "https://generativelanguage.googleapis.com"
- return f"{base_url}/v1beta/models/{model}:countTokens"
+ return join_api_base(base_url, f"v1beta/models/{model}:countTokens")
def __init__(self, client: AsyncHTTPHandler | None = None) -> None:
super().__init__()
diff --git a/litellm/llms/gemini/files/transformation.py b/litellm/llms/gemini/files/transformation.py
index 2c62e04c5a3..f9a7e230ea9 100644
--- a/litellm/llms/gemini/files/transformation.py
+++ b/litellm/llms/gemini/files/transformation.py
@@ -15,7 +15,7 @@ from typing_extensions import ReadOnly, Required
from litellm._logging import verbose_logger
from litellm.litellm_core_utils.prompt_templates.common_utils import extract_file_data
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.files.transformation import (
BaseFilesConfig,
LiteLLMLoggingObj,
@@ -106,8 +106,7 @@ class GoogleAIStudioFilesHandler(GeminiModelInfo, BaseFilesConfig):
if not final_api_key:
raise ValueError("api_key is required")
- url: Final = f"{api_base}/{endpoint}"
- return url
+ return join_api_base(api_base, endpoint)
def get_supported_openai_params(self, model: str) -> list[OpenAICreateFileRequestOptionalParams]:
return []
@@ -235,10 +234,11 @@ class GoogleAIStudioFilesHandler(GeminiModelInfo, BaseFilesConfig):
file_part: Final = self._normalize_gemini_file_id(file_id)
- api_base = self.get_api_base(litellm_params.get("api_base")) or "https://generativelanguage.googleapis.com"
- api_base = api_base.rstrip("/")
+ api_base: Final = (
+ self.get_api_base(litellm_params.get("api_base")) or "https://generativelanguage.googleapis.com"
+ )
- url: Final = f"{api_base}/v1beta/{file_part}"
+ url: Final = join_api_base(api_base, f"v1beta/{file_part}")
# API key is passed via x-goog-api-key header (set in validate_environment)
return url, {}
@@ -345,7 +345,7 @@ class GoogleAIStudioFilesHandler(GeminiModelInfo, BaseFilesConfig):
file_name: Final = self._normalize_gemini_file_id(file_id)
# Construct the delete URL
- url: Final = f"{api_base}/v1beta/{file_name}"
+ url: Final = join_api_base(api_base, f"v1beta/{file_name}")
# Add API key as header (Google AI Studio uses x-goog-api-key header)
params: Final[dict] = {}
diff --git a/litellm/llms/gemini/image_edit/transformation.py b/litellm/llms/gemini/image_edit/transformation.py
index 31d3963c70c..450c38a2206 100644
--- a/litellm/llms/gemini/image_edit/transformation.py
+++ b/litellm/llms/gemini/image_edit/transformation.py
@@ -7,6 +7,7 @@ import httpx
from httpx._types import RequestFiles
from litellm.images.utils import ImageEditRequestUtils
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_edit.transformation import BaseImageEditConfig
from litellm.llms.gemini.common_utils import (
get_gemini_image_generation_config,
@@ -79,9 +80,8 @@ class GeminiImageEditConfig(BaseImageEditConfig):
api_base: str | None,
litellm_params: dict,
) -> str:
- base_url = api_base or get_secret_str("GEMINI_API_BASE") or self.DEFAULT_BASE_URL
- base_url = base_url.rstrip("/")
- return f"{base_url}/models/{model}:generateContent"
+ base_url: Final = api_base or get_secret_str("GEMINI_API_BASE") or self.DEFAULT_BASE_URL
+ return join_api_base(base_url, f"models/{model}:generateContent")
def transform_image_edit_request(
self,
diff --git a/litellm/llms/gemini/image_generation/transformation.py b/litellm/llms/gemini/image_generation/transformation.py
index fe82a9cae81..1aa3d074b1a 100644
--- a/litellm/llms/gemini/image_generation/transformation.py
+++ b/litellm/llms/gemini/image_generation/transformation.py
@@ -4,6 +4,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
from pydantic import ConfigDict, TypeAdapter
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
)
@@ -88,14 +89,12 @@ class GoogleImageGenConfig(BaseImageGenerationConfig):
"""
complete_url: str = api_base or get_secret_str("GEMINI_API_BASE") or self.DEFAULT_BASE_URL
- complete_url = complete_url.rstrip("/")
-
# Gemini Flash Image Preview models use generateContent endpoint
if is_gemini_image_model(model):
- complete_url = f"{complete_url}/models/{model}:generateContent"
+ complete_url = join_api_base(complete_url, f"models/{model}:generateContent")
else:
# All other Imagen models use predict endpoint
- complete_url = f"{complete_url}/models/{model}:predict"
+ complete_url = join_api_base(complete_url, f"models/{model}:predict")
return complete_url
diff --git a/litellm/llms/gemini/interactions/transformation.py b/litellm/llms/gemini/interactions/transformation.py
index a58037f691f..6f3e9a56b92 100644
--- a/litellm/llms/gemini/interactions/transformation.py
+++ b/litellm/llms/gemini/interactions/transformation.py
@@ -16,7 +16,7 @@ from typing_extensions import ReadOnly, TypedDict
from litellm._logging import verbose_logger
from litellm.litellm_core_utils.core_helpers import process_response_headers
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.interactions.transformation import BaseInteractionsAPIConfig
from litellm.llms.gemini.common_utils import GeminiError, GeminiModelInfo
from litellm.types.interactions import (
@@ -153,10 +153,8 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig):
if not api_key:
raise ValueError("Google API key is required. Set GOOGLE_API_KEY or GEMINI_API_KEY environment variable.")
- if stream:
- return f"{api_base}/{self.api_version}/interactions?alt=sse"
-
- return f"{api_base}/{self.api_version}/interactions"
+ collection_url: Final = join_api_base(api_base, f"{self.api_version}/interactions")
+ return f"{collection_url}?alt=sse" if stream else collection_url
def transform_request(
self,
@@ -304,7 +302,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig):
raise ValueError("Google API key is required")
encoded_interaction_id: Final = encode_url_path_segment(interaction_id, field_name="interaction_id")
return (
- f"{resolved_api_base}/{self.api_version}/interactions/{encoded_interaction_id}",
+ join_api_base(resolved_api_base, f"{self.api_version}/interactions/{encoded_interaction_id}"),
{},
)
@@ -344,7 +342,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig):
raise ValueError("Google API key is required")
encoded_interaction_id: Final = encode_url_path_segment(interaction_id, field_name="interaction_id")
return (
- f"{resolved_api_base}/{self.api_version}/interactions/{encoded_interaction_id}",
+ join_api_base(resolved_api_base, f"{self.api_version}/interactions/{encoded_interaction_id}"),
{},
)
@@ -375,7 +373,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig):
raise ValueError("Google API key is required")
encoded_interaction_id: Final = encode_url_path_segment(interaction_id, field_name="interaction_id")
return (
- f"{resolved_api_base}/{self.api_version}/interactions/{encoded_interaction_id}/cancel",
+ join_api_base(resolved_api_base, f"{self.api_version}/interactions/{encoded_interaction_id}/cancel"),
{},
)
diff --git a/litellm/llms/gemini/realtime/transformation.py b/litellm/llms/gemini/realtime/transformation.py
index f4e2941670a..9af7edeac86 100644
--- a/litellm/llms/gemini/realtime/transformation.py
+++ b/litellm/llms/gemini/realtime/transformation.py
@@ -13,6 +13,7 @@ import litellm
from litellm import verbose_logger
from litellm._uuid import uuid
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.realtime.transformation import BaseRealtimeConfig
from litellm.llms.vertex_ai.gemini.vertex_and_google_ai_studio_gemini import (
VertexGeminiConfig,
@@ -212,7 +213,10 @@ class GeminiRealtimeConfig(BaseRealtimeConfig):
# so the API key must remain as a query parameter here. This is an accepted
# limitation; httpx is not used for WebSocket so MaskedHTTPStatusError
# already covers the main leak vector.
- return f"{api_base}/ws/google.ai.generativelanguage.v1beta.GenerativeService.BidiGenerateContent?key={api_key}"
+ live_url: Final = join_api_base(
+ api_base, "ws/google.ai.generativelanguage.v1beta.GenerativeService.BidiGenerateContent"
+ )
+ return f"{live_url}?key={api_key}"
def map_model_turn_event(self, model_turn: HttpxContentType) -> OpenAIRealtimeEventTypes:
"""
diff --git a/litellm/llms/gemini/vector_stores/transformation.py b/litellm/llms/gemini/vector_stores/transformation.py
index 34bd9660c43..11e619eb4c7 100644
--- a/litellm/llms/gemini/vector_stores/transformation.py
+++ b/litellm/llms/gemini/vector_stores/transformation.py
@@ -11,6 +11,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
from typing_extensions import ReadOnly, TypedDict
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.vector_store.transformation import BaseVectorStoreConfig
from litellm.llms.gemini.common_utils import (
GeminiError,
@@ -141,15 +142,7 @@ class GeminiVectorStoreConfig(BaseVectorStoreConfig):
Note: This returns the base URL WITHOUT the API key.
The API key will be appended to specific endpoint URLs in the transform methods.
"""
- if api_base is None:
- api_base = GeminiModelInfo.get_api_base()
-
- if api_base is None:
- raise ValueError("GEMINI_API_BASE is not set")
-
- # Ensure we're using the v1beta version for File Search
- api_version: Final = "v1beta"
- return f"{api_base}/{api_version}"
+ return join_api_base(GeminiModelInfo.get_api_base(api_base), "v1beta")
def get_error_class(self, error_message: str, status_code: int, headers: dict | httpx.Headers) -> GeminiError:
"""Return Gemini-specific error class."""
@@ -187,7 +180,7 @@ class GeminiVectorStoreConfig(BaseVectorStoreConfig):
api_key: Final = litellm_params.get("api_key") or GeminiModelInfo.get_api_key()
if not api_key:
raise ValueError("GEMINI_API_KEY or GOOGLE_API_KEY is required")
- url: Final = f"{api_base}/models/{model}:generateContent"
+ url: Final = join_api_base(api_base, f"models/{model}:generateContent")
# Build file_search tool configuration (using snake_case as per Gemini docs)
file_search_config: Final[dict[str, object]] = {"file_search_store_names": [vector_store_id]}
@@ -328,7 +321,7 @@ class GeminiVectorStoreConfig(BaseVectorStoreConfig):
"""
Transform create request to Gemini's fileSearchStores format.
"""
- url: Final = f"{api_base}/fileSearchStores"
+ url: Final = join_api_base(api_base, "fileSearchStores")
# API key is passed via x-goog-api-key header (set in validate_environment)
diff --git a/litellm/llms/gemini/videos/transformation.py b/litellm/llms/gemini/videos/transformation.py
index 53ca5ed4adc..0656095927c 100644
--- a/litellm/llms/gemini/videos/transformation.py
+++ b/litellm/llms/gemini/videos/transformation.py
@@ -8,6 +8,7 @@ from httpx._types import RequestFiles
import litellm
from litellm.constants import DEFAULT_GOOGLE_VIDEO_DURATION_SECONDS
from litellm.images.utils import ImageEditRequestUtils
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.videos.transformation import BaseVideoConfig
from litellm.llms.vertex_ai.videos.transformation import veo_video_count_from_parameters
from litellm.secret_managers.main import get_secret_str
@@ -241,7 +242,7 @@ class GeminiVideoConfig(BaseVideoConfig):
return api_base.rstrip("/")
model_name: Final = model.replace("gemini/", "")
- url: Final = f"{api_base.rstrip('/')}/v1beta/models/{model_name}:predictLongRunning"
+ url: Final = join_api_base(api_base, f"v1beta/models/{model_name}:predictLongRunning")
return url
@@ -377,7 +378,7 @@ class GeminiVideoConfig(BaseVideoConfig):
GET https://generativelanguage.googleapis.com/v1beta/{operation_name}
"""
operation_name: Final = extract_original_video_id(video_id)
- url: Final = f"{api_base.rstrip('/')}/v1beta/{operation_name}"
+ url: Final = join_api_base(api_base, f"v1beta/{operation_name}")
params: Final[dict[str, object]] = {}
return url, params
@@ -451,7 +452,7 @@ class GeminiVideoConfig(BaseVideoConfig):
"""
operation_name: Final = extract_original_video_id(video_id)
- status_url: Final = f"{api_base.rstrip('/')}/v1beta/{operation_name}"
+ status_url: Final = join_api_base(api_base, f"v1beta/{operation_name}")
client: Final = litellm.module_level_client
status_response: Final = client.get(url=status_url, headers=headers)
status_response.raise_for_status()
diff --git a/litellm/llms/gigachat/chat/transformation.py b/litellm/llms/gigachat/chat/transformation.py
index 1fe7de0ba85..936e6339865 100644
--- a/litellm/llms/gigachat/chat/transformation.py
+++ b/litellm/llms/gigachat/chat/transformation.py
@@ -17,6 +17,7 @@ import httpx
from pydantic import TypeAdapter
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.llms.gigachat.utils import convert_usage, get_api_base
from litellm.secret_managers.main import get_secret_str
@@ -108,8 +109,7 @@ class GigaChatConfig(BaseConfig):
stream: bool | None = None,
) -> str:
"""Get complete API URL for chat completions."""
- base: Final = get_api_base(api_base)
- return f"{base}/chat/completions"
+ return join_api_base(get_api_base(api_base), "chat/completions")
def validate_environment(
self,
diff --git a/litellm/llms/gigachat/embedding/transformation.py b/litellm/llms/gigachat/embedding/transformation.py
index e0e746427ad..30ca7147865 100644
--- a/litellm/llms/gigachat/embedding/transformation.py
+++ b/litellm/llms/gigachat/embedding/transformation.py
@@ -16,6 +16,7 @@ from pydantic import ConfigDict, TypeAdapter
from litellm import LlmProviders
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.llms.gigachat.utils import get_api_base
@@ -98,8 +99,7 @@ class GigaChatEmbeddingConfig(BaseEmbeddingConfig):
stream: bool | None = None,
) -> str:
"""Get the complete URL for embeddings endpoint."""
- base: Final = get_api_base(api_base)
- return f"{base}/embeddings"
+ return join_api_base(get_api_base(api_base), "embeddings")
def transform_embedding_request(
self,
diff --git a/litellm/llms/gigachat/file_handler.py b/litellm/llms/gigachat/file_handler.py
index c051d693720..aff0465d1ec 100644
--- a/litellm/llms/gigachat/file_handler.py
+++ b/litellm/llms/gigachat/file_handler.py
@@ -13,6 +13,7 @@ from collections.abc import Mapping
from typing import Final
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.custom_httpx.http_handler import (
get_async_httpx_client,
get_httpx_client,
@@ -125,8 +126,7 @@ def upload_file_sync(
access_token: Final = get_access_token(credentials=credentials, litellm_params=litellm_params)
# Upload to GigaChat
- base_url: Final = get_api_base(api_base)
- upload_url: Final = f"{base_url}/files"
+ upload_url: Final = join_api_base(get_api_base(api_base), "files")
client: Final = get_httpx_client(params={"ssl_verify": False})
response: Final = client.post(
@@ -191,8 +191,7 @@ async def upload_file_async(
access_token: Final = await get_access_token_async(credentials=credentials, litellm_params=litellm_params)
# Upload to GigaChat
- base_url: Final = get_api_base(api_base)
- upload_url: Final = f"{base_url}/files"
+ upload_url: Final = join_api_base(get_api_base(api_base), "files")
client: Final = get_async_httpx_client(
llm_provider=LlmProviders.GIGACHAT,
diff --git a/litellm/llms/gigachat/passthrough/transformation.py b/litellm/llms/gigachat/passthrough/transformation.py
index f16d263fd4d..872eee6dbf9 100644
--- a/litellm/llms/gigachat/passthrough/transformation.py
+++ b/litellm/llms/gigachat/passthrough/transformation.py
@@ -40,7 +40,7 @@ class GigaChatPassthroughConfig(BasePassthroughConfig):
if base_target_url is None:
raise Exception("GigaChat api base not found")
- complete_url: Final = f"{base_target_url}/{endpoint.lstrip('/')}"
+ complete_url: Final = f"{base_target_url.rstrip('/')}/{endpoint.lstrip('/')}"
return (
httpx.URL(complete_url),
@@ -193,8 +193,7 @@ class GigaChatPassthroughConfig(BasePassthroughConfig):
@staticmethod
def get_api_base(api_base: str | None = None) -> str | None:
- resolved_api_base: Final = api_base or get_secret_str("GIGACHAT_API_BASE") or GIGACHAT_BASE_URL
- return resolved_api_base.rstrip("/") if resolved_api_base else None
+ return api_base or get_secret_str("GIGACHAT_API_BASE") or GIGACHAT_BASE_URL
@staticmethod
def get_api_key(
diff --git a/litellm/llms/gigachat/utils.py b/litellm/llms/gigachat/utils.py
index ab1890ad73b..55ca4260221 100644
--- a/litellm/llms/gigachat/utils.py
+++ b/litellm/llms/gigachat/utils.py
@@ -21,5 +21,5 @@ def convert_usage(usage_data: Mapping[str, int]) -> Usage:
)
-def get_api_base(api_base: str | None = None) -> str | None:
+def get_api_base(api_base: str | None = None) -> str:
return (api_base or get_secret_str("GIGACHAT_API_BASE") or GIGACHAT_BASE_URL).rstrip("/")
diff --git a/litellm/llms/github_copilot/embedding/transformation.py b/litellm/llms/github_copilot/embedding/transformation.py
index 7f9209afd65..f16564ba505 100644
--- a/litellm/llms/github_copilot/embedding/transformation.py
+++ b/litellm/llms/github_copilot/embedding/transformation.py
@@ -20,6 +20,7 @@ from pydantic import ConfigDict, TypeAdapter
from litellm._logging import verbose_logger
from litellm.exceptions import AuthenticationError
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.types.llms.openai import AllEmbeddingInputValues
@@ -121,7 +122,7 @@ class GithubCopilotEmbeddingConfig(BaseEmbeddingConfig):
user_session: Final = require_github_copilot_user_session(
cast("dict[str, object]", litellm_params) # cast-ok: litellm_params arrives as an untyped request dict
)
- effective_api_base = (
+ effective_api_base: Final = (
(user_session.api_base if user_session is not None else None)
or api_base
or self.authenticator.get_api_base()
@@ -129,11 +130,7 @@ class GithubCopilotEmbeddingConfig(BaseEmbeddingConfig):
or DEFAULT_GITHUB_COPILOT_API_BASE
)
- # Remove trailing slashes
- effective_api_base = effective_api_base.rstrip("/")
-
- # Return the embeddings endpoint
- return f"{effective_api_base}/embeddings"
+ return join_api_base(effective_api_base, "embeddings")
def transform_embedding_request(
self,
diff --git a/litellm/llms/github_copilot/messages/transformation.py b/litellm/llms/github_copilot/messages/transformation.py
index be138755f85..56c3693a02c 100644
--- a/litellm/llms/github_copilot/messages/transformation.py
+++ b/litellm/llms/github_copilot/messages/transformation.py
@@ -1,6 +1,7 @@
from typing import Any, Final, cast # noqa: TID251 # narrows untyped request dicts at the session boundary
from litellm.exceptions import AuthenticationError
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.anthropic.pass_through.messages.transformation import (
AnthropicMessagesConfig,
)
@@ -140,5 +141,5 @@ class GithubCopilotAnthropicMessagesConfig(AnthropicMessagesConfig):
or api_base
or self.authenticator.get_api_base()
or DEFAULT_GITHUB_COPILOT_API_BASE
- ).rstrip("/")
- return resolved if resolved.endswith("/v1/messages") else f"{resolved}/v1/messages"
+ )
+ return ensure_api_route(resolved, "v1/messages")
diff --git a/litellm/llms/github_copilot/responses/transformation.py b/litellm/llms/github_copilot/responses/transformation.py
index d468889b626..bf0b8c79068 100644
--- a/litellm/llms/github_copilot/responses/transformation.py
+++ b/litellm/llms/github_copilot/responses/transformation.py
@@ -20,6 +20,7 @@ import litellm
from litellm._logging import verbose_logger
from litellm.constants import DEFAULT_MAX_RECURSE_DEPTH
from litellm.exceptions import AuthenticationError
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.types.llms.openai import (
ResponseInputParam,
@@ -266,7 +267,7 @@ class GithubCopilotResponsesAPIConfig(OpenAIResponsesAPIConfig):
user_session: Final = require_github_copilot_user_session(
cast("dict[str, object]", litellm_params) # cast-ok: litellm_params arrives as an untyped request dict
)
- effective_api_base = (
+ effective_api_base: Final = (
(user_session.api_base if user_session is not None else None)
or api_base
or self.authenticator.get_api_base()
@@ -274,11 +275,7 @@ class GithubCopilotResponsesAPIConfig(OpenAIResponsesAPIConfig):
or DEFAULT_GITHUB_COPILOT_API_BASE
)
- # Remove trailing slashes
- effective_api_base = effective_api_base.rstrip("/")
-
- # Return the responses endpoint
- return f"{effective_api_base}/responses"
+ return join_api_base(effective_api_base, "responses")
def _handle_reasoning_item(self, item: dict[str, object]) -> dict[str, object]:
"""
diff --git a/litellm/llms/gradient_ai/chat/transformation.py b/litellm/llms/gradient_ai/chat/transformation.py
index fcee55a27e9..0ba5ca45c2a 100644
--- a/litellm/llms/gradient_ai/chat/transformation.py
+++ b/litellm/llms/gradient_ai/chat/transformation.py
@@ -1,5 +1,6 @@
from typing import Final, Literal
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import (
AllMessageValues,
@@ -105,10 +106,10 @@ class GradientAIConfig(OpenAILikeChatConfig):
) -> str:
clean_api_base: Final = (api_base or "").rstrip("/")
if clean_api_base and clean_api_base != GRADIENT_AI_SERVERLESS_ENDPOINT:
- return f"{clean_api_base}/api/v1/chat/completions"
+ return join_api_base(clean_api_base, "api/v1/chat/completions")
agent_endpoint: Final = (get_secret_str("GRADIENT_AI_AGENT_ENDPOINT") or "").rstrip("/")
if agent_endpoint and agent_endpoint != GRADIENT_AI_SERVERLESS_ENDPOINT:
- return f"{agent_endpoint}/api/v1/chat/completions"
+ return join_api_base(agent_endpoint, "api/v1/chat/completions")
return f"{GRADIENT_AI_SERVERLESS_ENDPOINT}/v1/chat/completions"
def _get_openai_compatible_provider_info(
diff --git a/litellm/llms/heroku/chat/transformation.py b/litellm/llms/heroku/chat/transformation.py
index acd630c85d0..002078dd04b 100644
--- a/litellm/llms/heroku/chat/transformation.py
+++ b/litellm/llms/heroku/chat/transformation.py
@@ -6,11 +6,12 @@ this is OpenAI compatible - no translation needed / occurs
import os
from collections.abc import Coroutine
-from typing import Final, Literal, overload
+from typing import Literal, overload
from litellm.litellm_core_utils.prompt_templates.common_utils import (
handle_messages_with_content_list_to_str_conversion,
)
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.openai.chat.gpt_transformation import OpenAIGPTConfig
from litellm.types.llms.openai import AllMessageValues
@@ -78,9 +79,4 @@ class HerokuChatConfig(OpenAIGPTConfig):
"No api base was set. Please provide an api_base, or set the HEROKU_API_BASE environment variable."
)
- clean_api_base: Final = resolved_base.rstrip("/")
-
- if not clean_api_base.endswith("/v1/chat/completions"):
- return f"{clean_api_base}/v1/chat/completions"
-
- return clean_api_base
+ return ensure_api_route(resolved_base, "v1/chat/completions")
diff --git a/litellm/llms/hosted_vllm/embedding/transformation.py b/litellm/llms/hosted_vllm/embedding/transformation.py
index 3b54145a67b..d72a8d06f21 100644
--- a/litellm/llms/hosted_vllm/embedding/transformation.py
+++ b/litellm/llms/hosted_vllm/embedding/transformation.py
@@ -11,6 +11,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.secret_managers.main import get_secret_str
@@ -81,14 +82,7 @@ class HostedVLLMEmbeddingConfig(BaseEmbeddingConfig):
if api_base is None:
raise ValueError("api_base is required for hosted_vllm embeddings")
- # Remove trailing slashes
- api_base = api_base.rstrip("/")
-
- # Ensure the URL ends with /embeddings
- if not api_base.endswith("/embeddings"):
- api_base = f"{api_base}/embeddings"
-
- return api_base
+ return ensure_api_route(api_base, "embeddings")
def transform_embedding_request(
self,
diff --git a/litellm/llms/hosted_vllm/image_edit/transformation.py b/litellm/llms/hosted_vllm/image_edit/transformation.py
index 6804d3a0fb8..8e05b6dfb46 100644
--- a/litellm/llms/hosted_vllm/image_edit/transformation.py
+++ b/litellm/llms/hosted_vllm/image_edit/transformation.py
@@ -1,5 +1,6 @@
from typing import Final
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.image_edit.transformation import OpenAIImageEditConfig
from litellm.secret_managers.main import get_secret_str
@@ -39,5 +40,5 @@ class HostedVLLMImageEditConfig(OpenAIImageEditConfig):
)
trimmed: Final = resolved_api_base.rstrip("/")
if trimmed.endswith("/v1"):
- return f"{trimmed}/images/edits"
- return f"{trimmed}/v1/images/edits"
+ return join_api_base(trimmed, "images/edits")
+ return join_api_base(trimmed, "v1/images/edits")
diff --git a/litellm/llms/hosted_vllm/rerank/transformation.py b/litellm/llms/hosted_vllm/rerank/transformation.py
index a463fc1d3c9..4a5729945b4 100644
--- a/litellm/llms/hosted_vllm/rerank/transformation.py
+++ b/litellm/llms/hosted_vllm/rerank/transformation.py
@@ -12,6 +12,7 @@ from pydantic import ValidationError
from litellm._uuid import uuid
from litellm.exceptions import UnsupportedParamsError
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.rerank.transformation import BaseRerankConfig
from litellm.secret_managers.main import get_secret_str
@@ -62,7 +63,7 @@ class HostedVLLMRerankConfig(BaseRerankConfig):
if api_base.endswith("/v1/rerank"):
api_base = api_base.replace("/v1/rerank", "/rerank")
elif not api_base.endswith("/rerank"):
- api_base = f"{api_base}/rerank"
+ api_base = join_api_base(api_base, "rerank")
return api_base
raise ValueError("api_base must be provided for Hosted VLLM rerank")
diff --git a/litellm/llms/hosted_vllm/responses/transformation.py b/litellm/llms/hosted_vllm/responses/transformation.py
index 20d755a2273..8489bc0da84 100644
--- a/litellm/llms/hosted_vllm/responses/transformation.py
+++ b/litellm/llms/hosted_vllm/responses/transformation.py
@@ -8,6 +8,7 @@ the chat completions → responses conversion pipeline.
from typing import Final
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.router import GenericLiteLLMParams
@@ -64,9 +65,9 @@ class HostedVLLMResponsesAPIConfig(OpenAIResponsesAPIConfig):
# If api_base already ends with /v1, append /responses
# Otherwise append /v1/responses
if api_base.endswith("/v1"):
- return f"{api_base}/responses"
+ return join_api_base(api_base, "responses")
- return f"{api_base}/v1/responses"
+ return join_api_base(api_base, "v1/responses")
def supports_native_websocket(self) -> bool:
"""Hosted vLLM does not support native WebSocket for Responses API"""
diff --git a/litellm/llms/hosted_vllm/transcriptions/transformation.py b/litellm/llms/hosted_vllm/transcriptions/transformation.py
index 9a8c67fadc0..785ccd52a69 100644
--- a/litellm/llms/hosted_vllm/transcriptions/transformation.py
+++ b/litellm/llms/hosted_vllm/transcriptions/transformation.py
@@ -6,6 +6,7 @@ from typing import Final
import httpx
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.audio_transcription.transformation import (
AudioTranscriptionRequestData,
)
@@ -40,11 +41,7 @@ class HostedVLLMAudioTranscriptionConfig(OpenAIWhisperAudioTranscriptionConfig):
stream: bool | None = None,
) -> str:
if api_base:
- # Remove trailing slashes and ensure clean base URL
- api_base = api_base.rstrip("/")
- if not api_base.endswith("/v1/audio/transcriptions"):
- api_base = f"{api_base}/v1/audio/transcriptions"
- return api_base
+ return ensure_api_route(api_base, "v1/audio/transcriptions")
raise ValueError("api_base must be provided for Hosted VLLM rerank")
def transform_audio_transcription_request(
diff --git a/litellm/llms/hosted_vllm/videos/transformation.py b/litellm/llms/hosted_vllm/videos/transformation.py
index 22e4f4876fb..30ae6835263 100644
--- a/litellm/llms/hosted_vllm/videos/transformation.py
+++ b/litellm/llms/hosted_vllm/videos/transformation.py
@@ -10,7 +10,7 @@ from urllib.parse import urlparse
from httpx._types import FileTypes, RequestFiles
from litellm.images.utils import ImageEditRequestUtils
-from litellm.litellm_core_utils.url_utils import SSRFError, validate_url
+from litellm.litellm_core_utils.url_utils import SSRFError, join_api_base, validate_url
from litellm.llms.openai.videos.transformation import OpenAIVideoConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.router import GenericLiteLLMParams
@@ -177,8 +177,8 @@ class HostedVLLMVideoConfig(OpenAIVideoConfig):
)
trimmed: Final = resolved_api_base.rstrip("/")
if trimmed.endswith("/v1"):
- return f"{trimmed}/videos"
- return f"{trimmed}/v1/videos"
+ return join_api_base(trimmed, "videos")
+ return join_api_base(trimmed, "v1/videos")
def transform_video_create_request(
self,
diff --git a/litellm/llms/huggingface/chat/transformation.py b/litellm/llms/huggingface/chat/transformation.py
index b45c706f8bf..3fcb738eacb 100644
--- a/litellm/llms/huggingface/chat/transformation.py
+++ b/litellm/llms/huggingface/chat/transformation.py
@@ -4,6 +4,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.types.llms.openai import AllMessageValues, ChatCompletionRequest
if TYPE_CHECKING:
@@ -29,11 +30,11 @@ def _build_chat_completion_url(model_url: str) -> str:
# Append /chat/completions if not already present
if model_url.endswith("/v1"):
- model_url += "/chat/completions"
+ model_url = join_api_base(model_url, "chat/completions")
# Append /v1/chat/completions if not already present
if not model_url.endswith("/chat/completions"):
- model_url += "/v1/chat/completions"
+ model_url = join_api_base(model_url, "v1/chat/completions")
return model_url
diff --git a/litellm/llms/huggingface/rerank/transformation.py b/litellm/llms/huggingface/rerank/transformation.py
index 334c60ee848..1217a2540d4 100644
--- a/litellm/llms/huggingface/rerank/transformation.py
+++ b/litellm/llms/huggingface/rerank/transformation.py
@@ -7,6 +7,7 @@ from typing_extensions import TypedDict
import litellm
from litellm._uuid import uuid
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.rerank.transformation import BaseRerankConfig
from litellm.secret_managers.main import get_secret_str
@@ -69,15 +70,7 @@ class HuggingFaceRerankConfig(BaseRerankConfig):
"""
Get the complete URL for the API call, including the /rerank suffix if necessary.
"""
- # Get base URL from api_base or default
- base_url = self.get_api_base(model=model, api_base=api_base)
-
- # Remove trailing slashes and ensure we have the /rerank endpoint
- base_url = base_url.rstrip("/")
- if not base_url.endswith("/rerank"):
- base_url = f"{base_url}/rerank"
-
- return base_url
+ return ensure_api_route(self.get_api_base(model=model, api_base=api_base), "rerank")
def get_supported_cohere_rerank_params(self, model: str) -> list:
return [
diff --git a/litellm/llms/infinity/embedding/transformation.py b/litellm/llms/infinity/embedding/transformation.py
index f289918cb3c..a467c56fc74 100644
--- a/litellm/llms/infinity/embedding/transformation.py
+++ b/litellm/llms/infinity/embedding/transformation.py
@@ -5,6 +5,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.secret_managers.main import get_secret_str
@@ -36,11 +37,7 @@ class InfinityEmbeddingConfig(BaseEmbeddingConfig):
) -> str:
if api_base is None:
raise ValueError("api_base is required for Infinity embeddings")
- # Remove trailing slashes and ensure clean base URL
- api_base = api_base.rstrip("/")
- if not api_base.endswith("/embeddings"):
- api_base = f"{api_base}/embeddings"
- return api_base
+ return ensure_api_route(api_base, "embeddings")
def validate_environment(
self,
diff --git a/litellm/llms/infinity/rerank/transformation.py b/litellm/llms/infinity/rerank/transformation.py
index e089b3fecbe..07f3bb2c26b 100644
--- a/litellm/llms/infinity/rerank/transformation.py
+++ b/litellm/llms/infinity/rerank/transformation.py
@@ -13,6 +13,7 @@ from typing_extensions import NotRequired, ReadOnly, TypedDict
import litellm
from litellm._uuid import uuid
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.cohere.rerank.transformation import CohereRerankConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.rerank import (
@@ -61,11 +62,7 @@ class InfinityRerankConfig(CohereRerankConfig):
) -> str:
if api_base is None:
raise ValueError("api_base is required for Infinity rerank")
- # Remove trailing slashes and ensure clean base URL
- api_base = api_base.rstrip("/")
- if not api_base.endswith("/rerank"):
- api_base = f"{api_base}/rerank"
- return api_base
+ return ensure_api_route(api_base, "rerank")
def validate_environment(
self,
diff --git a/litellm/llms/jina_ai/embedding/transformation.py b/litellm/llms/jina_ai/embedding/transformation.py
index 9b156e5ae21..04ff8f941a3 100644
--- a/litellm/llms/jina_ai/embedding/transformation.py
+++ b/litellm/llms/jina_ai/embedding/transformation.py
@@ -15,6 +15,7 @@ from pydantic import ConfigDict, TypeAdapter
from litellm import LlmProviders
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm import BaseEmbeddingConfig
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.secret_managers.main import get_secret_str
@@ -115,12 +116,7 @@ class JinaAIEmbeddingConfig(BaseEmbeddingConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- if api_base:
- clean_api_base: Final = api_base.rstrip("/")
- if not clean_api_base.endswith("/embeddings"):
- return f"{clean_api_base}/embeddings"
- return clean_api_base
- return "https://api.jina.ai/v1/embeddings"
+ return ensure_api_route(api_base or "https://api.jina.ai/v1", "embeddings")
def transform_embedding_request(
self,
diff --git a/litellm/llms/langflow/chat/transformation.py b/litellm/llms/langflow/chat/transformation.py
index 1781082d36f..0938cc0d07b 100644
--- a/litellm/llms/langflow/chat/transformation.py
+++ b/litellm/llms/langflow/chat/transformation.py
@@ -9,6 +9,7 @@ from litellm._logging import verbose_logger
from litellm.litellm_core_utils.prompt_templates.common_utils import (
convert_content_list_to_str,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.types.llms.openai import AllMessageValues
from litellm.types.utils import Choices, Message, ModelResponse, Usage
@@ -107,9 +108,8 @@ class LangFlowConfig(BaseConfig):
"api_base is required for LangFlow. Set it via LANGFLOW_API_BASE env var or api_base parameter."
)
- api_base = api_base.rstrip("/")
flow_id: Final = quote(self._get_flow_id(model, optional_params), safe="")
- return f"{api_base}/api/v1/run/{flow_id}"
+ return join_api_base(api_base, f"api/v1/run/{flow_id}")
def _get_last_user_message(self, messages: list[AllMessageValues]) -> str:
"""Extract the text of the last user message to use as input_value."""
diff --git a/litellm/llms/langgraph/chat/transformation.py b/litellm/llms/langgraph/chat/transformation.py
index 4f93f435957..ce624eecc0e 100644
--- a/litellm/llms/langgraph/chat/transformation.py
+++ b/litellm/llms/langgraph/chat/transformation.py
@@ -18,6 +18,7 @@ from litellm._logging import verbose_logger
from litellm.litellm_core_utils.prompt_templates.common_utils import (
convert_content_list_to_str,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.llms.langgraph.chat.sse_iterator import LangGraphSSEStreamIterator
from litellm.types.llms.openai import AllMessageValues
@@ -116,14 +117,11 @@ class LangGraphConfig(BaseConfig):
"api_base is required for LangGraph. Set it via LANGGRAPH_API_BASE env var or api_base parameter."
)
- # Remove trailing slash if present
- api_base = api_base.rstrip("/")
-
# Choose endpoint based on streaming mode
if stream:
- return f"{api_base}/runs/stream"
+ return join_api_base(api_base, "runs/stream")
else:
- return f"{api_base}/runs/wait"
+ return join_api_base(api_base, "runs/wait")
def _get_assistant_id(self, model: str, optional_params: dict) -> str:
"""
diff --git a/litellm/llms/lemonade/chat/transformation.py b/litellm/llms/lemonade/chat/transformation.py
index a740050dd94..01fdc0b8bf8 100644
--- a/litellm/llms/lemonade/chat/transformation.py
+++ b/litellm/llms/lemonade/chat/transformation.py
@@ -12,6 +12,7 @@ from pydantic import ConfigDict, TypeAdapter
import litellm
from litellm._logging import verbose_logger
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import (
AllMessageValues,
@@ -105,7 +106,7 @@ class LemonadeChatConfig(OpenAILikeChatConfig):
# Getting the list of models from lemonade
try:
response: Final = litellm.module_level_client.get(
- url=f"{api_base}/models",
+ url=join_api_base(api_base, "models"),
headers=self._get_auth_headers(api_key),
)
except Exception as e:
@@ -193,7 +194,7 @@ class LemonadeChatConfig(OpenAILikeChatConfig):
try:
response: Final = litellm.module_level_client.get(
- url=f"{api_base}/models/{encoded_model}",
+ url=join_api_base(str(api_base), f"models/{encoded_model}"),
headers=self._get_auth_headers(api_key),
)
response.raise_for_status()
diff --git a/litellm/llms/linkup/search/transformation.py b/litellm/llms/linkup/search/transformation.py
index e219a227b1f..fd1ca792302 100644
--- a/litellm/llms/linkup/search/transformation.py
+++ b/litellm/llms/linkup/search/transformation.py
@@ -11,6 +11,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
SearchResponse,
@@ -87,11 +88,7 @@ class LinkupSearchConfig(BaseSearchConfig):
"""
Get complete URL for Search endpoint.
"""
- resolved_api_base: Final = api_base or get_secret_str("LINKUP_API_BASE") or self.LINKUP_API_BASE
- clean_api_base: Final = resolved_api_base.rstrip("/")
- if clean_api_base.endswith("/search"):
- return clean_api_base
- return f"{clean_api_base}/search"
+ return ensure_api_route(api_base or get_secret_str("LINKUP_API_BASE") or self.LINKUP_API_BASE, "search")
def transform_search_request(
self,
diff --git a/litellm/llms/litellm_proxy/image_edit/transformation.py b/litellm/llms/litellm_proxy/image_edit/transformation.py
index 9e5a2a2c343..f319e31ff3f 100644
--- a/litellm/llms/litellm_proxy/image_edit/transformation.py
+++ b/litellm/llms/litellm_proxy/image_edit/transformation.py
@@ -1,3 +1,4 @@
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.image_edit.transformation import OpenAIImageEditConfig
from litellm.secret_managers.main import get_secret_str
@@ -21,5 +22,4 @@ class LiteLLMProxyImageEditConfig(OpenAIImageEditConfig):
api_base = api_base or get_secret_str("LITELLM_PROXY_API_BASE")
if api_base is None:
raise ValueError("api_base not set for LiteLLM Proxy route. Set in env via `LITELLM_PROXY_API_BASE`")
- api_base = api_base.rstrip("/")
- return f"{api_base}/images/edits"
+ return join_api_base(api_base, "images/edits")
diff --git a/litellm/llms/litellm_proxy/image_generation/transformation.py b/litellm/llms/litellm_proxy/image_generation/transformation.py
index fcddeabfc85..05e536eece5 100644
--- a/litellm/llms/litellm_proxy/image_generation/transformation.py
+++ b/litellm/llms/litellm_proxy/image_generation/transformation.py
@@ -1,3 +1,4 @@
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.image_generation.gpt_transformation import (
GPTImageGenerationConfig,
)
@@ -33,5 +34,4 @@ class LiteLLMProxyImageGenerationConfig(GPTImageGenerationConfig):
api_base = api_base or get_secret_str("LITELLM_PROXY_API_BASE")
if api_base is None:
raise ValueError("api_base not set for LiteLLM Proxy route. Set in env via `LITELLM_PROXY_API_BASE`")
- api_base = api_base.rstrip("/")
- return f"{api_base}/images/generations"
+ return join_api_base(api_base, "images/generations")
diff --git a/litellm/llms/litellm_proxy/responses/transformation.py b/litellm/llms/litellm_proxy/responses/transformation.py
index 2928b07665a..3092b49b743 100644
--- a/litellm/llms/litellm_proxy/responses/transformation.py
+++ b/litellm/llms/litellm_proxy/responses/transformation.py
@@ -5,6 +5,7 @@ LiteLLM Proxy supports the OpenAI Responses API natively when the underlying mod
This config enables pass-through behavior to the proxy's /v1/responses endpoint.
"""
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.utils import LlmProviders
@@ -40,10 +41,7 @@ class LiteLLMProxyResponsesAPIConfig(OpenAIResponsesAPIConfig):
"Set via api_base parameter or LITELLM_PROXY_API_BASE environment variable"
)
- # Remove trailing slashes
- api_base = api_base.rstrip("/")
-
- return f"{api_base}/responses"
+ return join_api_base(api_base, "responses")
def supports_native_websocket(self) -> bool:
"""LiteLLM Proxy does not support native WebSocket for Responses API"""
diff --git a/litellm/llms/manus/files/transformation.py b/litellm/llms/manus/files/transformation.py
index 94b0e62199d..991aa4cbc32 100644
--- a/litellm/llms/manus/files/transformation.py
+++ b/litellm/llms/manus/files/transformation.py
@@ -21,7 +21,7 @@ from pydantic import ConfigDict, TypeAdapter
import litellm
from litellm._logging import verbose_logger
from litellm.litellm_core_utils.prompt_templates.common_utils import extract_file_data
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.files.transformation import (
BaseFilesConfig,
@@ -140,8 +140,8 @@ class ManusFilesConfig(BaseFilesConfig):
# Manus API uses /v1/files endpoint
if api_base.endswith("/v1"):
- return f"{api_base}/files"
- return f"{api_base}/v1/files"
+ return join_api_base(api_base, "files")
+ return join_api_base(api_base, "v1/files")
def get_error_class(
self,
@@ -423,7 +423,7 @@ class ManusFilesConfig(BaseFilesConfig):
litellm_params=litellm_params,
)
encoded_file_id: Final = encode_url_path_segment(file_id, field_name="file_id")
- return f"{api_base}/{encoded_file_id}/content", {}
+ return join_api_base(api_base, f"{encoded_file_id}/content"), {}
def transform_file_content_response(
self,
diff --git a/litellm/llms/manus/responses/transformation.py b/litellm/llms/manus/responses/transformation.py
index 9785a50bbce..ed06391a260 100644
--- a/litellm/llms/manus/responses/transformation.py
+++ b/litellm/llms/manus/responses/transformation.py
@@ -10,7 +10,7 @@ from litellm.litellm_core_utils.core_helpers import process_response_headers
from litellm.litellm_core_utils.llm_response_utils.convert_dict_to_response import (
safe_convert_created_field,
)
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.openai.common_utils import OpenAIError
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.secret_managers.main import get_secret_str
@@ -119,8 +119,8 @@ class ManusResponsesAPIConfig(OpenAIResponsesAPIConfig):
# Manus API uses /v1/responses endpoint (OpenAI-compatible)
if api_base.endswith("/v1"):
- return f"{api_base}/responses"
- return f"{api_base}/v1/responses"
+ return join_api_base(api_base, "responses")
+ return join_api_base(api_base, "v1/responses")
def transform_responses_api_request(
self,
diff --git a/litellm/llms/milvus/vector_stores/transformation.py b/litellm/llms/milvus/vector_stores/transformation.py
index 70d3649debd..67de3a06af8 100644
--- a/litellm/llms/milvus/vector_stores/transformation.py
+++ b/litellm/llms/milvus/vector_stores/transformation.py
@@ -5,6 +5,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.vector_store.transformation import (
BaseQueryEmbeddingVectorStoreConfig,
VectorStoreEmbeddingExecutor,
@@ -185,7 +186,7 @@ class MilvusVectorStoreConfig(BaseQueryEmbeddingVectorStoreConfig):
}
litellm_logging_obj.model_call_details["input"] = query_text
litellm_logging_obj.model_call_details["embedding_model"] = litellm_params.get("litellm_embedding_model")
- return f"{api_base}/v2/vectordb/entities/search", {
+ return join_api_base(api_base, "v2/vectordb/entities/search"), {
"collectionName": vector_store_id,
"data": [query_vector],
"annsField": "book_intro_vector",
diff --git a/litellm/llms/minimax/chat/transformation.py b/litellm/llms/minimax/chat/transformation.py
index c5aa8811f02..f397bf7fbb9 100644
--- a/litellm/llms/minimax/chat/transformation.py
+++ b/litellm/llms/minimax/chat/transformation.py
@@ -5,6 +5,7 @@ MiniMax OpenAI transformation config - extends OpenAI chat config for MiniMax's
from typing import Final
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.chat.gpt_transformation import OpenAIGPTConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import AllMessageValues, ChatCompletionToolParam
@@ -55,17 +56,14 @@ class MinimaxChatConfig(OpenAIGPTConfig):
Override to ensure we use MiniMax's endpoint.
"""
# Get the base URL (either provided or default MiniMax endpoint)
- base_url: Final = self.get_api_base(api_base=api_base)
+ base_url: Final = self.get_api_base(api_base=api_base).rstrip("/")
# Ensure it ends with /chat/completions
if base_url.endswith("/chat/completions"):
return base_url
- elif base_url.endswith("/v1"):
- return f"{base_url}/chat/completions"
- elif base_url.endswith("/"):
- return f"{base_url}v1/chat/completions"
- else:
- return f"{base_url}/v1/chat/completions"
+ if base_url.endswith("/v1"):
+ return join_api_base(base_url, "chat/completions")
+ return join_api_base(base_url, "v1/chat/completions")
def remove_cache_control_flag_from_messages_and_tools(
self,
diff --git a/litellm/llms/minimax/messages/transformation.py b/litellm/llms/minimax/messages/transformation.py
index d62b88a24c6..f9f3502237c 100644
--- a/litellm/llms/minimax/messages/transformation.py
+++ b/litellm/llms/minimax/messages/transformation.py
@@ -2,9 +2,10 @@
MiniMax Anthropic transformation config - extends AnthropicConfig for MiniMax's Anthropic-compatible API
"""
-from typing import Any, Final # noqa: TID251 # override below must mirror the legacy base signature
+from typing import Any # noqa: TID251 # override below must mirror the legacy base signature
import litellm
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.anthropic.pass_through.messages.transformation import (
AnthropicMessagesConfig,
)
@@ -82,15 +83,4 @@ class MinimaxMessagesConfig(AnthropicMessagesConfig):
Get the complete URL for MiniMax API.
Override to ensure we use MiniMax's endpoint, not Anthropic's.
"""
- # Get the base URL (either provided or default MiniMax endpoint)
- base_url: Final = self.get_api_base(api_base=api_base)
-
- # If the base URL already includes the full path, return it
- if base_url.endswith("/v1/messages"):
- return base_url
-
- # Otherwise append the messages endpoint
- if base_url.endswith("/"):
- return f"{base_url}v1/messages"
- else:
- return f"{base_url}/v1/messages"
+ return ensure_api_route(self.get_api_base(api_base=api_base), "v1/messages")
diff --git a/litellm/llms/minimax/text_to_speech/transformation.py b/litellm/llms/minimax/text_to_speech/transformation.py
index 1cabc43eb1c..6f2cb36ed03 100644
--- a/litellm/llms/minimax/text_to_speech/transformation.py
+++ b/litellm/llms/minimax/text_to_speech/transformation.py
@@ -13,6 +13,7 @@ from httpx import Headers
from pydantic import ConfigDict, TypeAdapter
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.text_to_speech.transformation import (
BaseTextToSpeechConfig,
@@ -404,10 +405,5 @@ class MinimaxTextToSpeechConfig(BaseTextToSpeechConfig):
"""
Construct the MiniMax endpoint URL.
"""
- base_url = api_base or get_secret_str("MINIMAX_API_BASE") or self.TTS_BASE_URL
- base_url = base_url.rstrip("/")
-
- # MiniMax uses a simple endpoint path
- url: Final = f"{base_url}{self.TTS_ENDPOINT_PATH}"
-
- return url
+ base_url: Final = api_base or get_secret_str("MINIMAX_API_BASE") or self.TTS_BASE_URL
+ return join_api_base(base_url, self.TTS_ENDPOINT_PATH)
diff --git a/litellm/llms/mistral/audio_speech/transformation.py b/litellm/llms/mistral/audio_speech/transformation.py
index 5a8960d0898..ac0e1a33620 100644
--- a/litellm/llms/mistral/audio_speech/transformation.py
+++ b/litellm/llms/mistral/audio_speech/transformation.py
@@ -13,6 +13,7 @@ from typing import TYPE_CHECKING, Final
import httpx
from pydantic import ConfigDict, TypeAdapter, ValidationError
+from litellm.litellm_core_utils.url_utils import ensure_api_route, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.text_to_speech.transformation import (
BaseTextToSpeechConfig,
@@ -118,9 +119,7 @@ class MistralTextToSpeechConfig(BaseTextToSpeechConfig):
api_base: str | None,
litellm_params: Mapping[str, object],
) -> str:
- configured_base: Final = (api_base or self.TTS_BASE_URL).rstrip("/")
- versioned_base: Final = configured_base if configured_base.endswith("/v1") else f"{configured_base}/v1"
- return f"{versioned_base}/audio/speech"
+ return join_api_base(ensure_api_route(api_base or self.TTS_BASE_URL, "v1"), "audio/speech")
def transform_text_to_speech_request(
self,
diff --git a/litellm/llms/mistral/audio_transcription/transformation.py b/litellm/llms/mistral/audio_transcription/transformation.py
index 4fce7a95350..3eb9e122265 100644
--- a/litellm/llms/mistral/audio_transcription/transformation.py
+++ b/litellm/llms/mistral/audio_transcription/transformation.py
@@ -9,6 +9,7 @@ from typing import Final
import httpx
from litellm.litellm_core_utils.audio_utils.utils import process_audio_file
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.audio_transcription.transformation import (
AudioTranscriptionRequestData,
BaseAudioTranscriptionConfig,
@@ -57,8 +58,7 @@ class MistralAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- api_base = "https://api.mistral.ai/v1" if api_base is None else api_base.rstrip("/")
- return f"{api_base}/audio/transcriptions"
+ return join_api_base("https://api.mistral.ai/v1" if api_base is None else api_base, "audio/transcriptions")
def get_error_class(self, error_message: str, status_code: int, headers: dict | httpx.Headers) -> BaseLLMException:
return MistralAudioTranscriptionException(
diff --git a/litellm/llms/mistral/batches/transformation.py b/litellm/llms/mistral/batches/transformation.py
index 86de2836048..18f5df9b6e9 100644
--- a/litellm/llms/mistral/batches/transformation.py
+++ b/litellm/llms/mistral/batches/transformation.py
@@ -18,7 +18,7 @@ from openai.types.batch_error import BatchError
from pydantic import ConfigDict
from typing_extensions import NotRequired, ReadOnly, TypedDict
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.batches.transformation import BaseBatchesConfig
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.types.llms.base import LiteLLMBaseModel
@@ -156,7 +156,7 @@ class MistralBatchesConfig(BaseBatchesConfig):
litellm_params: Mapping[str, object],
data: CreateBatchRequest,
) -> str:
- return f"{get_mistral_api_base(api_base)}/v1/batch/jobs"
+ return join_api_base(get_mistral_api_base(api_base), "v1/batch/jobs")
def transform_create_batch_request(
self,
@@ -199,7 +199,10 @@ class MistralBatchesConfig(BaseBatchesConfig):
api_key: Final = litellm_params.get("api_key")
request: Final = MistralPresignedRequest(
method="GET",
- url=f"{get_mistral_api_base(api_base if isinstance(api_base, str) else None)}/v1/batch/jobs/{encoded_batch_id}",
+ url=join_api_base(
+ get_mistral_api_base(api_base if isinstance(api_base, str) else None),
+ f"v1/batch/jobs/{encoded_batch_id}",
+ ),
headers=get_mistral_auth_headers(_NO_HEADERS, api_key if isinstance(api_key, str) else None),
)
return dict(request)
diff --git a/litellm/llms/mistral/chat/transformation.py b/litellm/llms/mistral/chat/transformation.py
index b8a6cb1beb9..2c0acacff23 100644
--- a/litellm/llms/mistral/chat/transformation.py
+++ b/litellm/llms/mistral/chat/transformation.py
@@ -19,6 +19,7 @@ from litellm.litellm_core_utils.prompt_templates.common_utils import (
strip_litellm_internal_message_fields,
strip_none_values_from_message,
)
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.openai.chat.gpt_transformation import (
OpenAIChatCompletionStreamingHandler,
OpenAIGPTConfig,
@@ -220,8 +221,7 @@ class MistralConfig(OpenAIGPTConfig):
)
# if api_base does not end with /v1 we add it
- if api_base is not None and not api_base.endswith("/v1"): # Mistral always needs a /v1 at the end
- api_base = api_base + "/v1"
+ api_base = ensure_api_route(api_base, "v1")
dynamic_api_key: Final = (
api_key
or get_secret_str("MISTRAL_AZURE_API_KEY") # for Azure AI Mistral
diff --git a/litellm/llms/mistral/files/transformation.py b/litellm/llms/mistral/files/transformation.py
index 86bce763917..908567e88fb 100644
--- a/litellm/llms/mistral/files/transformation.py
+++ b/litellm/llms/mistral/files/transformation.py
@@ -18,7 +18,7 @@ from pydantic import ConfigDict
from typing_extensions import ReadOnly, TypedDict
from litellm.litellm_core_utils.prompt_templates.common_utils import extract_file_data
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.files.transformation import BaseFilesConfig, LiteLLMLoggingObj
from litellm.types.llms.base import LiteLLMBaseModel
@@ -130,11 +130,11 @@ class MistralFilesConfig(BaseFilesConfig):
litellm_params: Mapping[str, object],
stream: bool | None = None,
) -> str:
- return f"{get_mistral_api_base(api_base)}/v1/files"
+ return join_api_base(get_mistral_api_base(api_base), "v1/files")
def _file_url(self, file_id: str, litellm_params: Mapping[str, object], suffix: str = "") -> str:
encoded_file_id: Final = encode_url_path_segment(file_id, field_name="file_id")
- return f"{_api_base_from(litellm_params)}/v1/files/{encoded_file_id}{suffix}"
+ return join_api_base(_api_base_from(litellm_params), f"v1/files/{encoded_file_id}{suffix}")
def get_error_class(
self, error_message: str, status_code: int, headers: Mapping[str, str] | httpx.Headers
@@ -233,7 +233,7 @@ class MistralFilesConfig(BaseFilesConfig):
optional_params: Mapping[str, object],
litellm_params: Mapping[str, object],
) -> tuple[str, dict[str, str]]: # mutable-ok: BaseFilesConfig signature
- url: Final = f"{_api_base_from(litellm_params)}/v1/files"
+ url: Final = join_api_base(_api_base_from(litellm_params), "v1/files")
if not purpose:
return url, _NO_QUERY_PARAMS
return url, {"purpose": _to_mistral_purpose(purpose)}
diff --git a/litellm/llms/modelscope/chat/transformation.py b/litellm/llms/modelscope/chat/transformation.py
index c54437b9a21..ec48b344bd9 100644
--- a/litellm/llms/modelscope/chat/transformation.py
+++ b/litellm/llms/modelscope/chat/transformation.py
@@ -7,6 +7,7 @@ from typing import Final, Literal, cast, overload
from typing_extensions import override
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import AllMessageValues
@@ -86,9 +87,4 @@ class ModelScopeChatConfig(OpenAIGPTConfig):
"""
If api_base is not provided, use the default ModelScope /chat/completions endpoint.
"""
- clean_api_base: Final = (api_base or self.DEFAULT_BASE_URL).rstrip("/")
-
- if not clean_api_base.endswith("/chat/completions"):
- return f"{clean_api_base}/chat/completions"
-
- return clean_api_base
+ return ensure_api_route(api_base or self.DEFAULT_BASE_URL, "chat/completions")
diff --git a/litellm/llms/modelscope/image_generation/transformation.py b/litellm/llms/modelscope/image_generation/transformation.py
index 08cdbacdad0..36c8705f256 100644
--- a/litellm/llms/modelscope/image_generation/transformation.py
+++ b/litellm/llms/modelscope/image_generation/transformation.py
@@ -13,6 +13,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from typing_extensions import override
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
@@ -90,11 +91,10 @@ class ModelScopeImageGenerationConfig(BaseImageGenerationConfig):
"""
Get the complete URL for the ModelScope image generation API request.
"""
- base_url: str = api_base or get_secret_str("MODELSCOPE_API_BASE") or self.DEFAULT_BASE_URL
- base_url = base_url.rstrip("/")
+ base_url: Final = api_base or get_secret_str("MODELSCOPE_API_BASE") or self.DEFAULT_BASE_URL
# Return the images endpoint
- return f"{base_url}/images/generations"
+ return join_api_base(base_url, "images/generations")
@override
def validate_environment(
diff --git a/litellm/llms/mongodb/vector_stores/transformation.py b/litellm/llms/mongodb/vector_stores/transformation.py
index 0d931ee78cb..7bf9d8b8441 100644
--- a/litellm/llms/mongodb/vector_stores/transformation.py
+++ b/litellm/llms/mongodb/vector_stores/transformation.py
@@ -9,6 +9,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter, ValidationError
from litellm.exceptions import AuthenticationError, BadRequestError, ServiceUnavailableError, Timeout
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.vector_store.transformation import (
BaseQueryEmbeddingVectorStoreConfig,
@@ -283,7 +284,7 @@ class MongoDBVectorStoreConfig(BaseQueryEmbeddingVectorStoreConfig):
raise config_error("The embedding model must return a non-empty, finite query vector.")
limit: Final = cls._limit(optional_params)
return (
- f"{api_base}/v1/vector_stores/{quote(vector_store_id, safe='')}/search",
+ join_api_base(api_base, f"v1/vector_stores/{quote(vector_store_id, safe='')}/search"),
{
"query": query_text,
"query_vector": tuple(vector),
diff --git a/litellm/llms/moonshot/chat/transformation.py b/litellm/llms/moonshot/chat/transformation.py
index 93496e775db..698db8b7c25 100644
--- a/litellm/llms/moonshot/chat/transformation.py
+++ b/litellm/llms/moonshot/chat/transformation.py
@@ -9,6 +9,7 @@ import litellm
from litellm.litellm_core_utils.prompt_templates.common_utils import (
handle_messages_with_content_list_to_str_conversion,
)
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import AllMessageValues
from litellm.utils import supports_reasoning
@@ -93,12 +94,7 @@ class MoonshotChatConfig(OpenAIGPTConfig):
"""
If api_base is not provided, use the default Moonshot AI /chat/completions endpoint.
"""
- clean_api_base: Final = (api_base or "https://api.moonshot.ai/v1").rstrip("/")
-
- if not clean_api_base.endswith("/chat/completions"):
- return f"{clean_api_base}/chat/completions"
-
- return clean_api_base
+ return ensure_api_route(api_base or "https://api.moonshot.ai/v1", "chat/completions")
def get_supported_openai_params(self, model: str) -> list:
"""
diff --git a/litellm/llms/nimble/search/transformation.py b/litellm/llms/nimble/search/transformation.py
index 30cfc38eb87..b4d9be909c3 100644
--- a/litellm/llms/nimble/search/transformation.py
+++ b/litellm/llms/nimble/search/transformation.py
@@ -13,6 +13,7 @@ from typing import TYPE_CHECKING, Final
import httpx
from pydantic import ConfigDict, TypeAdapter, ValidationError
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
@@ -124,10 +125,7 @@ class NimbleSearchConfig(BaseSearchConfig):
data: dict[str, object] | list[dict[str, object]] | None = None, # mutable-ok: base signature
**kwargs: object, # kwargs-ok: BaseSearchConfig.get_complete_url signature
) -> str:
- resolved_base: Final = (api_base or get_secret_str("NIMBLE_API_BASE") or self.NIMBLE_API_BASE).rstrip("/")
- if resolved_base.endswith("/search"):
- return resolved_base
- return f"{resolved_base}/search"
+ return ensure_api_route(api_base or get_secret_str("NIMBLE_API_BASE") or self.NIMBLE_API_BASE, "search")
def transform_search_request(
self,
diff --git a/litellm/llms/nvidia_nim/rerank/ranking_transformation.py b/litellm/llms/nvidia_nim/rerank/ranking_transformation.py
index 177d7883feb..39391be0fdb 100644
--- a/litellm/llms/nvidia_nim/rerank/ranking_transformation.py
+++ b/litellm/llms/nvidia_nim/rerank/ranking_transformation.py
@@ -11,6 +11,7 @@ from typing import Final
import httpx
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.nvidia_nim.rerank.transformation import NvidiaNimRerankConfig
from litellm.types.rerank import RerankResponse
@@ -76,7 +77,7 @@ class NvidiaNimRankingConfig(NvidiaNimRerankConfig):
api_base = api_base.removesuffix("/v1")
- return f"{api_base}/v1/ranking"
+ return join_api_base(api_base, "v1/ranking")
def map_cohere_rerank_params(
self,
diff --git a/litellm/llms/nvidia_nim/rerank/transformation.py b/litellm/llms/nvidia_nim/rerank/transformation.py
index c8e1a331c8c..12377d8f854 100644
--- a/litellm/llms/nvidia_nim/rerank/transformation.py
+++ b/litellm/llms/nvidia_nim/rerank/transformation.py
@@ -8,6 +8,7 @@ from typing_extensions import Required, TypedDict
import litellm
from litellm._uuid import uuid
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.rerank.transformation import BaseRerankConfig
from litellm.secret_managers.main import get_secret_str
@@ -106,7 +107,7 @@ class NvidiaNimRerankConfig(BaseRerankConfig):
# Strip nvidia_nim/ prefix from model name if present
clean_model: Final = self._get_clean_model_name(model)
- return f"{api_base}/v1/retrieval/{clean_model}/reranking"
+ return join_api_base(api_base, f"v1/retrieval/{clean_model}/reranking")
def get_supported_cohere_rerank_params(self, model: str) -> list:
"""
diff --git a/litellm/llms/oci/chat/transformation.py b/litellm/llms/oci/chat/transformation.py
index e62e5e63642..8e1502b2610 100644
--- a/litellm/llms/oci/chat/transformation.py
+++ b/litellm/llms/oci/chat/transformation.py
@@ -21,6 +21,7 @@ from litellm._logging import verbose_logger
from litellm.constants import DEFAULT_OCI_CHAT_MAX_TOKENS
from litellm.litellm_core_utils.logging_utils import track_llm_api_timing
from litellm.litellm_core_utils.streaming_handler import CustomStreamWrapper
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.llms.custom_httpx.http_handler import (
AsyncHTTPHandler,
@@ -538,7 +539,7 @@ class OCIChatConfig(BaseConfig):
stream: bool | None = None,
) -> str:
base: Final = get_oci_base_url(optional_params, api_base or litellm.api_base)
- return f"{base}/{OCI_API_VERSION}/actions/chat"
+ return join_api_base(base, f"{OCI_API_VERSION}/actions/chat")
def _get_optional_params(self, vendor: OCIVendors, optional_params: dict, model: str = "") -> dict:
param_map: Final = (
diff --git a/litellm/llms/oci/embed/transformation.py b/litellm/llms/oci/embed/transformation.py
index edadc293771..ffcb156bb16 100644
--- a/litellm/llms/oci/embed/transformation.py
+++ b/litellm/llms/oci/embed/transformation.py
@@ -29,6 +29,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.llms.oci.common_utils import (
@@ -156,7 +157,7 @@ class OCIEmbedConfig(BaseEmbeddingConfig):
stream: bool | None = None,
) -> str:
base: Final = get_oci_base_url(optional_params, api_base or litellm.api_base)
- return f"{base}/{OCI_API_VERSION}/actions/embedText"
+ return join_api_base(base, f"{OCI_API_VERSION}/actions/embedText")
def sign_request(
self,
diff --git a/litellm/llms/ollama/chat/transformation.py b/litellm/llms/ollama/chat/transformation.py
index 644ab187f15..f9e7071db41 100644
--- a/litellm/llms/ollama/chat/transformation.py
+++ b/litellm/llms/ollama/chat/transformation.py
@@ -14,6 +14,7 @@ from litellm.litellm_core_utils.prompt_templates.common_utils import (
extract_images_from_message,
extract_reasoning_content,
)
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.base_model_iterator import BaseModelResponseIterator
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.types.llms.ollama import (
@@ -218,10 +219,7 @@ class OllamaChatConfig(BaseConfig):
Some providers need `model` in `api_base`
"""
- clean_api_base: Final = (api_base or "http://localhost:11434").rstrip("/")
- if clean_api_base.endswith("/api/chat"):
- return clean_api_base
- return f"{clean_api_base}/api/chat"
+ return ensure_api_route(api_base or "http://localhost:11434", "api/chat")
def transform_request(
self,
diff --git a/litellm/llms/ollama/common_utils.py b/litellm/llms/ollama/common_utils.py
index 284a6ffea62..7894904c870 100644
--- a/litellm/llms/ollama/common_utils.py
+++ b/litellm/llms/ollama/common_utils.py
@@ -8,6 +8,7 @@ from pydantic import BaseModel, TypeAdapter
from litellm import verbose_logger
from litellm._uuid import uuid
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.types.utils import ModelResponse, ModelResponseStream
@@ -149,7 +150,7 @@ class OllamaModelInfo(BaseLLMModelInfo):
names: Final[set[str]] = set()
try:
- resp: Final = httpx.get(f"{base}/api/tags", headers=headers)
+ resp: Final = httpx.get(join_api_base(base, "api/tags"), headers=headers)
resp.raise_for_status()
data: Final = resp.json()
# Expecting a dict with a 'models' list
@@ -230,7 +231,7 @@ class OllamaModelInfo(BaseLLMModelInfo):
try:
response: Final = module_level_client.post(
- url=f"{api_base}/api/show",
+ url=join_api_base(api_base, "api/show"),
json={"name": model},
headers=headers,
)
diff --git a/litellm/llms/ollama/completion/handler.py b/litellm/llms/ollama/completion/handler.py
index 816028499e8..d4110e55bf0 100644
--- a/litellm/llms/ollama/completion/handler.py
+++ b/litellm/llms/ollama/completion/handler.py
@@ -10,6 +10,7 @@ from typing import Any, Final, Protocol, TypedDict
from typing_extensions import NotRequired, ReadOnly
import litellm
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.types.utils import EmbeddingResponse
@@ -26,13 +27,6 @@ class OllamaEmbeddingResponse(TypedDict):
prompt_eval_count: ReadOnly[NotRequired[int]]
-def _ollama_embed_url(api_base: str) -> str:
- clean_api_base: Final = api_base.rstrip("/")
- if clean_api_base.endswith("/api/embed"):
- return clean_api_base
- return f"{clean_api_base}/api/embed"
-
-
def _prepare_ollama_embedding_payload(
model: str, prompts: list[str], optional_params: Mapping[str, object]
) -> dict[str, object]:
@@ -99,7 +93,7 @@ async def ollama_aembeddings(
):
data: Final = _prepare_ollama_embedding_payload(model, prompts, optional_params)
- response: Final = await litellm.module_level_aclient.post(url=_ollama_embed_url(api_base), json=data)
+ response: Final = await litellm.module_level_aclient.post(url=ensure_api_route(api_base, "api/embed"), json=data)
response_json: Final[OllamaEmbeddingResponse] = response.json()
return _process_ollama_embedding_response(
@@ -123,7 +117,7 @@ def ollama_embeddings(
):
data: Final = _prepare_ollama_embedding_payload(model, prompts, optional_params)
- response: Final = litellm.module_level_client.post(url=_ollama_embed_url(api_base), json=data)
+ response: Final = litellm.module_level_client.post(url=ensure_api_route(api_base, "api/embed"), json=data)
response_json: Final[OllamaEmbeddingResponse] = response.json()
return _process_ollama_embedding_response(
diff --git a/litellm/llms/ollama/completion/transformation.py b/litellm/llms/ollama/completion/transformation.py
index ac41310fb61..30de22b17ff 100644
--- a/litellm/llms/ollama/completion/transformation.py
+++ b/litellm/llms/ollama/completion/transformation.py
@@ -23,6 +23,7 @@ from litellm.litellm_core_utils.prompt_templates.image_handling import (
async_inline_remote_media,
inline_remote_image_urls,
)
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.base_model_iterator import BaseModelResponseIterator
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.types.llms.base import LiteLLMBaseModel
@@ -500,10 +501,7 @@ class OllamaConfig(BaseConfig):
Some providers need `model` in `api_base`
"""
- clean_api_base: Final = (api_base or "http://localhost:11434").rstrip("/")
- if clean_api_base.endswith("/api/generate"):
- return clean_api_base
- return f"{clean_api_base}/api/generate"
+ return ensure_api_route(api_base or "http://localhost:11434", "api/generate")
def get_model_response_iterator(
self,
diff --git a/litellm/llms/oobabooga/chat/oobabooga.py b/litellm/llms/oobabooga/chat/oobabooga.py
index e21a165d757..8879c20c5eb 100644
--- a/litellm/llms/oobabooga/chat/oobabooga.py
+++ b/litellm/llms/oobabooga/chat/oobabooga.py
@@ -3,6 +3,7 @@ from collections.abc import Callable
from typing import TYPE_CHECKING, Final
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.custom_httpx.http_handler import get_httpx_client
from litellm.utils import EmbeddingResponse, ModelResponse, Usage
@@ -49,7 +50,7 @@ def completion(
)
model = model
- completion_url = completion_url + "/v1/chat/completions"
+ completion_url = join_api_base(completion_url, "v1/chat/completions")
data: Final = oobabooga_config.transform_request(
model=model,
messages=messages,
@@ -103,7 +104,7 @@ def embedding(
if model.startswith(("http://", "https://")):
embeddings_url = model
elif api_base:
- embeddings_url = f"{api_base}/v1/embeddings"
+ embeddings_url = join_api_base(api_base, "v1/embeddings")
else:
raise OobaboogaError(
status_code=404,
diff --git a/litellm/llms/openai/chat/gpt_transformation.py b/litellm/llms/openai/chat/gpt_transformation.py
index 4d7ef833181..b4c92848752 100644
--- a/litellm/llms/openai/chat/gpt_transformation.py
+++ b/litellm/llms/openai/chat/gpt_transformation.py
@@ -34,6 +34,7 @@ from litellm.litellm_core_utils.prompt_templates.image_handling import (
async_convert_url_to_base64,
convert_url_to_base64,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.base_model_iterator import BaseModelResponseIterator
from litellm.llms.base_llm.base_utils import BaseLLMModelInfo
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
@@ -807,7 +808,7 @@ class OpenAIGPTConfig(BaseLLMModelInfo, BaseConfig):
if endpoint in api_base:
return api_base
- return f"{api_base}/{endpoint}"
+ return join_api_base(api_base, endpoint)
def validate_environment(
self,
diff --git a/litellm/llms/openai/containers/transformation.py b/litellm/llms/openai/containers/transformation.py
index 6cb2396b85d..5d03585189d 100644
--- a/litellm/llms/openai/containers/transformation.py
+++ b/litellm/llms/openai/containers/transformation.py
@@ -9,7 +9,7 @@ from litellm.litellm_core_utils.hidden_params import get_or_create_hidden_params
from litellm.litellm_core_utils.llm_cost_calc.tool_call_cost_tracking import (
StandardBuiltInToolCostTracking,
)
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.secret_managers.main import get_secret_str
from litellm.types.containers.main import (
ContainerCreateOptionalRequestParams,
@@ -130,7 +130,7 @@ class OpenAIContainerConfig(BaseContainerConfig):
or "https://api.openai.com/v1"
)
- return f"{api_base.rstrip('/')}/containers"
+ return join_api_base(api_base, "containers")
def transform_container_create_request(
self,
diff --git a/litellm/llms/openai/evals/transformation.py b/litellm/llms/openai/evals/transformation.py
index 24de7f6d479..581e5e8ad87 100644
--- a/litellm/llms/openai/evals/transformation.py
+++ b/litellm/llms/openai/evals/transformation.py
@@ -8,7 +8,7 @@ from typing import Final
import httpx
from litellm._logging import verbose_logger
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.evals.transformation import (
BaseEvalsAPIConfig,
LiteLLMLoggingObj,
@@ -67,8 +67,8 @@ class OpenAIEvalsConfig(BaseEvalsAPIConfig):
if eval_id:
encoded_eval_id: Final = encode_url_path_segment(eval_id, field_name="eval_id")
- return f"{api_base}/v1/evals/{encoded_eval_id}"
- return f"{api_base}/v1/{endpoint}"
+ return join_api_base(api_base, f"v1/evals/{encoded_eval_id}")
+ return join_api_base(api_base, f"v1/{endpoint}")
def transform_create_eval_request(
self,
@@ -226,7 +226,9 @@ class OpenAIEvalsConfig(BaseEvalsAPIConfig):
headers: dict,
) -> tuple[str, dict, dict]:
"""Transform cancel eval request for OpenAI"""
- url: Final = f"{self.get_complete_url(api_base=api_base, endpoint='evals', eval_id=eval_id)}/cancel"
+ url: Final = join_api_base(
+ self.get_complete_url(api_base=api_base, endpoint="evals", eval_id=eval_id), "cancel"
+ )
# Empty body for cancel request
request_body: Final[dict[str, object]] = {}
@@ -260,7 +262,7 @@ class OpenAIEvalsConfig(BaseEvalsAPIConfig):
api_base = litellm_params.api_base
encoded_eval_id: Final = encode_url_path_segment(eval_id, field_name="eval_id")
- url: Final = f"{api_base}/v1/evals/{encoded_eval_id}/runs"
+ url: Final = join_api_base(api_base, f"v1/evals/{encoded_eval_id}/runs")
# Build request body
request_body: Final = {k: v for k, v in create_request.items() if v is not None}
@@ -293,7 +295,7 @@ class OpenAIEvalsConfig(BaseEvalsAPIConfig):
api_base = litellm_params.api_base
encoded_eval_id: Final = encode_url_path_segment(eval_id, field_name="eval_id")
- url: Final = f"{api_base}/v1/evals/{encoded_eval_id}/runs"
+ url: Final = join_api_base(api_base, f"v1/evals/{encoded_eval_id}/runs")
# Build query parameters
query_params: Final[dict[str, object]] = {}
@@ -335,7 +337,7 @@ class OpenAIEvalsConfig(BaseEvalsAPIConfig):
"""Transform get run request for OpenAI"""
encoded_eval_id: Final = encode_url_path_segment(eval_id, field_name="eval_id")
encoded_run_id: Final = encode_url_path_segment(run_id, field_name="run_id")
- url: Final = f"{api_base}/v1/evals/{encoded_eval_id}/runs/{encoded_run_id}"
+ url: Final = join_api_base(api_base, f"v1/evals/{encoded_eval_id}/runs/{encoded_run_id}")
verbose_logger.debug("Get run request - URL: %s", url)
@@ -363,7 +365,7 @@ class OpenAIEvalsConfig(BaseEvalsAPIConfig):
"""Transform cancel run request for OpenAI"""
encoded_eval_id: Final = encode_url_path_segment(eval_id, field_name="eval_id")
encoded_run_id: Final = encode_url_path_segment(run_id, field_name="run_id")
- url: Final = f"{api_base}/v1/evals/{encoded_eval_id}/runs/{encoded_run_id}/cancel"
+ url: Final = join_api_base(api_base, f"v1/evals/{encoded_eval_id}/runs/{encoded_run_id}/cancel")
# Empty body for cancel request
request_body: Final[dict[str, object]] = {}
@@ -394,7 +396,7 @@ class OpenAIEvalsConfig(BaseEvalsAPIConfig):
"""Transform delete run request for OpenAI"""
encoded_eval_id: Final = encode_url_path_segment(eval_id, field_name="eval_id")
encoded_run_id: Final = encode_url_path_segment(run_id, field_name="run_id")
- url: Final = f"{api_base}/v1/evals/{encoded_eval_id}/runs/{encoded_run_id}"
+ url: Final = join_api_base(api_base, f"v1/evals/{encoded_eval_id}/runs/{encoded_run_id}")
# Empty body for delete request
request_body: Final[dict[str, object]] = {}
diff --git a/litellm/llms/openai/image_edit/transformation.py b/litellm/llms/openai/image_edit/transformation.py
index a9e5b1b9f34..9def833052b 100644
--- a/litellm/llms/openai/image_edit/transformation.py
+++ b/litellm/llms/openai/image_edit/transformation.py
@@ -6,6 +6,7 @@ from httpx._types import RequestFiles
import litellm
from litellm.images.utils import ImageEditRequestUtils
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_edit.transformation import BaseImageEditConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.images.main import (
@@ -184,7 +185,4 @@ class OpenAIImageEditConfig(BaseImageEditConfig):
or "https://api.openai.com/v1"
)
- # Remove trailing slashes
- api_base = api_base.rstrip("/")
-
- return f"{api_base}/images/edits"
+ return join_api_base(api_base, "images/edits")
diff --git a/litellm/llms/openai/realtime/http_transformation.py b/litellm/llms/openai/realtime/http_transformation.py
index 61dbf20397f..49d33582c6f 100644
--- a/litellm/llms/openai/realtime/http_transformation.py
+++ b/litellm/llms/openai/realtime/http_transformation.py
@@ -1,6 +1,7 @@
"""OpenAI realtime HTTP transformation config (client_secrets + realtime_calls)."""
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.realtime.http_transformation import BaseRealtimeHTTPConfig
from litellm.secret_managers.main import get_secret_str
@@ -15,17 +16,17 @@ class OpenAIRealtimeHTTPConfig(BaseRealtimeHTTPConfig):
def get_complete_url(self, api_base: str | None, model: str, api_version: str | None = None) -> str:
base = self.get_api_base(api_base).rstrip("/")
base = base.removesuffix("/v1")
- return f"{base}/v1/realtime/client_secrets"
+ return join_api_base(base, "v1/realtime/client_secrets")
def get_realtime_calls_url(self, api_base: str | None, model: str, api_version: str | None = None) -> str:
base = self.get_api_base(api_base).rstrip("/")
base = base.removesuffix("/v1")
- return f"{base}/v1/realtime/calls"
+ return join_api_base(base, "v1/realtime/calls")
def get_transcription_session_url(self, api_base: str | None, model: str, api_version: str | None = None) -> str:
base = self.get_api_base(api_base).rstrip("/")
base = base.removesuffix("/v1")
- return f"{base}/v1/realtime/transcription_sessions"
+ return join_api_base(base, "v1/realtime/transcription_sessions")
def validate_environment(
self,
diff --git a/litellm/llms/openai/responses/count_tokens/transformation.py b/litellm/llms/openai/responses/count_tokens/transformation.py
index 6d882e0eeb1..c1ee42444c5 100644
--- a/litellm/llms/openai/responses/count_tokens/transformation.py
+++ b/litellm/llms/openai/responses/count_tokens/transformation.py
@@ -13,6 +13,7 @@ from pydantic import JsonValue, TypeAdapter, ValidationError
from typing_extensions import ReadOnly, TypedDict
from litellm.litellm_core_utils.tool_fields import DropPolicy
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.anthropic.pass_through.convertible_blocks import malformed_tool_result_text_error
from litellm.llms.anthropic.pass_through.messages.utils import normalize_native_messages
from litellm.llms.anthropic.pass_through.responses_adapters.transformation import (
@@ -148,9 +149,8 @@ class OpenAICountTokensConfig:
"""
def get_openai_count_tokens_endpoint(self, api_base: str | None = None) -> str:
- base = api_base or "https://api.openai.com/v1"
- base = base.rstrip("/")
- return f"{base}/responses/input_tokens"
+ base: Final = api_base or "https://api.openai.com/v1"
+ return join_api_base(base, "responses/input_tokens")
def transform_request_to_count_tokens(
self,
diff --git a/litellm/llms/openai/responses/transformation.py b/litellm/llms/openai/responses/transformation.py
index 8f29a71d63b..b90f6aa01da 100644
--- a/litellm/llms/openai/responses/transformation.py
+++ b/litellm/llms/openai/responses/transformation.py
@@ -23,7 +23,7 @@ from litellm.litellm_core_utils.prompt_templates.common_utils import (
with_prompt_cache_breakpoint,
)
from litellm.litellm_core_utils.safe_json_loads import safe_json_loads
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.responses.transformation import BaseResponsesAPIConfig
from litellm.llms.openai.chat.gpt_5_transformation import (
is_gpt_reasoning_series_name,
@@ -881,10 +881,7 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig):
or "https://api.openai.com/v1"
)
- # Remove trailing slashes
- api_base = api_base.rstrip("/")
-
- return f"{api_base}/responses"
+ return join_api_base(api_base, "responses")
def transform_streaming_response(
self,
@@ -1111,7 +1108,7 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig):
order: Literal["asc", "desc"] = "desc",
) -> tuple[str, dict]:
encoded_response_id: Final = encode_url_path_segment(response_id, field_name="response_id")
- url: Final = f"{api_base}/{encoded_response_id}/input_items"
+ url: Final = join_api_base(api_base, f"{encoded_response_id}/input_items")
params: Final[dict[str, object]] = {}
if after is not None:
params["after"] = after
@@ -1152,7 +1149,7 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig):
- POST /v1/responses/{response_id}/cancel
"""
encoded_response_id: Final = encode_url_path_segment(response_id, field_name="response_id")
- url: Final = f"{api_base}/{encoded_response_id}/cancel"
+ url: Final = join_api_base(api_base, f"{encoded_response_id}/cancel")
data: Final[dict] = {}
return url, data
diff --git a/litellm/llms/openai/transcriptions/whisper_transformation.py b/litellm/llms/openai/transcriptions/whisper_transformation.py
index 8f96f688189..eb4f1b6c9ce 100644
--- a/litellm/llms/openai/transcriptions/whisper_transformation.py
+++ b/litellm/llms/openai/transcriptions/whisper_transformation.py
@@ -4,6 +4,7 @@ from typing import Final
from httpx import Headers, Response
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.audio_transcription.transformation import (
AudioTranscriptionRequestData,
BaseAudioTranscriptionConfig,
@@ -46,9 +47,9 @@ class OpenAIWhisperAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
# if endswith "/v1"
if api_base and api_base.endswith("/v1"):
- api_base = f"{api_base}/audio/transcriptions"
+ api_base = join_api_base(api_base, "audio/transcriptions")
else:
- api_base = f"{api_base}/v1/audio/transcriptions"
+ api_base = join_api_base(api_base, "v1/audio/transcriptions")
return api_base or ""
diff --git a/litellm/llms/openai/vector_store_files/transformation.py b/litellm/llms/openai/vector_store_files/transformation.py
index 7f65ade224e..bd8145bb460 100644
--- a/litellm/llms/openai/vector_store_files/transformation.py
+++ b/litellm/llms/openai/vector_store_files/transformation.py
@@ -4,7 +4,7 @@ from typing import Final, cast
import httpx
import litellm
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.vector_store_files.transformation import (
BaseVectorStoreFilesConfig,
)
@@ -92,16 +92,15 @@ class OpenAIVectorStoreFilesConfig(BaseVectorStoreFilesConfig):
vector_store_id: str,
litellm_params: Mapping[str, object],
) -> str:
- base_url = (
+ base_url: Final = (
api_base
or litellm.api_base
or get_secret_str("OPENAI_BASE_URL")
or get_secret_str("OPENAI_API_BASE")
or "https://api.openai.com/v1"
)
- base_url = base_url.rstrip("/")
encoded_vector_store_id: Final = encode_url_path_segment(vector_store_id, field_name="vector_store_id")
- return f"{base_url}/vector_stores/{encoded_vector_store_id}/files"
+ return join_api_base(base_url, f"vector_stores/{encoded_vector_store_id}/files")
def transform_create_vector_store_file_request(
self,
@@ -187,7 +186,7 @@ class OpenAIVectorStoreFilesConfig(BaseVectorStoreFilesConfig):
api_base: str,
) -> tuple[str, dict[str, object]]:
encoded_file_id: Final = encode_url_path_segment(file_id, field_name="file_id")
- return f"{api_base}/{encoded_file_id}/content", {}
+ return join_api_base(api_base, f"{encoded_file_id}/content"), {}
def transform_retrieve_vector_store_file_content_response(
self,
diff --git a/litellm/llms/openai/vector_stores/transformation.py b/litellm/llms/openai/vector_stores/transformation.py
index 46105a66be3..d108802f7f9 100644
--- a/litellm/llms/openai/vector_stores/transformation.py
+++ b/litellm/llms/openai/vector_stores/transformation.py
@@ -3,7 +3,7 @@ from typing import TYPE_CHECKING, Any, Final, cast
import httpx
import litellm
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.vector_store.transformation import BaseVectorStoreConfig
from litellm.llms.openai.workload_identity import resolve_openai_bearer_token
from litellm.secret_managers.main import get_secret_str
@@ -88,10 +88,7 @@ class OpenAIVectorStoreConfig(BaseVectorStoreConfig):
or "https://api.openai.com/v1"
)
- # Remove trailing slashes
- api_base = api_base.rstrip("/")
-
- return f"{api_base}/vector_stores"
+ return join_api_base(api_base, "vector_stores")
def transform_search_vector_store_request(
self,
@@ -105,7 +102,7 @@ class OpenAIVectorStoreConfig(BaseVectorStoreConfig):
) -> tuple[str, dict]:
encoded_vector_store_id: Final = encode_url_path_segment(vector_store_id, field_name="vector_store_id")
base_url, query_separator, query_string = api_base.partition("?")
- url: Final = f"{base_url}/{encoded_vector_store_id}/search{query_separator}{query_string}"
+ url: Final = f"{join_api_base(base_url, f'{encoded_vector_store_id}/search')}{query_separator}{query_string}"
typed_request_body: Final = VectorStoreSearchRequest(
query=query,
filters=vector_store_search_optional_params.get("filters", None),
diff --git a/litellm/llms/openai/videos/transformation.py b/litellm/llms/openai/videos/transformation.py
index d422e51f649..0560e1c2e3a 100644
--- a/litellm/llms/openai/videos/transformation.py
+++ b/litellm/llms/openai/videos/transformation.py
@@ -8,7 +8,7 @@ from urllib.parse import quote
import httpx
from httpx._types import FileContent, FileTypes, RequestFiles
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.videos.transformation import BaseVideoConfig
from litellm.llms.openai.image_edit.transformation import ImageEditRequestUtils
from litellm.llms.openai.workload_identity import resolve_openai_bearer_token
@@ -98,7 +98,7 @@ class OpenAIVideoConfig(BaseVideoConfig):
if api_base is None:
api_base = "https://api.openai.com/v1"
- return f"{api_base.rstrip('/')}/videos"
+ return join_api_base(api_base, "videos")
def use_multipart_form_data(self) -> bool:
return True
@@ -214,7 +214,7 @@ class OpenAIVideoConfig(BaseVideoConfig):
encoded_video_id: Final = encode_url_path_segment(original_video_id, field_name="video_id")
# Construct the URL for video content download
- url = f"{api_base.rstrip('/')}/{encoded_video_id}/content"
+ url = join_api_base(api_base, f"{encoded_video_id}/content")
if variant is not None:
# Encode the user-controlled ``variant`` so a value like
# ``thumbnail&extra=1`` cannot inject additional query params
@@ -246,7 +246,7 @@ class OpenAIVideoConfig(BaseVideoConfig):
encoded_video_id: Final = encode_url_path_segment(original_video_id, field_name="video_id")
# Construct the URL for video remix
- url: Final = f"{api_base.rstrip('/')}/{encoded_video_id}/remix"
+ url: Final = join_api_base(api_base, f"{encoded_video_id}/remix")
# Prepare the request data
data: Final[dict[str, object]] = {"prompt": prompt}
@@ -465,7 +465,7 @@ class OpenAIVideoConfig(BaseVideoConfig):
litellm_params: GenericLiteLLMParams,
headers: dict,
) -> tuple[str, list]:
- url: Final = f"{api_base.rstrip('/')}/characters"
+ url: Final = join_api_base(api_base, "characters")
files_list: Final[list[tuple[str, FileTypes]]] = [("name", (None, name))]
self._add_video_to_files(files_list, video, "video")
return url, files_list
@@ -486,7 +486,7 @@ class OpenAIVideoConfig(BaseVideoConfig):
) -> tuple[str, dict]:
original_character_id: Final = extract_original_character_id(character_id)
encoded_character_id: Final = encode_url_path_segment(original_character_id, field_name="character_id")
- url: Final = f"{api_base.rstrip('/')}/characters/{encoded_character_id}"
+ url: Final = join_api_base(api_base, f"characters/{encoded_character_id}")
return url, {}
def transform_video_get_character_response(
@@ -507,7 +507,7 @@ class OpenAIVideoConfig(BaseVideoConfig):
extra_body: dict[str, object] | None = None,
prefetched_source_data: dict[str, object] | None = None,
) -> tuple[str, Mapping[str, object], RequestFiles | None]:
- url: Final = f"{api_base.rstrip('/')}/edits"
+ url: Final = join_api_base(api_base, "edits")
if video_file is not None:
files: Final[RequestFiles] = (self._video_file_tuple(video_file, "video"),)
@@ -547,7 +547,7 @@ class OpenAIVideoConfig(BaseVideoConfig):
extra_body: dict[str, object] | None = None,
) -> tuple[str, dict]:
original_video_id: Final = extract_original_video_id(video_id)
- url: Final = f"{api_base.rstrip('/')}/extensions"
+ url: Final = join_api_base(api_base, "extensions")
data: Final[dict[str, object]] = {
"prompt": prompt,
"seconds": seconds,
diff --git a/litellm/llms/openai_like/common_utils.py b/litellm/llms/openai_like/common_utils.py
index 11b85e52af5..5d3085a3901 100644
--- a/litellm/llms/openai_like/common_utils.py
+++ b/litellm/llms/openai_like/common_utils.py
@@ -2,6 +2,8 @@ from typing import Literal
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
+
class OpenAILikeError(Exception):
def __init__(self, status_code, message):
@@ -48,7 +50,7 @@ class OpenAILikeBase:
if not custom_endpoint:
if endpoint_type == "chat_completions":
- api_base = f"{api_base}/chat/completions"
+ api_base = join_api_base(api_base, "chat/completions")
elif endpoint_type == "embeddings":
- api_base = f"{api_base}/embeddings"
+ api_base = join_api_base(api_base, "embeddings")
return api_base, headers
diff --git a/litellm/llms/openai_like/dynamic_config.py b/litellm/llms/openai_like/dynamic_config.py
index f0bf710b951..685561f55a5 100644
--- a/litellm/llms/openai_like/dynamic_config.py
+++ b/litellm/llms/openai_like/dynamic_config.py
@@ -9,6 +9,7 @@ from litellm._logging import verbose_logger
from litellm.litellm_core_utils.prompt_templates.common_utils import (
handle_messages_with_content_list_to_str_conversion,
)
+from litellm.litellm_core_utils.url_utils import ensure_api_route, join_api_base
from litellm.llms.openai.chat.gpt_transformation import OpenAIGPTConfig
from litellm.llms.openai_like.chat.transformation import OpenAILikeChatConfig
from litellm.secret_managers.main import get_secret_str
@@ -91,13 +92,9 @@ def create_config_class(provider: SimpleProviderConfig):
if not api_base:
api_base = provider.base_url
- if api_base is None:
- raise ValueError(f"api_base is required for provider {provider.slug}")
-
- if not api_base.endswith("/chat/completions"):
- api_base = f"{api_base}/chat/completions"
-
- return api_base
+ if isinstance(api_base, str):
+ return ensure_api_route(api_base, "chat/completions")
+ raise ValueError(f"api_base is required for provider {provider.slug}")
def get_supported_openai_params(self, model: str) -> list:
"""Get supported OpenAI params, excluding tool-related params for models
@@ -226,11 +223,9 @@ def create_responses_config_class(provider: SimpleProviderConfig) -> "type[OpenA
if not api_base:
api_base = provider.base_url
- if api_base is None:
- raise ValueError(f"api_base is required for provider {provider.slug}")
-
- api_base = api_base.rstrip("/")
- return f"{api_base}/responses"
+ if isinstance(api_base, str):
+ return join_api_base(api_base, "responses")
+ raise ValueError(f"api_base is required for provider {provider.slug}")
def transform_responses_api_request(
self,
diff --git a/litellm/llms/openai_like/messages/transformation.py b/litellm/llms/openai_like/messages/transformation.py
index ef501720e2d..12c8b1da166 100644
--- a/litellm/llms/openai_like/messages/transformation.py
+++ b/litellm/llms/openai_like/messages/transformation.py
@@ -1,6 +1,7 @@
from typing import Any, Final
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.anthropic.common_utils import normalize_cache_control_in_anthropic_payload
from litellm.llms.anthropic.pass_through.messages.transformation import (
AnthropicMessagesConfig,
@@ -113,7 +114,7 @@ class OpenAILikeAnthropicMessagesConfig(AnthropicMessagesConfig):
if base.endswith("/v1/messages"):
return base
base = base.removesuffix("/v1")
- return f"{base}/v1/messages"
+ return join_api_base(base, "v1/messages")
class JSONProviderAnthropicMessagesConfig(OpenAILikeAnthropicMessagesConfig):
diff --git a/litellm/llms/openai_like/responses/transformation.py b/litellm/llms/openai_like/responses/transformation.py
index 9655772e953..68de4065b09 100644
--- a/litellm/llms/openai_like/responses/transformation.py
+++ b/litellm/llms/openai_like/responses/transformation.py
@@ -8,6 +8,7 @@ provider-specific resolution (slug, API key env var, base URL).
from typing import Final
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.router import GenericLiteLLMParams
@@ -47,5 +48,4 @@ class OpenAILikeResponsesConfig(OpenAIResponsesAPIConfig):
api_base = api_base or get_secret_str("OPENAI_LIKE_API_BASE")
if not api_base:
raise ValueError("api_base is required for openai_like provider")
- api_base = api_base.rstrip("/")
- return f"{api_base}/responses"
+ return join_api_base(api_base, "responses")
diff --git a/litellm/llms/openrouter/embedding/transformation.py b/litellm/llms/openrouter/embedding/transformation.py
index 14b0e462ea7..d1530a29515 100644
--- a/litellm/llms/openrouter/embedding/transformation.py
+++ b/litellm/llms/openrouter/embedding/transformation.py
@@ -11,6 +11,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.types.llms.openai import AllEmbeddingInputValues
from litellm.types.utils import EmbeddingResponse
@@ -84,15 +85,7 @@ class OpenrouterEmbeddingConfig(BaseEmbeddingConfig):
"""
Get the complete URL for OpenRouter Embedding API endpoint.
"""
- # api_base is already set to https://openrouter.ai/api/v1 in main.py
- # Remove trailing slashes
- if api_base:
- api_base = api_base.rstrip("/")
- else:
- api_base = "https://openrouter.ai/api/v1"
-
- # Return the embeddings endpoint
- return f"{api_base}/embeddings"
+ return join_api_base(api_base or "https://openrouter.ai/api/v1", "embeddings")
def transform_embedding_request(
self,
diff --git a/litellm/llms/openrouter/image_edit/transformation.py b/litellm/llms/openrouter/image_edit/transformation.py
index a73a2fe4192..9644ee3f280 100644
--- a/litellm/llms/openrouter/image_edit/transformation.py
+++ b/litellm/llms/openrouter/image_edit/transformation.py
@@ -52,6 +52,7 @@ from typing_extensions import NotRequired, ReadOnly, TypedDict
import litellm
from litellm.images.utils import ImageEditRequestUtils
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.image_edit.transformation import BaseImageEditConfig
from litellm.llms.openrouter.common_utils import OpenRouterException
@@ -187,11 +188,8 @@ class OpenRouterImageEditConfig(BaseImageEditConfig):
api_base: str | None,
litellm_params: dict,
) -> str:
- base_url = api_base or get_secret_str("OPENROUTER_API_BASE") or "https://openrouter.ai/api/v1"
- base_url = base_url.rstrip("/")
- if not base_url.endswith("/chat/completions"):
- return f"{base_url}/chat/completions"
- return base_url
+ base_url: Final = api_base or get_secret_str("OPENROUTER_API_BASE") or "https://openrouter.ai/api/v1"
+ return ensure_api_route(base_url, "chat/completions")
def transform_image_edit_request(
self,
diff --git a/litellm/llms/openrouter/image_generation/transformation.py b/litellm/llms/openrouter/image_generation/transformation.py
index 355df81b87a..ce6e62aaa88 100644
--- a/litellm/llms/openrouter/image_generation/transformation.py
+++ b/litellm/llms/openrouter/image_generation/transformation.py
@@ -35,6 +35,7 @@ from pydantic import ConfigDict, TypeAdapter
import litellm
from litellm.litellm_core_utils.hidden_params import HIDDEN_PARAMS_ATTR, set_hidden_params
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
@@ -269,10 +270,7 @@ class OpenRouterImageGenerationConfig(BaseImageGenerationConfig):
Default: https://openrouter.ai/api/v1/chat/completions
"""
if api_base:
- if not api_base.endswith("/chat/completions"):
- api_base = api_base.rstrip("/")
- return f"{api_base}/chat/completions"
- return api_base
+ return ensure_api_route(api_base, "chat/completions")
return "https://openrouter.ai/api/v1/chat/completions"
diff --git a/litellm/llms/openrouter/responses/transformation.py b/litellm/llms/openrouter/responses/transformation.py
index aeb5b09e14a..e166ccaad2d 100644
--- a/litellm/llms/openrouter/responses/transformation.py
+++ b/litellm/llms/openrouter/responses/transformation.py
@@ -18,6 +18,7 @@ import litellm
from litellm.litellm_core_utils.core_helpers import RESPONSE_COST_HEADER
from litellm.litellm_core_utils.hidden_params import set_hidden_params
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import (
@@ -138,9 +139,7 @@ class OpenRouterResponsesAPIConfig(OpenAIResponsesAPIConfig):
api_base or litellm.api_base or get_secret_str("OPENROUTER_API_BASE") or "https://openrouter.ai/api/v1"
)
- api_base = api_base.rstrip("/")
-
- return f"{api_base}/responses"
+ return join_api_base(api_base, "responses")
def supports_native_websocket(self) -> bool:
"""OpenRouter does not support native WebSocket for Responses API"""
diff --git a/litellm/llms/opensandbox/sandbox/transformation.py b/litellm/llms/opensandbox/sandbox/transformation.py
index 13c9d805638..574fbfa9e5b 100644
--- a/litellm/llms/opensandbox/sandbox/transformation.py
+++ b/litellm/llms/opensandbox/sandbox/transformation.py
@@ -21,6 +21,7 @@ from litellm.constants import (
OPEN_SANDBOX_READY_TIMEOUT,
)
from litellm.litellm_core_utils.hidden_params import HIDDEN_PARAMS_ATTR, set_hidden_params
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.sandbox.transformation import (
SANDBOX_MAX_OUTPUT_BYTES,
BaseSandboxConfig,
@@ -93,7 +94,7 @@ class OpenSandboxSandboxConfig(BaseSandboxConfig):
)
response: Final = await self._http(client).post(
- url=f"{base}/sandboxes",
+ url=join_api_base(base, "sandboxes"),
headers=self._lifecycle_headers(key),
json=body,
)
@@ -192,7 +193,7 @@ class OpenSandboxSandboxConfig(BaseSandboxConfig):
key: Final = self._api_key(api_key=api_key, handle=handle)
try:
response: Final = await self._http(client).delete(
- url=f"{base}/sandboxes/{handle.id}",
+ url=join_api_base(base, f"sandboxes/{handle.id}"),
headers=self._lifecycle_headers(key),
)
except httpx.HTTPStatusError as e:
@@ -258,7 +259,7 @@ class OpenSandboxSandboxConfig(BaseSandboxConfig):
deadline: Final = time.monotonic() + ready_timeout
while True:
response = await self._http(client).get(
- url=f"{api_base}/sandboxes/{sandbox_id}",
+ url=join_api_base(api_base, f"sandboxes/{sandbox_id}"),
headers=headers,
)
data = _JSON_VALUE.validate_python(response.json())
@@ -316,7 +317,7 @@ class OpenSandboxSandboxConfig(BaseSandboxConfig):
client: AsyncHTTPHandler | None,
) -> tuple[str, dict[str, str]]:
response: Final = await self._http(client).get(
- url=f"{api_base}/sandboxes/{sandbox_id}/endpoints/{OPEN_SANDBOX_EXECD_PORT}",
+ url=join_api_base(api_base, f"sandboxes/{sandbox_id}/endpoints/{OPEN_SANDBOX_EXECD_PORT}"),
headers=headers,
params={"use_server_proxy": use_server_proxy},
)
diff --git a/litellm/llms/ovhcloud/audio_transcription/transformation.py b/litellm/llms/ovhcloud/audio_transcription/transformation.py
index 3ae5f6fa69c..dd61bfb0c13 100644
--- a/litellm/llms/ovhcloud/audio_transcription/transformation.py
+++ b/litellm/llms/ovhcloud/audio_transcription/transformation.py
@@ -12,6 +12,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.audio_utils.utils import process_audio_file
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.audio_transcription.transformation import (
AudioTranscriptionRequestData,
BaseAudioTranscriptionConfig,
@@ -63,9 +64,8 @@ class OVHCloudAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- api_base = "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1" if api_base is None else api_base.rstrip("/")
- complete_url: Final = f"{api_base}/audio/transcriptions"
- return complete_url
+ base_url: Final = "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1" if api_base is None else api_base
+ return join_api_base(base_url, "audio/transcriptions")
def get_error_class(self, error_message: str, status_code: int, headers: dict | httpx.Headers) -> BaseLLMException:
return OVHCloudException(
diff --git a/litellm/llms/ovhcloud/chat/transformation.py b/litellm/llms/ovhcloud/chat/transformation.py
index 3d3953eed34..91b3dbe6268 100644
--- a/litellm/llms/ovhcloud/chat/transformation.py
+++ b/litellm/llms/ovhcloud/chat/transformation.py
@@ -9,6 +9,7 @@ from typing import Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.base_model_iterator import BaseModelResponseIterator
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.openai.chat.gpt_transformation import OpenAIGPTConfig
@@ -31,9 +32,8 @@ class OVHCloudChatConfig(OpenAIGPTConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- api_base = "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1" if api_base is None else api_base.rstrip("/")
- complete_url: Final = f"{api_base}/chat/completions"
- return complete_url
+ base_url: Final = "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1" if api_base is None else api_base
+ return join_api_base(base_url, "chat/completions")
def get_error_class(self, error_message: str, status_code: int, headers: dict | httpx.Headers) -> BaseLLMException:
return OVHCloudException(
diff --git a/litellm/llms/ovhcloud/embedding/transformation.py b/litellm/llms/ovhcloud/embedding/transformation.py
index cdf095e6ad1..487c40afa52 100644
--- a/litellm/llms/ovhcloud/embedding/transformation.py
+++ b/litellm/llms/ovhcloud/embedding/transformation.py
@@ -10,6 +10,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.secret_managers.main import get_secret_str
@@ -35,9 +36,8 @@ class OVHCloudEmbeddingConfig(BaseEmbeddingConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- api_base = "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1" if api_base is None else api_base.rstrip("/")
- complete_url: Final = f"{api_base}/embeddings"
- return complete_url
+ base_url: Final = "https://oai.endpoints.kepler.ai.cloud.ovh.net/v1" if api_base is None else api_base
+ return join_api_base(base_url, "embeddings")
def validate_environment(
self,
diff --git a/litellm/llms/perplexity/embedding/transformation.py b/litellm/llms/perplexity/embedding/transformation.py
index 945a58d4d75..7b4da933976 100644
--- a/litellm/llms/perplexity/embedding/transformation.py
+++ b/litellm/llms/perplexity/embedding/transformation.py
@@ -20,6 +20,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.secret_managers.main import get_secret_str
@@ -68,7 +69,7 @@ class PerplexityEmbeddingConfig(BaseEmbeddingConfig):
if api_base:
clean_api_base: Final = api_base.rstrip("/")
if not clean_api_base.endswith("/embeddings"):
- return f"{clean_api_base}/v1/embeddings"
+ return join_api_base(clean_api_base, "v1/embeddings")
return clean_api_base
return "https://api.perplexity.ai/v1/embeddings"
diff --git a/litellm/llms/perplexity/responses/transformation.py b/litellm/llms/perplexity/responses/transformation.py
index 91d40d8247f..a2b7eceea40 100644
--- a/litellm/llms/perplexity/responses/transformation.py
+++ b/litellm/llms/perplexity/responses/transformation.py
@@ -14,6 +14,7 @@ from typing import Any, Final
import httpx
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.secret_managers.main import get_secret_str
@@ -51,7 +52,7 @@ class PerplexityResponsesConfig(OpenAIResponsesAPIConfig):
def get_complete_url(self, api_base: str | None, litellm_params: dict) -> str:
api_base = api_base or get_secret_str("PERPLEXITY_API_BASE") or "https://api.perplexity.ai"
- return f"{api_base.rstrip('/')}/v1/responses"
+ return join_api_base(api_base, "v1/responses")
def _ensure_message_type(self, input: str | ResponseInputParam) -> str | ResponseInputParam:
"""Ensure list input items have type='message' (required by Perplexity)."""
diff --git a/litellm/llms/perplexity/search/transformation.py b/litellm/llms/perplexity/search/transformation.py
index b46e4225309..1f3b0fe729d 100644
--- a/litellm/llms/perplexity/search/transformation.py
+++ b/litellm/llms/perplexity/search/transformation.py
@@ -9,6 +9,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
SearchResponse,
@@ -78,11 +79,7 @@ class PerplexitySearchConfig(BaseSearchConfig):
"""
Get complete URL for Search endpoint.
"""
- resolved_api_base: Final = api_base or get_secret_str("PERPLEXITY_API_BASE") or self.PERPLEXITY_API_BASE
- clean_api_base: Final = resolved_api_base.rstrip("/")
- if clean_api_base.endswith("/search"):
- return clean_api_base
- return f"{clean_api_base}/search"
+ return ensure_api_route(api_base or get_secret_str("PERPLEXITY_API_BASE") or self.PERPLEXITY_API_BASE, "search")
def transform_search_request(
self,
diff --git a/litellm/llms/pg_vector/vector_stores/transformation.py b/litellm/llms/pg_vector/vector_stores/transformation.py
index e4b06c36bf4..eeecd02653d 100644
--- a/litellm/llms/pg_vector/vector_stores/transformation.py
+++ b/litellm/llms/pg_vector/vector_stores/transformation.py
@@ -1,6 +1,6 @@
from typing import TYPE_CHECKING, Any, Final
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.openai.vector_stores.transformation import OpenAIVectorStoreConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.router import GenericLiteLLMParams
@@ -66,10 +66,7 @@ class PGVectorStoreConfig(OpenAIVectorStoreConfig):
"PG Vector API base URL is required. Set PG_VECTOR_API_BASE environment variable or pass api_base in litellm_params."
)
- # Remove trailing slashes
- api_base = api_base.rstrip("/")
-
- return f"{api_base}/v1/vector_stores"
+ return join_api_base(api_base, "v1/vector_stores")
def transform_search_vector_store_request(
self,
@@ -82,7 +79,7 @@ class PGVectorStoreConfig(OpenAIVectorStoreConfig):
extra_body: dict[str, Any] | None = None,
) -> tuple[str, dict]:
encoded_vector_store_id: Final = encode_url_path_segment(vector_store_id, field_name="vector_store_id")
- url: Final = f"{api_base}/{encoded_vector_store_id}/search"
+ url: Final = join_api_base(api_base, f"{encoded_vector_store_id}/search")
_, request_body = super().transform_search_vector_store_request(
vector_store_id=vector_store_id,
query=query,
diff --git a/litellm/llms/predibase/chat/transformation.py b/litellm/llms/predibase/chat/transformation.py
index 0dd7e27722e..b30713fa119 100644
--- a/litellm/llms/predibase/chat/transformation.py
+++ b/litellm/llms/predibase/chat/transformation.py
@@ -13,6 +13,7 @@ from litellm.litellm_core_utils.prompt_templates.factory import (
prompt_factory,
render_custom_prompt_template,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.types.llms.openai import AllMessageValues
from litellm.types.utils import Choices, Message, ModelResponse, Usage
@@ -337,12 +338,12 @@ class PredibaseConfig(BaseConfig):
elif "PREDIBASE_API_BASE" in os.environ:
base_url = os.getenv("PREDIBASE_API_BASE", "")
- completion_url = f"{base_url}/{tenant_id}/deployments/v2/llms/{model}"
+ completion_url = join_api_base(base_url, f"{tenant_id}/deployments/v2/llms/{model}")
should_stream: Final = stream if stream is not None else optional_params.get("stream", False)
if should_stream is True:
- completion_url += "/generate_stream"
+ completion_url = join_api_base(completion_url, "generate_stream")
else:
- completion_url += "/generate"
+ completion_url = join_api_base(completion_url, "generate")
return completion_url
def get_error_class(self, error_message: str, status_code: int, headers: dict | Headers) -> BaseLLMException:
diff --git a/litellm/llms/ragflow/chat/transformation.py b/litellm/llms/ragflow/chat/transformation.py
index 97b723f7dd5..f626cd8e55f 100644
--- a/litellm/llms/ragflow/chat/transformation.py
+++ b/litellm/llms/ragflow/chat/transformation.py
@@ -13,7 +13,7 @@ Model name format:
from typing import Final
import litellm
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.openai.openai import OpenAIConfig
from litellm.secret_managers.main import get_secret, get_secret_str
from litellm.types.llms.openai import AllMessageValues
@@ -122,7 +122,7 @@ class RAGFlowConfig(OpenAIConfig):
if not path.startswith("/"):
path = "/" + path
- return f"{api_base}{path}"
+ return join_api_base(api_base, path)
def _get_openai_compatible_provider_info(
self,
diff --git a/litellm/llms/ragflow/vector_stores/transformation.py b/litellm/llms/ragflow/vector_stores/transformation.py
index 007b047cc1b..ef44b6e4c0f 100644
--- a/litellm/llms/ragflow/vector_stores/transformation.py
+++ b/litellm/llms/ragflow/vector_stores/transformation.py
@@ -5,6 +5,7 @@ import httpx
from pydantic import ConfigDict, RootModel, with_config
from typing_extensions import NotRequired, ReadOnly, TypedDict
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.vector_store.transformation import BaseVectorStoreConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.router import GenericLiteLLMParams
@@ -95,15 +96,16 @@ class RAGFlowVectorStoreConfig(BaseVectorStoreConfig):
- api_base in litellm_params
- Default: http://localhost:9380
"""
- api_base = (
- api_base or litellm_params.get("api_base") or get_secret_str("RAGFLOW_API_BASE") or "http://localhost:9380"
+ params_api_base: Final = litellm_params.get("api_base")
+ resolved_api_base: Final = (
+ api_base
+ or (f"{params_api_base}" if params_api_base else None)
+ or get_secret_str("RAGFLOW_API_BASE")
+ or "http://localhost:9380"
)
- # Remove trailing slashes
- api_base = api_base.rstrip("/")
-
# RAGFlow datasets API endpoint
- return f"{api_base}/api/v1/datasets"
+ return join_api_base(resolved_api_base, "api/v1/datasets")
def transform_search_vector_store_request(
self,
diff --git a/litellm/llms/recraft/image_edit/transformation.py b/litellm/llms/recraft/image_edit/transformation.py
index 46905580382..74bfa52c5e0 100644
--- a/litellm/llms/recraft/image_edit/transformation.py
+++ b/litellm/llms/recraft/image_edit/transformation.py
@@ -5,6 +5,7 @@ import httpx
from httpx._types import RequestFiles
from litellm.images.utils import ImageEditRequestUtils
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_edit.transformation import BaseImageEditConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.images.main import ImageEditOptionalRequestParams
@@ -68,11 +69,8 @@ class RecraftImageEditConfig(BaseImageEditConfig):
Some providers need `model` in `api_base`
"""
- complete_url: str = api_base or get_secret_str("RECRAFT_API_BASE") or self.DEFAULT_BASE_URL
-
- complete_url = complete_url.rstrip("/")
- complete_url = f"{complete_url}/{self.IMAGE_EDIT_ENDPOINT}"
- return complete_url
+ base_url: Final = api_base or get_secret_str("RECRAFT_API_BASE") or self.DEFAULT_BASE_URL
+ return join_api_base(base_url, self.IMAGE_EDIT_ENDPOINT)
def validate_environment(
self,
diff --git a/litellm/llms/recraft/image_generation/transformation.py b/litellm/llms/recraft/image_generation/transformation.py
index 0fcfafb79fb..7b1a7714ec6 100644
--- a/litellm/llms/recraft/image_generation/transformation.py
+++ b/litellm/llms/recraft/image_generation/transformation.py
@@ -4,6 +4,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
from pydantic import ConfigDict, TypeAdapter
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
)
@@ -72,11 +73,8 @@ class RecraftImageGenerationConfig(BaseImageGenerationConfig):
Some providers need `model` in `api_base`
"""
- complete_url: str = api_base or get_secret_str("RECRAFT_API_BASE") or self.DEFAULT_BASE_URL
-
- complete_url = complete_url.rstrip("/")
- complete_url = f"{complete_url}/{self.IMAGE_GENERATION_ENDPOINT}"
- return complete_url
+ base_url: Final = api_base or get_secret_str("RECRAFT_API_BASE") or self.DEFAULT_BASE_URL
+ return join_api_base(base_url, self.IMAGE_GENERATION_ENDPOINT)
def validate_environment(
self,
diff --git a/litellm/llms/runwayml/image_generation/transformation.py b/litellm/llms/runwayml/image_generation/transformation.py
index ddc1b9474ac..e721ac678a1 100644
--- a/litellm/llms/runwayml/image_generation/transformation.py
+++ b/litellm/llms/runwayml/image_generation/transformation.py
@@ -11,6 +11,7 @@ from litellm.constants import (
RUNWAYML_DEFAULT_API_VERSION,
RUNWAYML_POLLING_TIMEOUT,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
)
@@ -62,12 +63,8 @@ class RunwayMLImageGenerationConfig(BaseImageGenerationConfig):
Some providers need `model` in `api_base`
"""
- complete_url: str = api_base or get_secret_str("RUNWAYML_API_BASE") or self.DEFAULT_BASE_URL
-
- complete_url = complete_url.rstrip("/")
- if self.IMAGE_GENERATION_ENDPOINT:
- complete_url = f"{complete_url}/{self.IMAGE_GENERATION_ENDPOINT}"
- return complete_url
+ base_url: Final = api_base or get_secret_str("RUNWAYML_API_BASE") or self.DEFAULT_BASE_URL
+ return join_api_base(base_url, self.IMAGE_GENERATION_ENDPOINT)
def validate_environment(
self,
@@ -226,8 +223,7 @@ class RunwayMLImageGenerationConfig(BaseImageGenerationConfig):
start_time: Final = time.time()
# Build task status URL
- api_base = api_base.rstrip("/")
- task_url: Final = f"{api_base}/v1/tasks/{task_id}"
+ task_url: Final = join_api_base(api_base, f"v1/tasks/{task_id}")
verbose_logger.debug("Polling RunwayML task: %s", task_url)
@@ -275,8 +271,7 @@ class RunwayMLImageGenerationConfig(BaseImageGenerationConfig):
start_time: Final = time.time()
# Build task status URL
- api_base = api_base.rstrip("/")
- task_url: Final = f"{api_base}/v1/tasks/{task_id}"
+ task_url: Final = join_api_base(api_base, f"v1/tasks/{task_id}")
verbose_logger.debug("Polling RunwayML task (async): %s", task_url)
diff --git a/litellm/llms/runwayml/text_to_speech/transformation.py b/litellm/llms/runwayml/text_to_speech/transformation.py
index d41ffea17dc..1f052648f13 100644
--- a/litellm/llms/runwayml/text_to_speech/transformation.py
+++ b/litellm/llms/runwayml/text_to_speech/transformation.py
@@ -18,6 +18,7 @@ from litellm.constants import (
RUNWAYML_DEFAULT_API_VERSION,
RUNWAYML_POLLING_TIMEOUT,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.text_to_speech.transformation import (
BaseTextToSpeechConfig,
TextToSpeechRequestData,
@@ -230,10 +231,8 @@ class RunwayMLTextToSpeechConfig(BaseTextToSpeechConfig):
"""
Get the complete URL for RunwayML TTS request
"""
- complete_url = api_base or get_secret_str("RUNWAYML_API_BASE") or self.DEFAULT_BASE_URL
-
- complete_url = complete_url.rstrip("/")
- return f"{complete_url}/{self.TTS_ENDPOINT_PATH}"
+ base_url: Final = api_base or get_secret_str("RUNWAYML_API_BASE") or self.DEFAULT_BASE_URL
+ return join_api_base(base_url, self.TTS_ENDPOINT_PATH)
@staticmethod
def _check_timeout(start_time: float, timeout_secs: float) -> None:
@@ -311,8 +310,7 @@ class RunwayMLTextToSpeechConfig(BaseTextToSpeechConfig):
start_time: Final = time.time()
# Build task status URL
- api_base = api_base.rstrip("/")
- task_url: Final = f"{api_base}/v1/tasks/{task_id}"
+ task_url: Final = join_api_base(api_base, f"v1/tasks/{task_id}")
verbose_logger.debug("Polling RunwayML TTS task: %s", task_url)
@@ -359,8 +357,7 @@ class RunwayMLTextToSpeechConfig(BaseTextToSpeechConfig):
start_time: Final = time.time()
# Build task status URL
- api_base = api_base.rstrip("/")
- task_url: Final = f"{api_base}/v1/tasks/{task_id}"
+ task_url: Final = join_api_base(api_base, f"v1/tasks/{task_id}")
verbose_logger.debug("Polling RunwayML TTS task (async): %s", task_url)
diff --git a/litellm/llms/runwayml/videos/transformation.py b/litellm/llms/runwayml/videos/transformation.py
index 8b765b1390b..106567f6fba 100644
--- a/litellm/llms/runwayml/videos/transformation.py
+++ b/litellm/llms/runwayml/videos/transformation.py
@@ -9,7 +9,7 @@ from typing_extensions import ReadOnly, TypedDict
import litellm
from litellm.constants import RUNWAYML_DEFAULT_API_VERSION
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.videos.transformation import BaseVideoConfig
from litellm.llms.custom_httpx.http_handler import (
@@ -274,7 +274,7 @@ class RunwayMLVideoConfig(BaseVideoConfig):
files_list: Final[RequestFiles] = []
- return request_data, files_list, f"{api_base}/{endpoint}"
+ return request_data, files_list, join_api_base(api_base, endpoint)
def _select_generation_endpoint(self, request_data: Mapping[str, object]) -> str:
if request_data.get("promptVideo") is not None or request_data.get("videoUri") is not None:
@@ -413,7 +413,7 @@ class RunwayMLVideoConfig(BaseVideoConfig):
encoded_video_id: Final = encode_url_path_segment(original_video_id, field_name="video_id")
# Get task status to retrieve video URL
- url: Final = f"{api_base}/tasks/{encoded_video_id}"
+ url: Final = join_api_base(api_base, f"tasks/{encoded_video_id}")
return url, dict[str, str]()
@@ -571,7 +571,7 @@ class RunwayMLVideoConfig(BaseVideoConfig):
encoded_video_id: Final = encode_url_path_segment(original_video_id, field_name="video_id")
# Construct the URL for task cancellation
- url: Final = f"{api_base}/tasks/{encoded_video_id}/cancel"
+ url: Final = join_api_base(api_base, f"tasks/{encoded_video_id}/cancel")
return url, dict[str, str]()
@@ -608,7 +608,7 @@ class RunwayMLVideoConfig(BaseVideoConfig):
encoded_video_id: Final = encode_url_path_segment(original_video_id, field_name="video_id")
# Construct the full URL for task status retrieval
- url: Final = f"{api_base}/tasks/{encoded_video_id}"
+ url: Final = join_api_base(api_base, f"tasks/{encoded_video_id}")
# Empty dict for GET request (no body)
return url, dict[str, str]()
diff --git a/litellm/llms/s3_vectors/vector_stores/transformation.py b/litellm/llms/s3_vectors/vector_stores/transformation.py
index 044315168d2..2b0de3dc9ea 100644
--- a/litellm/llms/s3_vectors/vector_stores/transformation.py
+++ b/litellm/llms/s3_vectors/vector_stores/transformation.py
@@ -3,6 +3,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.vector_store.transformation import (
BaseQueryEmbeddingVectorStoreConfig,
VectorStoreEmbeddingExecutor,
@@ -131,7 +132,7 @@ class S3VectorsVectorStoreConfig(BaseQueryEmbeddingVectorStoreConfig, BaseAWSLLM
litellm_logging_obj: LiteLLMLoggingObj,
) -> tuple[str, dict[str, object]]:
litellm_logging_obj.model_call_details["query"] = query_text
- return f"{api_base}/QueryVectors", {
+ return join_api_base(api_base, "QueryVectors"), {
"vectorBucketName": bucket_name,
"indexName": index_name,
"queryVector": {"float32": list(query_vector)},
diff --git a/litellm/llms/sambanova/embedding/transformation.py b/litellm/llms/sambanova/embedding/transformation.py
index cf39c0a092f..93cbb13ac47 100644
--- a/litellm/llms/sambanova/embedding/transformation.py
+++ b/litellm/llms/sambanova/embedding/transformation.py
@@ -10,6 +10,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.secret_managers.main import get_secret_str
@@ -37,11 +38,7 @@ class SambaNovaEmbeddingConfig(BaseEmbeddingConfig):
) -> str:
if api_base is None:
raise ValueError("api_base is required for SambaNova embeddings")
- # Remove trailing slashes and ensure clean base URL
- api_base = api_base.rstrip("/")
- if not api_base.endswith("/embeddings"):
- api_base = f"{api_base}/embeddings"
- return api_base
+ return ensure_api_route(api_base, "embeddings")
def validate_environment(
self,
diff --git a/litellm/llms/sap/chat/transformation.py b/litellm/llms/sap/chat/transformation.py
index a82dfe72322..8a7baf18e8f 100755
--- a/litellm/llms/sap/chat/transformation.py
+++ b/litellm/llms/sap/chat/transformation.py
@@ -11,6 +11,7 @@ import httpx
from pydantic import BaseModel
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.types.llms.openai import AllMessageValues
from litellm.types.utils import ModelResponse
@@ -185,12 +186,12 @@ class GenAIHubOrchestrationConfig(OpenAIGPTConfig):
# Keep a short, tight client lifecycle here to avoid fd leaks
client: Final = litellm.module_level_client
# with httpx.Client(timeout=30) as client:
- deployments: Final = client.get(f"{self.base_url}/lm/deployments", headers=self.headers).json()
+ deployments: Final = client.get(join_api_base(self.base_url, "lm/deployments"), headers=self.headers).json()
valid: Final[list[tuple[str, str]]] = []
for dep in deployments.get("resources", []):
if dep.get("scenarioId") == "orchestration":
cfg = client.get(
- f"{self.base_url}/lm/configurations/{dep['configurationId']}",
+ join_api_base(self.base_url, f"lm/configurations/{dep['configurationId']}"),
headers=self.headers,
).json()
if cfg.get("executableId") == "orchestration":
@@ -263,7 +264,7 @@ class GenAIHubOrchestrationConfig(OpenAIGPTConfig):
litellm_params: dict,
stream: bool | None = None,
):
- api_base_: Final = f"{self.deployment_url}/v2/completion"
+ api_base_: Final = join_api_base(self.deployment_url, "v2/completion")
return api_base_
def _build_prompt_module(
diff --git a/litellm/llms/sap/credentials.py b/litellm/llms/sap/credentials.py
index b1618f38623..2622d0fa9b6 100644
--- a/litellm/llms/sap/credentials.py
+++ b/litellm/llms/sap/credentials.py
@@ -16,6 +16,7 @@ from typing_extensions import NotRequired, ReadOnly, TypedDict
import litellm
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.custom_httpx.http_handler import HTTPHandler, get_httpx_client
AUTH_ENDPOINT_SUFFIX: Final = "/oauth/token"
@@ -112,7 +113,7 @@ CREDENTIAL_VALUES: Final[list[CredentialsValue]] = [
CredentialsValue(
"base_url",
("serviceurls", "AI_API_URL"),
- transform_fn=lambda url: url.rstrip("/") + ("" if url.endswith("/v2") else "/v2"),
+ transform_fn=lambda url: ensure_api_route(url, "v2"),
),
CredentialsValue(
"cert_url",
diff --git a/litellm/llms/sap/embed/transformation.py b/litellm/llms/sap/embed/transformation.py
index e91dfb91081..107fef0b355 100644
--- a/litellm/llms/sap/embed/transformation.py
+++ b/litellm/llms/sap/embed/transformation.py
@@ -8,6 +8,7 @@ from typing import Final, Literal
import httpx
from pydantic import Field
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.embedding.transformation import (
BaseEmbeddingConfig,
LiteLLMLoggingObj,
@@ -101,11 +102,11 @@ class GenAIHubEmbeddingConfig(BaseEmbeddingConfig):
def deployment_url(self) -> str:
with httpx.Client(timeout=30) as client:
valid_deployments: Final = []
- deployments: Final = client.get(self.base_url + "/lm/deployments", headers=self.headers).json()
+ deployments: Final = client.get(join_api_base(self.base_url, "lm/deployments"), headers=self.headers).json()
for deployment in deployments.get("resources", []):
if deployment["scenarioId"] == "orchestration":
config_details = client.get(
- self.base_url + f"/lm/configurations/{deployment['configurationId']}",
+ join_api_base(self.base_url, f"lm/configurations/{deployment['configurationId']}"),
headers=self.headers,
).json()
if config_details["executableId"] == "orchestration":
@@ -144,7 +145,7 @@ class GenAIHubEmbeddingConfig(BaseEmbeddingConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- url: Final = self.deployment_url.rstrip("/") + "/v2/embeddings"
+ url: Final = join_api_base(self.deployment_url, "v2/embeddings")
return url
def transform_embedding_request(
diff --git a/litellm/llms/scaledown/chat/transformation.py b/litellm/llms/scaledown/chat/transformation.py
index 44eb4785615..3224bc97148 100644
--- a/litellm/llms/scaledown/chat/transformation.py
+++ b/litellm/llms/scaledown/chat/transformation.py
@@ -9,6 +9,7 @@ from urllib.parse import unquote
import httpx
from pydantic import JsonValue, TypeAdapter
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import AllMessageValues
@@ -103,8 +104,8 @@ class ScaleDownChatConfig(BaseConfig):
base: Final = _root(api_base or _trusted_root())
operation: Final = _operation(model)
if operation in DECISIONS_MODELS:
- return f"{base}/v1/scaledown"
- return f"{base}{NATIVE_PATHS[operation]}"
+ return join_api_base(base, "v1/scaledown")
+ return join_api_base(base, NATIVE_PATHS[operation])
def get_supported_openai_params(self, model: str) -> list[str]: # mutable-ok: BaseConfig signature
operation: Final = _operation(model)
diff --git a/litellm/llms/scaleway/audio_transcription/transformation.py b/litellm/llms/scaleway/audio_transcription/transformation.py
index 7ebf4475355..8913cc362b3 100644
--- a/litellm/llms/scaleway/audio_transcription/transformation.py
+++ b/litellm/llms/scaleway/audio_transcription/transformation.py
@@ -9,6 +9,7 @@ from typing import Final
import httpx
from litellm.litellm_core_utils.audio_utils.utils import process_audio_file
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.audio_transcription.transformation import (
AudioTranscriptionRequestData,
BaseAudioTranscriptionConfig,
@@ -58,8 +59,7 @@ class ScalewayAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- api_base = "https://api.scaleway.ai/v1" if api_base is None else api_base.rstrip("/")
- return f"{api_base}/audio/transcriptions"
+ return join_api_base("https://api.scaleway.ai/v1" if api_base is None else api_base, "audio/transcriptions")
def get_error_class(self, error_message: str, status_code: int, headers: dict | httpx.Headers) -> BaseLLMException:
return ScalewayAudioTranscriptionException(
diff --git a/litellm/llms/scaleway/rerank/transformation.py b/litellm/llms/scaleway/rerank/transformation.py
index 921252d7091..1de5d71f17a 100644
--- a/litellm/llms/scaleway/rerank/transformation.py
+++ b/litellm/llms/scaleway/rerank/transformation.py
@@ -9,6 +9,7 @@ API reference: https://www.scaleway.com/en/developers/api/generative-apis/#path-
from collections.abc import Mapping
from typing import Final
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.jina_ai.rerank.transformation import JinaAIRerankConfig
from litellm.secret_managers.main import get_secret_str
@@ -25,8 +26,8 @@ class ScalewayRerankConfig(JinaAIRerankConfig):
model: str,
optional_params: Mapping[str, object] | None = None,
) -> str:
- base: Final = SCALEWAY_API_BASE if api_base is None else api_base.rstrip("/")
- return f"{base}/rerank"
+ base: Final = SCALEWAY_API_BASE if api_base is None else api_base
+ return join_api_base(base, "rerank")
def validate_environment(
self,
diff --git a/litellm/llms/searxng/search/transformation.py b/litellm/llms/searxng/search/transformation.py
index 265a3db72f5..b1753821cd2 100644
--- a/litellm/llms/searxng/search/transformation.py
+++ b/litellm/llms/searxng/search/transformation.py
@@ -11,6 +11,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
SearchResponse,
@@ -101,12 +102,7 @@ class SearXNGSearchConfig(BaseSearchConfig):
"or pass `api_base` parameter. Example: os.environ['SEARXNG_API_BASE'] = 'https://your-searxng-instance.com'"
)
- # Append "/search" to the api base if it's not already there
- if not api_base.endswith("/search"):
- if api_base.endswith("/"):
- api_base = f"{api_base}search"
- else:
- api_base = f"{api_base}/search"
+ api_base = ensure_api_route(api_base, "search")
# Build query parameters from the transformed request body
if data and isinstance(data, dict) and "_searxng_params" in data:
diff --git a/litellm/llms/serper/search/transformation.py b/litellm/llms/serper/search/transformation.py
index 2539610ea59..e82427e4d0c 100644
--- a/litellm/llms/serper/search/transformation.py
+++ b/litellm/llms/serper/search/transformation.py
@@ -11,6 +11,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
SearchResponse,
@@ -84,12 +85,7 @@ class SerperSearchConfig(BaseSearchConfig):
Get complete URL for Search endpoint.
"""
api_base = api_base or get_secret_str("SERPER_API_BASE") or self.SERPER_API_BASE
- api_base = api_base.rstrip("/")
-
- if not api_base.endswith("/search"):
- api_base = f"{api_base}/search"
-
- return api_base
+ return ensure_api_route(api_base, "search")
def transform_search_request(
self,
diff --git a/litellm/llms/snowflake/chat/transformation.py b/litellm/llms/snowflake/chat/transformation.py
index 6a480904a3e..030413030eb 100644
--- a/litellm/llms/snowflake/chat/transformation.py
+++ b/litellm/llms/snowflake/chat/transformation.py
@@ -25,6 +25,7 @@ from litellm.litellm_core_utils.prompt_templates.factory import (
select_anthropic_content_block_type_for_file,
)
from litellm.litellm_core_utils.prompt_templates.image_handling import async_inline_remote_media
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.anthropic.chat.handler import ModelResponseIterator as AnthropicStreamParser
from litellm.llms.anthropic.chat.transformation import AnthropicConfig
from litellm.llms.anthropic.common_utils import normalize_cache_control_in_anthropic_payload
@@ -239,8 +240,8 @@ class SnowflakeConfig(SnowflakeBaseConfig, OpenAIGPTConfig):
) -> str:
api_base = self._get_api_base(api_base, optional_params)
if is_snowflake_claude_model(model):
- return f"{api_base}/cortex/v1/messages"
- return f"{api_base}/cortex/v1/chat/completions"
+ return join_api_base(api_base, "cortex/v1/messages")
+ return join_api_base(api_base, "cortex/v1/chat/completions")
def validate_environment(
self,
diff --git a/litellm/llms/snowflake/embedding/transformation.py b/litellm/llms/snowflake/embedding/transformation.py
index ec482337e52..45bec1283a3 100644
--- a/litellm/llms/snowflake/embedding/transformation.py
+++ b/litellm/llms/snowflake/embedding/transformation.py
@@ -5,6 +5,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.types.llms.openai import AllEmbeddingInputValues
@@ -31,7 +32,7 @@ class SnowflakeEmbeddingConfig(SnowflakeBaseConfig, BaseEmbeddingConfig):
) -> str:
api_base = self._get_api_base(api_base, optional_params)
- return f"{api_base}/cortex/inference:embed"
+ return join_api_base(api_base, "cortex/inference:embed")
def transform_embedding_request(
self,
diff --git a/litellm/llms/snowflake/utils.py b/litellm/llms/snowflake/utils.py
index a8c14d1d7f3..7cc88ab66e1 100644
--- a/litellm/llms/snowflake/utils.py
+++ b/litellm/llms/snowflake/utils.py
@@ -1,6 +1,7 @@
import re
from typing import TYPE_CHECKING, Any, Final
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import AllMessageValues
@@ -53,7 +54,7 @@ class SnowflakeBaseConfig:
optional_params[param] = value
return optional_params
- def _get_api_base(self, api_base, optional_params):
+ def _get_api_base(self, api_base: str | None, optional_params) -> str:
if not api_base:
if "account_id" in optional_params:
account_id = optional_params.pop("account_id")
@@ -65,10 +66,7 @@ class SnowflakeBaseConfig:
raise ValueError("Invalid account_id format")
api_base = f"https://{account_id}.snowflakecomputing.com/api/v2"
- api_base = api_base.rstrip("/")
- if not api_base.endswith("/api/v2"):
- api_base += "/api/v2"
- return api_base
+ return ensure_api_route(api_base, "api/v2")
def validate_environment(
self,
diff --git a/litellm/llms/soniox/audio_transcription/handler.py b/litellm/llms/soniox/audio_transcription/handler.py
index f86bc142898..a48e3e3ed4c 100644
--- a/litellm/llms/soniox/audio_transcription/handler.py
+++ b/litellm/llms/soniox/audio_transcription/handler.py
@@ -31,6 +31,7 @@ from litellm.litellm_core_utils.audio_utils.utils import (
get_audio_file_name,
process_audio_file,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.custom_httpx.http_handler import (
AsyncHTTPHandler,
HTTPHandler,
@@ -329,7 +330,7 @@ class SonioxAudioTranscriptionHandler:
input=None,
api_key=api_key,
additional_args={
- "api_base": f"{api_base}/v1/transcriptions",
+ "api_base": join_api_base(api_base, "v1/transcriptions"),
"atranscription": True,
"complete_input_dict": SonioxAudioTranscriptionHandler._redact_body_for_logging(body),
},
@@ -451,7 +452,7 @@ class SonioxAudioTranscriptionHandler:
self._safe_log_pre_call(logging_obj, api_key, base_url, body)
create_resp: Final = http_client.post(
- url=f"{base_url}/v1/transcriptions",
+ url=join_api_base(base_url, "v1/transcriptions"),
headers=auth_headers,
json=body,
timeout=timeout,
@@ -472,7 +473,7 @@ class SonioxAudioTranscriptionHandler:
)
transcript_resp: Final = http_client.get(
- url=f"{base_url}/v1/transcriptions/{transcription_id}/transcript",
+ url=join_api_base(base_url, f"v1/transcriptions/{transcription_id}/transcript"),
headers=auth_headers,
timeout=timeout,
)
@@ -528,7 +529,7 @@ class SonioxAudioTranscriptionHandler:
# `Authorization` header is fine; httpx sets multipart Content-Type.
upload_headers: Final = {"Authorization": auth_headers["Authorization"]}
resp: Final = http_client.post(
- url=f"{base_url}/v1/files",
+ url=join_api_base(base_url, "v1/files"),
headers=upload_headers,
files=files,
timeout=timeout,
@@ -550,7 +551,7 @@ class SonioxAudioTranscriptionHandler:
) -> _TranscriptionMeta:
for _ in range(max_attempts):
resp = http_client.get(
- url=f"{base_url}/v1/transcriptions/{transcription_id}",
+ url=join_api_base(base_url, f"v1/transcriptions/{transcription_id}"),
headers=auth_headers,
timeout=timeout,
)
@@ -593,7 +594,7 @@ class SonioxAudioTranscriptionHandler:
if "transcription" in cleanup and transcription_id:
try:
http_client.delete(
- url=f"{base_url}/v1/transcriptions/{transcription_id}",
+ url=join_api_base(base_url, f"v1/transcriptions/{transcription_id}"),
headers=auth_headers,
timeout=timeout,
)
@@ -605,7 +606,7 @@ class SonioxAudioTranscriptionHandler:
if "file" in cleanup and file_id_to_cleanup:
try:
http_client.delete(
- url=f"{base_url}/v1/files/{file_id_to_cleanup}",
+ url=join_api_base(base_url, f"v1/files/{file_id_to_cleanup}"),
headers=auth_headers,
timeout=timeout,
)
@@ -685,7 +686,7 @@ class SonioxAudioTranscriptionHandler:
self._safe_log_pre_call(logging_obj, api_key, base_url, body)
create_resp: Final = await http_client.post(
- url=f"{base_url}/v1/transcriptions",
+ url=join_api_base(base_url, "v1/transcriptions"),
headers=auth_headers,
json=body,
timeout=timeout,
@@ -706,7 +707,7 @@ class SonioxAudioTranscriptionHandler:
)
transcript_resp: Final = await http_client.get(
- url=f"{base_url}/v1/transcriptions/{transcription_id}/transcript",
+ url=join_api_base(base_url, f"v1/transcriptions/{transcription_id}/transcript"),
headers=auth_headers,
timeout=timeout,
)
@@ -761,7 +762,7 @@ class SonioxAudioTranscriptionHandler:
}
upload_headers: Final = {"Authorization": auth_headers["Authorization"]}
resp: Final = await http_client.post(
- url=f"{base_url}/v1/files",
+ url=join_api_base(base_url, "v1/files"),
headers=upload_headers,
files=files,
timeout=timeout,
@@ -783,7 +784,7 @@ class SonioxAudioTranscriptionHandler:
) -> _TranscriptionMeta:
for _ in range(max_attempts):
resp = await http_client.get(
- url=f"{base_url}/v1/transcriptions/{transcription_id}",
+ url=join_api_base(base_url, f"v1/transcriptions/{transcription_id}"),
headers=auth_headers,
timeout=timeout,
)
@@ -826,7 +827,7 @@ class SonioxAudioTranscriptionHandler:
if "transcription" in cleanup and transcription_id:
try:
await http_client.delete(
- url=f"{base_url}/v1/transcriptions/{transcription_id}",
+ url=join_api_base(base_url, f"v1/transcriptions/{transcription_id}"),
headers=auth_headers,
timeout=timeout,
)
@@ -838,7 +839,7 @@ class SonioxAudioTranscriptionHandler:
if "file" in cleanup and file_id_to_cleanup:
try:
await http_client.delete(
- url=f"{base_url}/v1/files/{file_id_to_cleanup}",
+ url=join_api_base(base_url, f"v1/files/{file_id_to_cleanup}"),
headers=auth_headers,
timeout=timeout,
)
diff --git a/litellm/llms/stability/image_generation/transformation.py b/litellm/llms/stability/image_generation/transformation.py
index 1856523ce49..a53fd087098 100644
--- a/litellm/llms/stability/image_generation/transformation.py
+++ b/litellm/llms/stability/image_generation/transformation.py
@@ -12,6 +12,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
from pydantic import ConfigDict, TypeAdapter
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_generation.transformation import (
BaseImageGenerationConfig,
)
@@ -131,11 +132,10 @@ class StabilityImageGenerationConfig(BaseImageGenerationConfig):
"""
Get the complete URL for the Stability AI API request.
"""
- base_url: str = api_base or get_secret_str("STABILITY_API_BASE") or self.DEFAULT_BASE_URL
- base_url = base_url.rstrip("/")
+ base_url: Final = api_base or get_secret_str("STABILITY_API_BASE") or self.DEFAULT_BASE_URL
endpoint: Final = self._get_model_endpoint(model)
- return f"{base_url}{endpoint}"
+ return join_api_base(base_url, endpoint)
def validate_environment(
self,
diff --git a/litellm/llms/tavily/search/transformation.py b/litellm/llms/tavily/search/transformation.py
index d0531e4c041..b46f974e0db 100644
--- a/litellm/llms/tavily/search/transformation.py
+++ b/litellm/llms/tavily/search/transformation.py
@@ -11,6 +11,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
SearchResponse,
@@ -90,11 +91,7 @@ class TavilySearchConfig(BaseSearchConfig):
"""
Get complete URL for Search endpoint.
"""
- resolved_api_base: Final = api_base or get_secret_str("TAVILY_API_BASE") or self.TAVILY_API_BASE
- clean_api_base: Final = resolved_api_base.rstrip("/")
- if clean_api_base.endswith("/search"):
- return clean_api_base
- return f"{clean_api_base}/search"
+ return ensure_api_route(api_base or get_secret_str("TAVILY_API_BASE") or self.TAVILY_API_BASE, "search")
def transform_search_request(
self,
diff --git a/litellm/llms/tencent/chat/transformation.py b/litellm/llms/tencent/chat/transformation.py
index 7e80b0012df..1f0bcd218d6 100644
--- a/litellm/llms/tencent/chat/transformation.py
+++ b/litellm/llms/tencent/chat/transformation.py
@@ -9,6 +9,7 @@ from typing import Final, TypedDict
from typing_extensions import ReadOnly
import litellm
+from litellm.litellm_core_utils.url_utils import ensure_api_route, join_api_base
from litellm.secret_managers.main import get_secret_str
from litellm.utils import supports_reasoning
@@ -137,7 +138,4 @@ class TencentChatConfig(OpenAIGPTConfig):
if api_base.endswith("/chat/completions"):
return api_base
- if not api_base.endswith("/v1"):
- api_base = f"{api_base}/v1"
-
- return f"{api_base}/chat/completions"
+ return join_api_base(ensure_api_route(api_base, "v1"), "chat/completions")
diff --git a/litellm/llms/tencent/messages/transformation.py b/litellm/llms/tencent/messages/transformation.py
index c56ecaeb51c..1e52bc37d47 100644
--- a/litellm/llms/tencent/messages/transformation.py
+++ b/litellm/llms/tencent/messages/transformation.py
@@ -8,6 +8,7 @@ alongside its standard OpenAI-compatible chat completions endpoint.
from typing import Any
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.anthropic.pass_through.messages.transformation import (
AnthropicMessagesConfig,
)
@@ -82,4 +83,4 @@ class TencentAnthropicMessagesConfig(AnthropicMessagesConfig):
elif base_url.endswith("/v1"):
base_url = base_url[: -len("/v1")]
- return f"{base_url}/v1/messages"
+ return join_api_base(base_url, "v1/messages")
diff --git a/litellm/llms/topaz/image_variations/transformation.py b/litellm/llms/topaz/image_variations/transformation.py
index 8d03b73e94e..b3dded36f49 100644
--- a/litellm/llms/topaz/image_variations/transformation.py
+++ b/litellm/llms/topaz/image_variations/transformation.py
@@ -7,6 +7,7 @@ from typing import TYPE_CHECKING, Final
from aiohttp import ClientResponse
from httpx import Headers, Response
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import (
BaseLLMException,
LiteLLMLoggingObj,
@@ -39,8 +40,7 @@ class TopazImageVariationConfig(TopazModelInfo, BaseImageVariationConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- clean_api_base: Final = (api_base or "https://api.topazlabs.com").rstrip("/")
- return f"{clean_api_base}/image/v1/enhance"
+ return join_api_base(api_base or "https://api.topazlabs.com", "image/v1/enhance")
def map_openai_params(
self,
diff --git a/litellm/llms/vercel_ai_gateway/chat/transformation.py b/litellm/llms/vercel_ai_gateway/chat/transformation.py
index 19c4c2659ca..3fc74b6acc1 100644
--- a/litellm/llms/vercel_ai_gateway/chat/transformation.py
+++ b/litellm/llms/vercel_ai_gateway/chat/transformation.py
@@ -11,6 +11,7 @@ from typing import Final
import httpx
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import AllMessageValues
@@ -92,7 +93,7 @@ class VercelAIGatewayConfig(OpenAIGPTConfig):
if api_base is None:
api_base = "https://ai-gateway.vercel.sh/v1"
- models_url: Final = f"{api_base}/models"
+ models_url: Final = join_api_base(api_base, "models")
response: Final = litellm.module_level_client.get(url=models_url)
if response.status_code != 200:
diff --git a/litellm/llms/vercel_ai_gateway/embedding/transformation.py b/litellm/llms/vercel_ai_gateway/embedding/transformation.py
index 9cfaaab89f0..01d7b7766c0 100644
--- a/litellm/llms/vercel_ai_gateway/embedding/transformation.py
+++ b/litellm/llms/vercel_ai_gateway/embedding/transformation.py
@@ -12,6 +12,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.secret_managers.main import get_secret_str
@@ -77,12 +78,8 @@ class VercelAIGatewayEmbeddingConfig(BaseEmbeddingConfig):
"""
Get the complete URL for Vercel AI Gateway Embedding API endpoint.
"""
- if api_base:
- api_base = api_base.rstrip("/")
- else:
- api_base = get_secret_str("VERCEL_AI_GATEWAY_API_BASE") or "https://ai-gateway.vercel.sh/v1"
-
- return f"{api_base}/embeddings"
+ base: Final = api_base or get_secret_str("VERCEL_AI_GATEWAY_API_BASE") or "https://ai-gateway.vercel.sh/v1"
+ return join_api_base(base, "embeddings")
def transform_embedding_request(
self,
diff --git a/litellm/llms/vertex_ai/audio_transcription/gemini_transcribe_transformation.py b/litellm/llms/vertex_ai/audio_transcription/gemini_transcribe_transformation.py
index f4db5eb110c..39d2e9b4c49 100644
--- a/litellm/llms/vertex_ai/audio_transcription/gemini_transcribe_transformation.py
+++ b/litellm/llms/vertex_ai/audio_transcription/gemini_transcribe_transformation.py
@@ -10,6 +10,7 @@ from litellm.litellm_core_utils.audio_utils.utils import (
normalize_transcription_language_to_bcp47,
process_audio_file,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.audio_transcription.transformation import (
AudioTranscriptionRequestData,
BaseAudioTranscriptionConfig,
@@ -129,10 +130,10 @@ class VertexGeminiAudioTranscriptionConfig(BaseAudioTranscriptionConfig, VertexB
project_id: Final = validate_vertex_transcription_project_id(
self.safe_get_vertex_ai_project(vertex_params) or self._resolve_project_id_from_credentials(vertex_params)
)
- base_url: Final = (api_base or get_vertex_base_url(location)).rstrip("/")
+ base_url: Final = api_base or get_vertex_base_url(location)
bare_model: Final = model.removeprefix("vertex_ai/")
model_path: Final = f"projects/{project_id}/locations/{location}/publishers/google/models/{bare_model}"
- return f"{base_url}/v1/{model_path}:generateContent"
+ return join_api_base(base_url, f"v1/{model_path}:generateContent")
def _resolve_project_id_from_credentials(self, litellm_params: Mapping[str, object]) -> str:
vertex_params: Final = dict(litellm_params)
diff --git a/litellm/llms/vertex_ai/audio_transcription/transformation.py b/litellm/llms/vertex_ai/audio_transcription/transformation.py
index 5603be81095..4fdf93a6307 100644
--- a/litellm/llms/vertex_ai/audio_transcription/transformation.py
+++ b/litellm/llms/vertex_ai/audio_transcription/transformation.py
@@ -9,6 +9,7 @@ from litellm.litellm_core_utils.audio_utils.utils import (
normalize_transcription_language_to_bcp47,
process_audio_file,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.audio_transcription.transformation import (
AudioTranscriptionRequestData,
BaseAudioTranscriptionConfig,
@@ -126,8 +127,8 @@ class VertexAIAudioTranscriptionConfig(BaseAudioTranscriptionConfig, VertexBase)
project_id: Final = validate_vertex_transcription_project_id(
self.safe_get_vertex_ai_project(litellm_params) or self._resolve_project_id_from_credentials(litellm_params)
)
- base_url: Final = (api_base or f"https://{speech_to_text_host(location)}").rstrip("/")
- return f"{base_url}/v2/projects/{project_id}/locations/{location}/recognizers/_:recognize"
+ base_url: Final = api_base or f"https://{speech_to_text_host(location)}"
+ return join_api_base(base_url, f"v2/projects/{project_id}/locations/{location}/recognizers/_:recognize")
def _resolve_project_id_from_credentials(self, litellm_params: dict) -> str:
_, project_id = self._ensure_access_token(
diff --git a/litellm/llms/vertex_ai/batches/handler.py b/litellm/llms/vertex_ai/batches/handler.py
index de5fe39cfc9..7c4238e3498 100644
--- a/litellm/llms/vertex_ai/batches/handler.py
+++ b/litellm/llms/vertex_ai/batches/handler.py
@@ -10,6 +10,7 @@ import litellm
from litellm.litellm_core_utils.url_utils import (
async_safe_get,
encode_url_path_segment,
+ join_api_base,
safe_get,
)
from litellm.llms.custom_httpx.http_handler import (
@@ -194,7 +195,7 @@ class VertexAIBatchPrediction(VertexLLM):
api_base_path: Final = urlparse(api_base).path.rstrip("/")
if api_base_path in ("/v1", "/v1beta1"):
return graft_default_vertex_path(api_base=api_base, default_url=default_endpoint_url)
- return api_base.rstrip("/") + urlparse(default_endpoint_url).path
+ return join_api_base(api_base, urlparse(default_endpoint_url).path)
def _resolve_fine_tuned_endpoint_model(
self,
diff --git a/litellm/llms/vertex_ai/files/transformation.py b/litellm/llms/vertex_ai/files/transformation.py
index d1e3f10fdd4..e36aba00293 100644
--- a/litellm/llms/vertex_ai/files/transformation.py
+++ b/litellm/llms/vertex_ai/files/transformation.py
@@ -37,6 +37,7 @@ from litellm.litellm_core_utils.prompt_templates.common_utils import (
extract_file_data,
extract_file_metadata,
)
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.base_utils import map_developer_role_to_system_role
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.files.batch_records import responses_batch_body_to_chat_body
@@ -1083,12 +1084,12 @@ class VertexAIFilesConfig(VertexBase, BaseFilesConfig):
if object_prefix:
object_name = f"{object_prefix}/{object_name}"
encoded_object_name: Final = encode_gcs_object_name_for_url(object_name)
- endpoint: Final = f"upload/storage/v1/b/{bucket_name}/o?uploadType=media&name={encoded_object_name}"
+ upload_path: Final = f"upload/storage/v1/b/{bucket_name}/o"
api_base = api_base or "https://storage.googleapis.com"
if not api_base:
raise ValueError("api_base is required")
- return f"{api_base}/{endpoint}"
+ return f"{join_api_base(api_base, upload_path)}?uploadType=media&name={encoded_object_name}"
def get_supported_openai_params(self, model: str) -> list[OpenAICreateFileRequestOptionalParams]:
return []
diff --git a/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py b/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py
index f916d494aad..085ff715c30 100644
--- a/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py
+++ b/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py
@@ -12,6 +12,7 @@ from pydantic import ConfigDict, TypeAdapter
import litellm
from litellm.constants import DEFAULT_MAX_RECURSE_DEPTH
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.image_edit.transformation import BaseImageEditConfig
from litellm.llms.vertex_ai.common_utils import get_vertex_base_url
from litellm.llms.vertex_ai.gemini.vertex_and_google_ai_studio_gemini import VertexLLM
@@ -143,12 +144,12 @@ class VertexAIImagenImageEditConfig(BaseImageEditConfig, VertexLLM):
if model.startswith("vertex_ai/"):
model_name = model.replace("vertex_ai/", "")
- if api_base:
- base_url = api_base.rstrip("/")
- else:
- base_url = get_vertex_base_url(vertex_location)
+ base_url: Final = api_base or get_vertex_base_url(vertex_location)
- return f"{base_url}/v1/projects/{vertex_project}/locations/{vertex_location}/publishers/google/models/{model_name}:predict"
+ return join_api_base(
+ base_url,
+ f"v1/projects/{vertex_project}/locations/{vertex_location}/publishers/google/models/{model_name}:predict",
+ )
def transform_image_edit_request(
self,
diff --git a/litellm/llms/vertex_ai/interactions/transformation.py b/litellm/llms/vertex_ai/interactions/transformation.py
index 36965fe666f..b8e13059917 100644
--- a/litellm/llms/vertex_ai/interactions/transformation.py
+++ b/litellm/llms/vertex_ai/interactions/transformation.py
@@ -2,7 +2,7 @@ from collections.abc import Callable, Mapping
from dataclasses import dataclass
from typing import Final
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.gemini.interactions.transformation import GoogleAIStudioInteractionsConfig
from litellm.llms.vertex_ai.common_utils import validate_vertex_location
from litellm.llms.vertex_ai.vertex_llm_base import VertexBase
@@ -22,9 +22,9 @@ class VertexInteractionsTarget:
@property
def collection_url(self) -> str:
- return (
- f"{self.base_url}/{VERTEX_INTERACTIONS_API_VERSION}"
- f"/projects/{self.project_id}/locations/{self.location}/interactions"
+ return join_api_base(
+ self.base_url,
+ f"{VERTEX_INTERACTIONS_API_VERSION}/projects/{self.project_id}/locations/{self.location}/interactions",
)
def interaction_url(self, interaction_id: str) -> str:
diff --git a/litellm/llms/vertex_ai/realtime/transformation.py b/litellm/llms/vertex_ai/realtime/transformation.py
index 55f1cf58cbe..13f547b3b92 100644
--- a/litellm/llms/vertex_ai/realtime/transformation.py
+++ b/litellm/llms/vertex_ai/realtime/transformation.py
@@ -21,6 +21,7 @@ from collections.abc import Awaitable, Callable
from typing import Final
from litellm import verbose_logger
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.gemini.realtime.transformation import GeminiRealtimeConfig
from litellm.llms.vertex_ai.audio_transcription.realtime_transformation import (
VertexChirpRealtimeConfig,
@@ -68,9 +69,8 @@ class VertexAIRealtimeConfig(GeminiRealtimeConfig):
"""
if api_base:
# Allow callers to supply a fully-qualified wss:// base URL.
- base = api_base.rstrip("/")
- base = base.replace("https://", "wss://").replace("http://", "ws://")
- return f"{base}/ws/google.cloud.aiplatform.v1.LlmBidiService/BidiGenerateContent"
+ base: Final = api_base.replace("https://", "wss://").replace("http://", "ws://")
+ return join_api_base(base, "ws/google.cloud.aiplatform.v1.LlmBidiService/BidiGenerateContent")
host: Final = get_vertex_base_url(self._location).removeprefix("https://")
return f"wss://{host}/ws/google.cloud.aiplatform.v1.LlmBidiService/BidiGenerateContent"
diff --git a/litellm/llms/vertex_ai/text_to_speech/transformation.py b/litellm/llms/vertex_ai/text_to_speech/transformation.py
index 1441d01e41e..c0082bc5ca6 100644
--- a/litellm/llms/vertex_ai/text_to_speech/transformation.py
+++ b/litellm/llms/vertex_ai/text_to_speech/transformation.py
@@ -23,7 +23,7 @@ from litellm.litellm_core_utils.audio_utils.utils import (
speech_media_type_from_audio_bytes,
)
from litellm.litellm_core_utils.core_helpers import as_str_mapping, drops_unsupported_params
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.text_to_speech.transformation import (
BaseTextToSpeechConfig,
@@ -1075,13 +1075,14 @@ class VertexAILyriaTextToSpeechConfig(VertexAITextToSpeechConfig):
},
)
location: Final = self.safe_get_vertex_ai_location(litellm_params) or self.get_default_vertex_location()
- base_url: Final = self.get_api_base(api_base=api_base, vertex_location=location).rstrip("/")
+ base_url: Final = self.get_api_base(api_base=api_base, vertex_location=location)
encoded_project: Final = encode_url_path_segment(project, field_name="project")
encoded_location: Final = encode_url_path_segment(location, field_name="location")
encoded_model: Final = encode_url_path_segment(base_model, field_name="model")
- return (
- f"{base_url}/v1/projects/{encoded_project}/locations/{encoded_location}"
- f"/publishers/google/models/{encoded_model}:predict"
+ return join_api_base(
+ base_url,
+ f"v1/projects/{encoded_project}/locations/{encoded_location}"
+ f"/publishers/google/models/{encoded_model}:predict",
)
def transform_text_to_speech_request(
diff --git a/litellm/llms/vertex_ai/vector_stores/rag_api/transformation.py b/litellm/llms/vertex_ai/vector_stores/rag_api/transformation.py
index 5c250fc1a7e..440604688bb 100644
--- a/litellm/llms/vertex_ai/vector_stores/rag_api/transformation.py
+++ b/litellm/llms/vertex_ai/vector_stores/rag_api/transformation.py
@@ -4,6 +4,7 @@ from typing import TYPE_CHECKING, Any, Final, Protocol
import httpx
from typing_extensions import ReadOnly, TypedDict
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.vector_store.transformation import BaseVectorStoreConfig
from litellm.llms.vertex_ai.common_utils import get_vertex_base_url
from litellm.llms.vertex_ai.vertex_llm_base import VertexBase
@@ -288,7 +289,7 @@ class VertexVectorStoreConfig(BaseVectorStoreConfig, VertexBase):
"""
Transform create request for Vertex AI RAG Corpus
"""
- url: Final = f"{api_base}/ragCorpora" # Base URL for creating RAG corpus
+ url: Final = join_api_base(api_base, "ragCorpora") # Base URL for creating RAG corpus
# Add metadata if provided
metadata: Final = vector_store_create_optional_params.get("metadata")
diff --git a/litellm/llms/vertex_ai/vertex_ai_partner_models/count_tokens/handler.py b/litellm/llms/vertex_ai/vertex_ai_partner_models/count_tokens/handler.py
index 7100c9efde3..4be41e26326 100644
--- a/litellm/llms/vertex_ai/vertex_ai_partner_models/count_tokens/handler.py
+++ b/litellm/llms/vertex_ai/vertex_ai_partner_models/count_tokens/handler.py
@@ -8,6 +8,7 @@ their respective publisher-specific count-tokens endpoints.
from typing import Any, Final
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.custom_httpx.http_handler import get_async_httpx_client
from litellm.llms.vertex_ai.common_utils import get_vertex_base_url
from litellm.llms.vertex_ai.vertex_llm_base import VertexBase
@@ -59,16 +60,14 @@ class VertexAIPartnerModelsTokenCounter(VertexBase):
publisher: Final = self._get_publisher_for_model(model)
# Use custom api_base if provided, otherwise construct default
- if api_base:
- base_url = api_base
- else:
- base_url = get_vertex_base_url(vertex_location)
+ base_url: Final = api_base or get_vertex_base_url(vertex_location)
# Construct the count-tokens endpoint
# Format: /v1/projects/{project}/locations/{location}/publishers/{publisher}/models/count-tokens:rawPredict
- endpoint: Final = (
- f"{base_url}/v1/projects/{project_id}/locations/{vertex_location}/"
- f"publishers/{publisher}/models/count-tokens:rawPredict"
+ endpoint: Final = join_api_base(
+ base_url,
+ f"v1/projects/{project_id}/locations/{vertex_location}/"
+ f"publishers/{publisher}/models/count-tokens:rawPredict",
)
return endpoint
diff --git a/litellm/llms/vertex_ai/vertex_llm_base.py b/litellm/llms/vertex_ai/vertex_llm_base.py
index 77508096e4a..6cff9729189 100644
--- a/litellm/llms/vertex_ai/vertex_llm_base.py
+++ b/litellm/llms/vertex_ai/vertex_llm_base.py
@@ -14,6 +14,7 @@ from urllib.parse import urlparse
import litellm
from litellm._logging import verbose_logger
from litellm.litellm_core_utils.asyncify import asyncify
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.custom_httpx.http_handler import AsyncHTTPHandler
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.vertex_ai import VERTEX_CREDENTIALS_TYPES, VertexPartnerProvider
@@ -295,23 +296,22 @@ class VertexBase:
if api_base is None:
api_base = get_vertex_base_url(vertex_location)
+ location_path: Final = f"projects/{vertex_project}/locations/{vertex_location}"
+ raw_predict_method: Final = "streamRawPredict" if stream else "rawPredict"
if partner == VertexPartnerProvider.llama:
- return f"{api_base}/v1/projects/{vertex_project}/locations/{vertex_location}/endpoints/openapi/chat/completions"
+ return join_api_base(api_base, f"v1/{location_path}/endpoints/openapi/chat/completions")
elif partner == VertexPartnerProvider.mistralai:
- if stream:
- return f"{api_base}/v1/projects/{vertex_project}/locations/{vertex_location}/publishers/mistralai/models/{model}:streamRawPredict"
- else:
- return f"{api_base}/v1/projects/{vertex_project}/locations/{vertex_location}/publishers/mistralai/models/{model}:rawPredict"
+ return join_api_base(
+ api_base, f"v1/{location_path}/publishers/mistralai/models/{model}:{raw_predict_method}"
+ )
elif partner == VertexPartnerProvider.ai21:
- if stream:
- return f"{api_base}/v1beta1/projects/{vertex_project}/locations/{vertex_location}/publishers/ai21/models/{model}:streamRawPredict"
- else:
- return f"{api_base}/v1beta1/projects/{vertex_project}/locations/{vertex_location}/publishers/ai21/models/{model}:rawPredict"
+ return join_api_base(
+ api_base, f"v1beta1/{location_path}/publishers/ai21/models/{model}:{raw_predict_method}"
+ )
elif partner == VertexPartnerProvider.claude:
- if stream:
- return f"{api_base}/v1/projects/{vertex_project}/locations/{vertex_location}/publishers/anthropic/models/{model}:streamRawPredict"
- else:
- return f"{api_base}/v1/projects/{vertex_project}/locations/{vertex_location}/publishers/anthropic/models/{model}:rawPredict"
+ return join_api_base(
+ api_base, f"v1/{location_path}/publishers/anthropic/models/{model}:{raw_predict_method}"
+ )
def get_complete_vertex_url(
self,
@@ -658,7 +658,7 @@ class VertexBase:
# For Gemini (Google AI Studio), construct the full path like other providers
if model is None:
raise ValueError("Model parameter is required for Gemini custom API base URLs")
- url = f"{api_base}/models/{model}:{endpoint}"
+ url = join_api_base(api_base, f"models/{model}:{endpoint}")
if gemini_api_key is None:
raise ValueError(
"Missing Gemini API key. Set the GEMINI_API_KEY or GOOGLE_API_KEY environment variable."
@@ -678,20 +678,17 @@ class VertexBase:
model_for_url: Final = get_vertex_base_model_name(model=model)
# Format: {api_base}/v1/projects/{project}/locations/{location}/endpoints/{model}:{endpoint}
version: Final = vertex_api_version or "v1"
- url = "{}/{}/projects/{}/locations/{}/endpoints/{}:{}".format(
- api_base.rstrip("/"),
- version,
- vertex_project,
- vertex_location,
- model_for_url,
- endpoint,
+ url = join_api_base(
+ api_base,
+ f"{version}/projects/{vertex_project}/locations/{vertex_location}"
+ f"/endpoints/{model_for_url}:{endpoint}",
)
elif urlparse(api_base).path in ("", "/"):
- url = api_base.rstrip("/") + urlparse(url).path
+ url = join_api_base(api_base, urlparse(url).path)
elif urlparse(api_base).path.rstrip("/") in ("/v1", "/v1beta1") and "/projects/" in urlparse(url).path:
url = graft_default_vertex_path(api_base=api_base, default_url=url)
else:
- url = f"{api_base}:{endpoint}"
+ url = f"{api_base.rstrip('/')}:{endpoint}"
if stream is True:
parsed_stream_url: Final = urlparse(url)
stream_query: Final = f"{parsed_stream_url.query}&alt=sse" if parsed_stream_url.query else "alt=sse"
diff --git a/litellm/llms/vertex_ai/videos/transformation.py b/litellm/llms/vertex_ai/videos/transformation.py
index de368cea5ac..080899d08b0 100644
--- a/litellm/llms/vertex_ai/videos/transformation.py
+++ b/litellm/llms/vertex_ai/videos/transformation.py
@@ -18,6 +18,7 @@ from typing_extensions import ReadOnly
import litellm
from litellm.constants import DEFAULT_GOOGLE_VIDEO_DURATION_SECONDS
from litellm.images.utils import ImageEditRequestUtils
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.videos.transformation import BaseVideoConfig
from litellm.llms.vertex_ai.common_utils import (
convert_vertex_datetime_to_openai_datetime,
@@ -325,13 +326,10 @@ class VertexAIVideoConfig(BaseVideoConfig, VertexBase):
model_name: Final = model.replace("vertex_ai/", "")
# Construct the URL
- if api_base:
- base_url = api_base.rstrip("/")
- else:
- base_url = get_vertex_base_url(vertex_location)
+ base_url: Final = api_base or get_vertex_base_url(vertex_location)
- url: Final = (
- f"{base_url}/v1/projects/{vertex_project}/locations/{vertex_location}/publishers/google/models/{model_name}"
+ url: Final = join_api_base(
+ base_url, f"v1/projects/{vertex_project}/locations/{vertex_location}/publishers/google/models/{model_name}"
)
return url
@@ -480,7 +478,7 @@ class VertexAIVideoConfig(BaseVideoConfig, VertexBase):
# Construct the full URL including model ID
# URL format: https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT/locations/LOCATION/publishers/google/models/MODEL:fetchPredictOperation
# Strip trailing slashes from api_base and append model
- url: Final = f"{api_base.rstrip('/')}/{model}:fetchPredictOperation"
+ url: Final = join_api_base(api_base, f"{model}:fetchPredictOperation")
# Request body contains the operation name
params: Final = {"operationName": operation_name}
@@ -778,7 +776,7 @@ class VertexAIVideoConfig(BaseVideoConfig, VertexBase):
if vertex_params:
request_data["parameters"] = vertex_params
- edit_url: Final = f"{api_base.rstrip('/')}/{model}:predictLongRunning"
+ edit_url: Final = join_api_base(api_base, f"{model}:predictLongRunning")
return edit_url, request_data, None
def transform_video_edit_response(
diff --git a/litellm/llms/volcengine/embedding/transformation.py b/litellm/llms/volcengine/embedding/transformation.py
index b298c9996f7..cc1e7d59caa 100644
--- a/litellm/llms/volcengine/embedding/transformation.py
+++ b/litellm/llms/volcengine/embedding/transformation.py
@@ -10,6 +10,7 @@ from typing import Final
import httpx
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.types.llms.openai import AllEmbeddingInputValues, AllMessageValues
@@ -76,12 +77,8 @@ class VolcEngineEmbeddingConfig(BaseEmbeddingConfig):
Returns:
Complete URL for the embedding API endpoint
"""
- base_url: Final = get_volcengine_base_url(api_base)
- # Construct the complete URL with /embeddings endpoint
- if base_url.endswith("/api/v3"):
- return f"{base_url}/embeddings"
- else:
- return f"{base_url}/api/v3/embeddings"
+ versioned_base_url: Final = ensure_api_route(get_volcengine_base_url(api_base), "api/v3")
+ return join_api_base(versioned_base_url, "embeddings")
def map_openai_params(
self,
diff --git a/litellm/llms/volcengine/responses/transformation.py b/litellm/llms/volcengine/responses/transformation.py
index 6e0aa88b6b6..74f6182bd89 100644
--- a/litellm/llms/volcengine/responses/transformation.py
+++ b/litellm/llms/volcengine/responses/transformation.py
@@ -11,7 +11,7 @@ from litellm.litellm_core_utils.core_helpers import process_response_headers
from litellm.litellm_core_utils.llm_response_utils.convert_dict_to_response import (
safe_convert_created_field,
)
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import (
@@ -137,8 +137,8 @@ class VolcEngineResponsesAPIConfig(OpenAIResponsesAPIConfig):
if base_url.endswith("/responses"):
return base_url
if base_url.endswith("/api/v3"):
- return f"{base_url}/responses"
- return f"{base_url}/api/v3/responses"
+ return join_api_base(base_url, "responses")
+ return join_api_base(base_url, "api/v3/responses")
def map_openai_params(
self,
@@ -345,7 +345,7 @@ class VolcEngineResponsesAPIConfig(OpenAIResponsesAPIConfig):
order: Literal["asc", "desc"] = "desc",
) -> tuple[str, dict]:
encoded_response_id: Final = encode_url_path_segment(response_id, field_name="response_id")
- url: Final = f"{api_base}/{encoded_response_id}/input_items"
+ url: Final = join_api_base(api_base, f"{encoded_response_id}/input_items")
params: Final[dict[str, str | int]] = {}
if after is not None:
params["after"] = after
@@ -380,7 +380,7 @@ class VolcEngineResponsesAPIConfig(OpenAIResponsesAPIConfig):
headers: dict,
) -> tuple[str, dict]:
encoded_response_id: Final = encode_url_path_segment(response_id, field_name="response_id")
- url: Final = f"{api_base}/{encoded_response_id}/cancel"
+ url: Final = join_api_base(api_base, f"{encoded_response_id}/cancel")
data: Final[dict] = {}
return url, data
diff --git a/litellm/llms/voyage/embedding/transformation.py b/litellm/llms/voyage/embedding/transformation.py
index 7e8425bb3d0..1ba66e5ce65 100644
--- a/litellm/llms/voyage/embedding/transformation.py
+++ b/litellm/llms/voyage/embedding/transformation.py
@@ -5,6 +5,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
@@ -50,12 +51,7 @@ class VoyageEmbeddingConfig(BaseEmbeddingConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- if api_base:
- clean_api_base: Final = api_base.rstrip("/")
- if not clean_api_base.endswith("/embeddings"):
- return f"{clean_api_base}/embeddings"
- return clean_api_base
- return f"{get_default_base_url(api_key)}/embeddings"
+ return ensure_api_route(api_base or get_default_base_url(api_key), "embeddings")
def get_supported_openai_params(self, model: str) -> list:
return [
diff --git a/litellm/llms/voyage/embedding/transformation_contextual.py b/litellm/llms/voyage/embedding/transformation_contextual.py
index 8e3802ccbc9..62c03139fbe 100644
--- a/litellm/llms/voyage/embedding/transformation_contextual.py
+++ b/litellm/llms/voyage/embedding/transformation_contextual.py
@@ -10,6 +10,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
@@ -58,12 +59,7 @@ class VoyageContextualEmbeddingConfig(BaseEmbeddingConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- if api_base:
- clean_api_base: Final = api_base.rstrip("/")
- if not clean_api_base.endswith("/contextualizedembeddings"):
- return f"{clean_api_base}/contextualizedembeddings"
- return clean_api_base
- return f"{get_default_base_url(api_key)}/contextualizedembeddings"
+ return ensure_api_route(api_base or get_default_base_url(api_key), "contextualizedembeddings")
def get_supported_openai_params(self, model: str) -> list: # mutable-ok: base class signature
return ["encoding_format", "dimensions"]
diff --git a/litellm/llms/voyage/embedding/transformation_multimodal.py b/litellm/llms/voyage/embedding/transformation_multimodal.py
index 29fdb9ecfe1..c04a2e1f26b 100644
--- a/litellm/llms/voyage/embedding/transformation_multimodal.py
+++ b/litellm/llms/voyage/embedding/transformation_multimodal.py
@@ -13,6 +13,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
@@ -59,12 +60,7 @@ class VoyageMultimodalEmbeddingConfig(BaseEmbeddingConfig):
litellm_params: dict,
stream: bool | None = None,
) -> str:
- if api_base:
- clean_api_base: Final = api_base.rstrip("/")
- if not clean_api_base.endswith("/multimodalembeddings"):
- return f"{clean_api_base}/multimodalembeddings"
- return clean_api_base
- return f"{get_default_base_url(api_key)}/multimodalembeddings"
+ return ensure_api_route(api_base or get_default_base_url(api_key), "multimodalembeddings")
def get_supported_openai_params(self, model: str) -> list:
return ["dimensions"]
diff --git a/litellm/llms/voyage/rerank/transformation.py b/litellm/llms/voyage/rerank/transformation.py
index 121c18e82ae..7fe643870d5 100644
--- a/litellm/llms/voyage/rerank/transformation.py
+++ b/litellm/llms/voyage/rerank/transformation.py
@@ -11,6 +11,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from litellm._uuid import uuid
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import LiteLLMLoggingObj
from litellm.llms.base_llm.rerank.transformation import BaseRerankConfig
from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
@@ -80,9 +81,9 @@ class VoyageRerankConfig(BaseRerankConfig):
api_base = api_base.rstrip("/")
if not api_base.endswith("/v1/rerank"):
if api_base.endswith("/v1"):
- api_base = f"{api_base}/rerank"
+ api_base = join_api_base(api_base, "rerank")
else:
- api_base = f"{api_base}/v1/rerank"
+ api_base = join_api_base(api_base, "v1/rerank")
return api_base
def transform_rerank_request(
diff --git a/litellm/llms/watsonx/audio_transcription/transformation.py b/litellm/llms/watsonx/audio_transcription/transformation.py
index 77ef03b8e10..e5da7dd412f 100644
--- a/litellm/llms/watsonx/audio_transcription/transformation.py
+++ b/litellm/llms/watsonx/audio_transcription/transformation.py
@@ -12,6 +12,7 @@ from pydantic import ConfigDict, TypeAdapter
import litellm
from litellm.litellm_core_utils.audio_utils.utils import process_audio_file
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.types.llms.openai import (
AllMessageValues,
OpenAIAudioTranscriptionOptionalParams,
@@ -150,10 +151,9 @@ class IBMWatsonXAudioTranscriptionConfig(IBMWatsonXMixin, OpenAIWhisperAudioTran
"""
# Get base URL
url = self._get_base_url(api_base=api_base)
- url = url.rstrip("/")
# Add the audio transcription endpoint
- url = f"{url}/ml/v1/audio/transcriptions"
+ url = join_api_base(url, "ml/v1/audio/transcriptions")
# Add version parameter (only version in query string, not project_id)
api_version: Final = optional_params.get("api_version", None) or litellm.WATSONX_DEFAULT_API_VERSION
diff --git a/litellm/llms/watsonx/chat/transformation.py b/litellm/llms/watsonx/chat/transformation.py
index 4ff2212711d..071ff7108b9 100644
--- a/litellm/llms/watsonx/chat/transformation.py
+++ b/litellm/llms/watsonx/chat/transformation.py
@@ -7,6 +7,7 @@ Docs: https://cloud.ibm.com/apidocs/watsonx-ai#text-chat
from typing import Final
from litellm import verbose_logger
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.watsonx import (
WatsonXAIEndpoint,
@@ -96,7 +97,7 @@ class IBMWatsonXChatConfig(IBMWatsonXMixin, OpenAIGPTConfig):
endpoint = endpoint.format(deployment_id=deployment_id)
else:
endpoint = WatsonXAIEndpoint.CHAT_STREAM.value if stream else WatsonXAIEndpoint.CHAT.value
- url = url.rstrip("/") + endpoint
+ url = join_api_base(url, endpoint)
## add api version
url = self._add_api_version_to_url(url=url, api_version=optional_params.pop("api_version", None))
diff --git a/litellm/llms/watsonx/completion/transformation.py b/litellm/llms/watsonx/completion/transformation.py
index 5c0416bd6ca..9ea062ea2f8 100644
--- a/litellm/llms/watsonx/completion/transformation.py
+++ b/litellm/llms/watsonx/completion/transformation.py
@@ -5,6 +5,7 @@ from typing import TYPE_CHECKING, Any, Final
import httpx
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.base_model_iterator import BaseModelResponseIterator
from litellm.types.llms.openai import AllMessageValues, ChatCompletionUsageBlock
from litellm.types.llms.watsonx import WatsonXAIEndpoint
@@ -342,7 +343,7 @@ class IBMWatsonXAIConfig(IBMWatsonXMixin, BaseConfig):
endpoint = endpoint.format(deployment_id=deployment_id)
else:
endpoint = WatsonXAIEndpoint.TEXT_GENERATION_STREAM if stream else WatsonXAIEndpoint.TEXT_GENERATION
- url = url.rstrip("/") + endpoint
+ url = join_api_base(url, endpoint)
## add api version
url = self._add_api_version_to_url(url=url, api_version=optional_params.pop("api_version", None))
diff --git a/litellm/llms/watsonx/embed/transformation.py b/litellm/llms/watsonx/embed/transformation.py
index 1b870c7f639..6ec3b50bbe9 100644
--- a/litellm/llms/watsonx/embed/transformation.py
+++ b/litellm/llms/watsonx/embed/transformation.py
@@ -9,6 +9,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
from typing_extensions import ReadOnly, TypedDict
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.embedding.transformation import (
BaseEmbeddingConfig,
LiteLLMLoggingObj,
@@ -92,7 +93,7 @@ class IBMWatsonXEmbeddingConfig(IBMWatsonXMixin, BaseEmbeddingConfig):
if model.startswith("deployment/"):
deployment_id: Final = "/".join(model.split("/")[1:])
endpoint = endpoint.format(deployment_id=deployment_id)
- url = url.rstrip("/") + endpoint
+ url = join_api_base(url, endpoint)
## add api version
url = self._add_api_version_to_url(url=url, api_version=optional_params.pop("api_version", None))
diff --git a/litellm/llms/watsonx/rerank/transformation.py b/litellm/llms/watsonx/rerank/transformation.py
index 5f337603763..591ddda3cbd 100644
--- a/litellm/llms/watsonx/rerank/transformation.py
+++ b/litellm/llms/watsonx/rerank/transformation.py
@@ -11,6 +11,7 @@ from typing import Final, cast
import httpx
from pydantic import ConfigDict, TypeAdapter
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import LiteLLMLoggingObj
from litellm.llms.base_llm.rerank.transformation import BaseRerankConfig
from litellm.secret_managers.main import get_secret_str
@@ -45,7 +46,7 @@ class IBMWatsonXRerankConfig(IBMWatsonXMixin, BaseRerankConfig):
base_url: Final = self._get_base_url(api_base=api_base)
endpoint: Final = WatsonXAIEndpoint.RERANK.value
- url: Final = base_url.rstrip("/") + endpoint
+ url: Final = join_api_base(base_url, endpoint)
params: Final = optional_params or {}
diff --git a/litellm/llms/xai/audio_transcription/transformation.py b/litellm/llms/xai/audio_transcription/transformation.py
index e34034b5ae9..e29c3d9334a 100644
--- a/litellm/llms/xai/audio_transcription/transformation.py
+++ b/litellm/llms/xai/audio_transcription/transformation.py
@@ -10,6 +10,7 @@ from pydantic import ConfigDict, TypeAdapter, ValidationError
import litellm
from litellm.litellm_core_utils.audio_utils.utils import process_audio_file
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.types.llms.base import LiteLLMBaseModel
from litellm.types.llms.openai import (
@@ -185,7 +186,7 @@ class XAIAudioTranscriptionConfig(BaseAudioTranscriptionConfig):
) -> str:
base: Final = (XAIModelInfo.get_api_base(api_base) or "").rstrip("/")
normalized: Final = base.removesuffix("/v1")
- return f"{normalized}/v1/stt"
+ return join_api_base(normalized, "v1/stt")
def validate_environment(
self,
diff --git a/litellm/llms/xai/batches/transformation.py b/litellm/llms/xai/batches/transformation.py
index fdd9939526e..1e803057e3d 100644
--- a/litellm/llms/xai/batches/transformation.py
+++ b/litellm/llms/xai/batches/transformation.py
@@ -20,7 +20,7 @@ from pydantic import ConfigDict
from typing_extensions import NotRequired, ReadOnly, TypedDict
from litellm.constants import XAI_API_BASE
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.xai.common_utils import XAIModelInfo
from litellm.secret_managers.main import get_secret_str
@@ -76,10 +76,11 @@ def get_xai_auth_headers(
def xai_batches_url(api_base: str | None, batch_id: str | None = None, suffix: str = "") -> str:
- base: Final = f"{get_xai_api_base(api_base)}/v1/batches"
+ root: Final = get_xai_api_base(api_base)
if batch_id is None:
- return base
- return f"{base}/{encode_url_path_segment(batch_id, field_name='batch_id')}{suffix}"
+ return join_api_base(root, "v1/batches")
+ encoded_batch_id: Final = encode_url_path_segment(batch_id, field_name="batch_id")
+ return join_api_base(root, f"v1/batches/{encoded_batch_id}{suffix}")
def is_xai_batch_results_id(file_id: str) -> bool:
diff --git a/litellm/llms/xai/common_utils.py b/litellm/llms/xai/common_utils.py
index 4843caeac59..06e3f32a767 100644
--- a/litellm/llms/xai/common_utils.py
+++ b/litellm/llms/xai/common_utils.py
@@ -5,6 +5,7 @@ import httpx
from pydantic import ConfigDict, TypeAdapter
import litellm
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.base_utils import BaseLLMModelInfo
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import AllMessageValues
@@ -91,7 +92,7 @@ class XAIModelInfo(BaseLLMModelInfo):
"XAI API base or key is not set. Set XAI_API_BASE and provide an xAI API key via api_key, litellm.xai_key, or XAI_API_KEY."
)
response: Final = litellm.module_level_client.get(
- url=f"{api_base}/v1/models",
+ url=join_api_base(api_base, "v1/models"),
headers={"Authorization": f"Bearer {api_key}"},
)
diff --git a/litellm/llms/xai/files/transformation.py b/litellm/llms/xai/files/transformation.py
index 33baca79005..88f2fc51983 100644
--- a/litellm/llms/xai/files/transformation.py
+++ b/litellm/llms/xai/files/transformation.py
@@ -14,7 +14,7 @@ from pydantic import ConfigDict
from typing_extensions import ReadOnly, TypedDict
from litellm.litellm_core_utils.prompt_templates.common_utils import extract_file_data
-from litellm.litellm_core_utils.url_utils import encode_url_path_segment
+from litellm.litellm_core_utils.url_utils import encode_url_path_segment, join_api_base
from litellm.llms.base_llm.chat.transformation import BaseLLMException
from litellm.llms.base_llm.files.transformation import BaseFilesConfig, LiteLLMLoggingObj
from litellm.types.llms.base import LiteLLMBaseModel
@@ -101,11 +101,11 @@ class XAIFilesConfig(BaseFilesConfig):
litellm_params: Mapping[str, object],
stream: bool | None = None,
) -> str:
- return f"{get_xai_api_base(api_base)}/v1/files"
+ return join_api_base(get_xai_api_base(api_base), "v1/files")
def _file_url(self, file_id: str, litellm_params: Mapping[str, object], suffix: str = "") -> str:
encoded_file_id: Final = encode_url_path_segment(file_id, field_name="file_id")
- return f"{_api_base_from(litellm_params)}/v1/files/{encoded_file_id}{suffix}"
+ return join_api_base(_api_base_from(litellm_params), f"v1/files/{encoded_file_id}{suffix}")
def get_error_class(
self, error_message: str, status_code: int, headers: Mapping[str, str] | httpx.Headers
@@ -204,7 +204,7 @@ class XAIFilesConfig(BaseFilesConfig):
optional_params: Mapping[str, object],
litellm_params: Mapping[str, object],
) -> tuple[str, dict[str, str]]: # mutable-ok: BaseFilesConfig signature
- return f"{_api_base_from(litellm_params)}/v1/files", _NO_QUERY_PARAMS
+ return join_api_base(_api_base_from(litellm_params), "v1/files"), _NO_QUERY_PARAMS
def transform_list_files_next_request(
self,
@@ -215,7 +215,7 @@ class XAIFilesConfig(BaseFilesConfig):
page: Final = XAIFileList.model_validate(raw_response.json())
if not page.pagination_token or not page.data:
return None
- return f"{_api_base_from(litellm_params)}/v1/files", {"pagination_token": page.pagination_token}
+ return join_api_base(_api_base_from(litellm_params), "v1/files"), {"pagination_token": page.pagination_token}
def transform_list_files_response(
self,
diff --git a/litellm/llms/xai/responses/transformation.py b/litellm/llms/xai/responses/transformation.py
index 2f638da49c8..b6ea516fb79 100644
--- a/litellm/llms/xai/responses/transformation.py
+++ b/litellm/llms/xai/responses/transformation.py
@@ -9,6 +9,7 @@ import litellm
from litellm._logging import verbose_logger
from litellm.constants import XAI_API_BASE
from litellm.exceptions import AuthenticationError
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.llms.xai.common_utils import XAIModelInfo, xai_reported_cost_in_usd
from litellm.secret_managers.main import get_secret_str
@@ -249,10 +250,7 @@ class XAIResponsesAPIConfig(OpenAIResponsesAPIConfig):
else:
api_base = api_base or litellm.api_base or get_secret_str("XAI_API_BASE") or XAI_API_BASE
- # Remove trailing slashes
- api_base = api_base.rstrip("/")
-
- return f"{api_base}/responses"
+ return join_api_base(api_base, "responses")
def transform_response_api_response(
self,
diff --git a/litellm/llms/you_com/search/transformation.py b/litellm/llms/you_com/search/transformation.py
index 8b2df949a93..9228deda26b 100644
--- a/litellm/llms/you_com/search/transformation.py
+++ b/litellm/llms/you_com/search/transformation.py
@@ -10,6 +10,7 @@ from typing import Final, TypedDict
import httpx
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.base_llm.search.transformation import (
BaseSearchConfig,
SearchResponse,
@@ -109,7 +110,7 @@ class YouComSearchConfig(BaseSearchConfig):
api_base = api_base.rstrip("/")
if not api_base.endswith("/v1/search") and not api_base.endswith("/v1/agents/search"):
- api_base = f"{api_base}/v1/search"
+ api_base = join_api_base(api_base, "v1/search")
return api_base
diff --git a/litellm/main.py b/litellm/main.py
index 079b0bcbd6c..27700cf9def 100644
--- a/litellm/main.py
+++ b/litellm/main.py
@@ -112,6 +112,7 @@ from litellm.litellm_core_utils.request_timeout_resolver import (
get_configured_request_timeout,
)
from litellm.litellm_core_utils.tokenizer import Encoding as Tokenizer
+from litellm.litellm_core_utils.url_utils import ensure_api_route
from litellm.llms.base_llm import BaseConfig, BaseImageGenerationConfig
from litellm.llms.base_llm.base_model_iterator import (
convert_model_response_to_streaming,
@@ -230,6 +231,7 @@ from .llms.azure.azure import AzureChatCompletion, check_dynamic_azure_params
from .llms.azure.chat.o_series_handler import AzureOpenAIO1ChatCompletion
from .llms.azure.completion.handler import AzureTextCompletion
from .llms.azure_ai.anthropic.handler import AzureAnthropicChatCompletion
+from .llms.azure_ai.anthropic.transformation import azure_anthropic_messages_url
from .llms.azure_ai.embed import AzureAIEmbedding
from .llms.bedrock.chat import BedrockConverseLLM
from .llms.bedrock.chat.converse_transformation import AmazonConverseConfig
@@ -1666,16 +1668,8 @@ def _complete_azure_ai(ctx: CompletionDispatchContext) -> _CompletionDispatchRes
)
api_key = AzureFoundryModelInfo.get_api_key(api_key)
- # Ensure the URL ends with /v1/messages for Anthropic
if api_base:
- api_base = api_base.rstrip("/")
- if not api_base.endswith("/v1/messages"):
- if "/anthropic" in api_base:
- parts: Final = api_base.split("/anthropic", 1)
- api_base = parts[0] + "/anthropic"
- else:
- api_base = api_base + "/anthropic"
- api_base = api_base + "/v1/messages"
+ api_base = azure_anthropic_messages_url(api_base)
response = azure_anthropic_chat_completions.completion(
model=model,
@@ -2895,8 +2889,8 @@ def _complete_anthropic_text(
# Check if we should disable automatic URL suffix appending
disable_url_suffix: Final = get_secret_bool("LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX")
- if api_base is not None and not disable_url_suffix and not api_base.endswith("/v1/complete"):
- api_base += "/v1/complete"
+ if api_base is not None and not disable_url_suffix:
+ api_base = ensure_api_route(api_base, "v1/complete")
elif disable_url_suffix:
verbose_logger.debug("LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX is set, skipping /v1/complete suffix")
@@ -2951,8 +2945,8 @@ def _complete_anthropic(ctx: CompletionDispatchContext) -> _CompletionDispatchRe
# Check if we should disable automatic URL suffix appending
disable_url_suffix: Final = get_secret_bool("LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX")
- if api_base is not None and not disable_url_suffix and not api_base.endswith("/v1/messages"):
- api_base += "/v1/messages"
+ if api_base is not None and not disable_url_suffix:
+ api_base = ensure_api_route(api_base, "v1/messages")
elif disable_url_suffix:
verbose_logger.debug("LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX is set, skipping /v1/messages suffix")
@@ -4090,9 +4084,7 @@ def _complete_text_completion_inception(
api_base = passed_api_base or get_secret_str("INCEPTION_API_BASE") or "https://api.inceptionlabs.ai/v1"
# FIM is served at `/v1/fim/completions`; the OpenAI client appends
# `/completions`, so point it at the `/v1/fim` base.
- api_base = api_base.rstrip("/")
- if not api_base.endswith("/fim"):
- api_base += "/fim"
+ api_base = ensure_api_route(api_base, "fim")
# Don't forward the server-managed Inception key to a caller-supplied
# api_base; only resolve it for the default/server base, or when the
diff --git a/litellm/rag/ingestion/gemini_ingestion.py b/litellm/rag/ingestion/gemini_ingestion.py
index 4ee16685e03..6320e477bf2 100644
--- a/litellm/rag/ingestion/gemini_ingestion.py
+++ b/litellm/rag/ingestion/gemini_ingestion.py
@@ -12,6 +12,7 @@ from typing import TYPE_CHECKING, Final, cast
from pydantic import ConfigDict
from litellm._logging import verbose_logger
+from litellm.litellm_core_utils.url_utils import join_api_base
from litellm.llms.custom_httpx.http_handler import (
get_async_httpx_client,
header_value,
@@ -107,7 +108,7 @@ class GeminiRAGIngestion(BaseRAGIngestion):
raise ValueError("GEMINI_API_BASE is required")
api_version: Final = "v1beta"
- base_url: Final = f"{api_base}/{api_version}"
+ base_url: Final = join_api_base(api_base, api_version)
# Create File Search store if not provided
if not vector_store_id:
@@ -148,7 +149,7 @@ class GeminiRAGIngestion(BaseRAGIngestion):
Returns:
Store name (format: fileSearchStores/xxxxxxx)
"""
- url: Final = f"{base_url}/fileSearchStores"
+ url: Final = join_api_base(base_url, "fileSearchStores")
request_body: Final = {"displayName": display_name}
@@ -236,7 +237,7 @@ class GeminiRAGIngestion(BaseRAGIngestion):
# base_url is like: https://generativelanguage.googleapis.com/v1beta
# We need: https://generativelanguage.googleapis.com/upload/v1beta/{store_id}:uploadToFileSearchStore
api_base: Final = base_url.replace("/v1beta", "") # Get base without version
- url: Final = f"{api_base}/upload/v1beta/{vector_store_id}:uploadToFileSearchStore"
+ url: Final = join_api_base(api_base, f"upload/v1beta/{vector_store_id}:uploadToFileSearchStore")
# Build request body with chunking config and metadata if provided
request_body: Final[dict[str, object]] = {"displayName": filename}
diff --git a/tests/integration/providers/_api_base_url_join_support.py b/tests/integration/providers/_api_base_url_join_support.py
new file mode 100644
index 00000000000..44bd6ec4dca
--- /dev/null
+++ b/tests/integration/providers/_api_base_url_join_support.py
@@ -0,0 +1,163 @@
+import json
+from collections.abc import Callable, Mapping
+from types import MappingProxyType
+from typing import Final
+from urllib.parse import urlsplit
+
+from integration._support.anthropic_sse import message_stream
+from integration._support.openai_wire import chat_reply
+from integration._support.upstream import aws_event_stream_frame
+from integration._support.wire import Reply, Request
+from integration.providers._api_base_trailing_slash_support import VECTOR, marker_of, reply_text, request_json
+from pydantic import JsonValue
+
+COUNTED_TOKENS: Final = 7
+RELEVANCE: Final = 0.75
+_USAGE: Final = {"input_tokens": 3, "output_tokens": 2}
+
+
+def _json(payload: Mapping[str, JsonValue]) -> Reply:
+ return Reply(body=json.dumps(payload).encode())
+
+
+def _path(request: Request) -> str:
+ return urlsplit(request.target).path
+
+
+def _wants_stream(request: Request) -> bool:
+ return request_json(request).get("stream") is True
+
+
+def anthropic_message(marker: str) -> dict[str, JsonValue]:
+ return {
+ "id": f"msg_{marker}",
+ "type": "message",
+ "role": "assistant",
+ "model": "scripted",
+ "content": [{"type": "text", "text": reply_text(marker)}],
+ "stop_reason": "end_turn",
+ "stop_sequence": None,
+ "usage": dict(_USAGE),
+ }
+
+
+def _anthropic(request: Request) -> Reply:
+ if _path(request).endswith("/count_tokens"):
+ return _json({"input_tokens": COUNTED_TOKENS})
+ marker: Final = marker_of(request)
+ if _wants_stream(request):
+ return Reply(
+ content_type="text/event-stream", chunks=message_stream(f"msg_{marker}", "scripted", reply_text(marker))
+ )
+ return _json(anthropic_message(marker))
+
+
+def _gemini_candidate(text: str, finished: bool) -> dict[str, JsonValue]:
+ finish: Final[dict[str, JsonValue]] = {"finishReason": "STOP"} if finished else {}
+ return {"content": {"role": "model", "parts": [{"text": text}]}, "index": 0, **finish}
+
+
+def _gemini_frame(text: str, finished: bool) -> bytes:
+ usage: Final = {"promptTokenCount": 3, "candidatesTokenCount": 2, "totalTokenCount": 5}
+ frame: Final = {"candidates": [_gemini_candidate(text, finished)], "usageMetadata": usage}
+ return b"data: " + json.dumps(frame).encode() + b"\r\n\r\n"
+
+
+def _gemini(request: Request) -> Reply:
+ path: Final = _path(request)
+ marker: Final = marker_of(request)
+ if path.endswith(":batchEmbedContents"):
+ return _json({"embeddings": [{"values": list(VECTOR)}]})
+ if path.endswith(":streamGenerateContent"):
+ return Reply(
+ content_type="text/event-stream",
+ chunks=(_gemini_frame("reply ", False), _gemini_frame(marker, True)),
+ )
+ return _json(
+ {
+ "candidates": [_gemini_candidate(reply_text(marker), True)],
+ "usageMetadata": {"promptTokenCount": 3, "candidatesTokenCount": 2, "totalTokenCount": 5},
+ }
+ )
+
+
+def _embedding_list(model: str) -> Reply:
+ return _json(
+ {
+ "object": "list",
+ "data": [{"object": "embedding", "index": 0, "embedding": list(VECTOR)}],
+ "model": model,
+ "usage": {"prompt_tokens": 2, "total_tokens": 2},
+ }
+ )
+
+
+def _openai_like(model: str) -> Callable[[Request], Reply]:
+ def respond(request: Request) -> Reply:
+ path: Final = _path(request)
+ marker: Final = marker_of(request)
+ if path.endswith("/embeddings"):
+ return _embedding_list(model)
+ if path.endswith("/audio/transcriptions"):
+ return _json({"text": reply_text(marker), "model": model})
+ return chat_reply(f"chatcmpl-{marker}", model, reply_text(marker), stream=_wants_stream(request))
+
+ return respond
+
+
+def _converse_event(kind: str, payload: Mapping[str, JsonValue]) -> bytes:
+ headers: Final = {":event-type": kind, ":content-type": "application/json", ":message-type": "event"}
+ return aws_event_stream_frame(headers, json.dumps(payload, separators=(",", ":")).encode())
+
+
+def _bedrock_converse(request: Request) -> Reply:
+ marker: Final = marker_of(request)
+ usage: Final[dict[str, JsonValue]] = {"inputTokens": 3, "outputTokens": 2, "totalTokens": 5}
+ if _path(request).endswith("/converse-stream"):
+ events: Final[tuple[tuple[str, Mapping[str, JsonValue]], ...]] = (
+ ("messageStart", {"role": "assistant"}),
+ ("contentBlockDelta", {"delta": {"text": "reply "}, "contentBlockIndex": 0}),
+ ("contentBlockDelta", {"delta": {"text": marker}, "contentBlockIndex": 0}),
+ ("contentBlockStop", {"contentBlockIndex": 0}),
+ ("messageStop", {"stopReason": "end_turn"}),
+ ("metadata", {"usage": usage, "metrics": {"latencyMs": 1}}),
+ )
+ return Reply(
+ content_type="application/vnd.amazon.eventstream",
+ chunks=tuple(_converse_event(kind, payload) for kind, payload in events),
+ )
+ return _json(
+ {
+ "output": {"message": {"role": "assistant", "content": [{"text": reply_text(marker)}]}},
+ "stopReason": "end_turn",
+ "usage": usage,
+ "metrics": {"latencyMs": 1},
+ }
+ )
+
+
+def _cohere_rerank(request: Request) -> Reply:
+ return _json(
+ {
+ "id": f"rerank-{marker_of(request)}",
+ "results": [{"index": 0, "relevance_score": RELEVANCE}],
+ "meta": {"billed_units": {"search_units": 1}},
+ }
+ )
+
+
+_BY_PREFIX: Final[Mapping[str, Callable[[Request], Reply]]] = MappingProxyType(
+ {
+ "an": _anthropic,
+ "bi": lambda request: _json(anthropic_message(marker_of(request))),
+ "bc": _bedrock_converse,
+ "gm": _gemini,
+ "mi": _openai_like("mistral-small-latest"),
+ "db": _openai_like("databricks-meta-llama-3-3-70b-instruct"),
+ "co": _cohere_rerank,
+ }
+)
+
+
+def join_peer(request: Request) -> Reply:
+ return _BY_PREFIX[request.target.split("/")[1]](request)
diff --git a/tests/integration/providers/test_api_base_url_join_chaos.py b/tests/integration/providers/test_api_base_url_join_chaos.py
new file mode 100644
index 00000000000..8ebcd165f4d
--- /dev/null
+++ b/tests/integration/providers/test_api_base_url_join_chaos.py
@@ -0,0 +1,176 @@
+import asyncio
+from collections.abc import Awaitable, Callable
+from itertools import chain, repeat
+from typing import Final, TypeAlias
+
+import anthropic
+import httpx
+import openai
+from integration._support.client import Gateway, Scenario, eventually
+from integration._support.database import read_rows
+from integration._support.wire import Reply, Request, wire_server
+from integration.providers._api_base_trailing_slash_support import (
+ KEY,
+ VECTOR,
+ assert_control_answers,
+ assert_one_request,
+ body_of,
+ marker_of,
+ new_marker,
+ of_marker,
+ reply_text,
+)
+from integration.providers._api_base_url_join_support import join_peer
+from pydantic import JsonValue
+
+_PER_KIND: Final = 3
+_HAIKU: Final = "anthropic.claude-haiku-5-5"
+_GENERATE: Final = "/gm/v1beta/models/gemini-3.8-flash:generateContent"
+_STREAM_GENERATE: Final = "/gm/v1beta/models/gemini-3.8-flash:streamGenerateContent?alt=sse"
+
+Answer: TypeAlias = tuple[str | None, str | None]
+Call: TypeAlias = Callable[[str], Awaitable[Answer]]
+
+
+def _deploy(scenario: Scenario, model: str, api_base: str) -> str:
+ if model.startswith("bedrock/"):
+ return scenario.model(model=model, api_base=api_base, api_key=KEY, aws_region_name="us-east-1")
+ return scenario.model(model=model, api_base=api_base, api_key=KEY)
+
+
+def _spend_api_bases(identities: tuple[str, ...]) -> dict[str, JsonValue]:
+ rows: Final = eventually(
+ lambda: read_rows(
+ "SELECT request_id, api_base FROM \"LiteLLM_SpendLogs\" WHERE request_id = ANY(string_to_array(%s, ','))",
+ (",".join(identities),),
+ ),
+ lambda found: len(found) == len(identities),
+ seconds=70,
+ )
+ return {str(row["request_id"]): row["api_base"] for row in rows}
+
+
+def test_mixed_burst_on_slash_bases_lands_each_call_once_on_its_single_slash_url(gateway: Gateway) -> None:
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ claude: Final = _deploy(scenario, "anthropic/claude-haiku-5-5", f"{wire.url}/an/")
+ gemini: Final = _deploy(scenario, "gemini/gemini-3.8-flash", f"{wire.url}/gm/v1beta/")
+ mistral: Final = _deploy(scenario, "mistral/mistral-small-latest", f"{wire.url}/mi/v1/")
+ converse: Final = _deploy(scenario, f"bedrock/converse/{_HAIKU}", f"{wire.url}/bc/")
+ databricks: Final = _deploy(
+ scenario, "databricks/databricks-meta-llama-3-3-70b-instruct", f"{wire.url}/db/serving-endpoints/"
+ )
+ embed_model: Final = _deploy(scenario, "mistral/mistral-embed", f"{wire.url}/mi/v1/")
+
+ def openai_client() -> openai.AsyncOpenAI:
+ return openai.AsyncOpenAI(base_url=str(gateway.client.base_url) + "/v1", api_key=gateway.key, max_retries=0)
+
+ def chat(model: str) -> Call:
+ async def call(marker: str) -> Answer:
+ async with openai_client() as client:
+ completion: Final = await client.chat.completions.create(
+ model=model, messages=[{"role": "user", "content": marker}]
+ )
+ return completion.choices[0].message.content, completion.id
+
+ return call
+
+ def streamed(model: str) -> Call:
+ async def call(marker: str) -> Answer:
+ async with openai_client() as client:
+ stream: Final = await client.chat.completions.create(
+ model=model, messages=[{"role": "user", "content": marker}], stream=True
+ )
+ parts: Final = [chunk.choices[0].delta.content or "" async for chunk in stream if chunk.choices]
+ return "".join(parts), None
+
+ return call
+
+ async def messages(marker: str) -> Answer:
+ async with anthropic.AsyncAnthropic(
+ base_url=str(gateway.client.base_url), api_key=gateway.key, max_retries=0
+ ) as client:
+ message: Final = await client.messages.create(
+ model=claude, max_tokens=16, messages=[{"role": "user", "content": marker}]
+ )
+ return "".join(block.text for block in message.content if block.type == "text"), None
+
+ async def embed(marker: str) -> Answer:
+ async with httpx.AsyncClient(base_url=str(gateway.client.base_url), timeout=60, trust_env=False) as client:
+ response: Final = await client.post(
+ "/v1/embeddings",
+ json={"model": embed_model, "input": marker},
+ headers={"Authorization": f"Bearer {gateway.key}"},
+ )
+ data: Final = body_of(response)["data"]
+ vector_matches: Final = data == [{"object": "embedding", "index": 0, "embedding": list(VECTOR)}]
+ return (reply_text(marker) if vector_matches else str(data)), None
+
+ kinds: Final[tuple[tuple[str, Call], ...]] = (
+ ("/an/v1/messages", chat(claude)),
+ ("/an/v1/messages", streamed(claude)),
+ ("/an/v1/messages", messages),
+ (_GENERATE, chat(gemini)),
+ (_STREAM_GENERATE, streamed(gemini)),
+ ("/mi/v1/chat/completions", chat(mistral)),
+ ("/mi/v1/chat/completions", streamed(mistral)),
+ (f"/bc/model/{_HAIKU}/converse", chat(converse)),
+ (f"/bc/model/{_HAIKU}/converse-stream", streamed(converse)),
+ ("/db/serving-endpoints/chat/completions", chat(databricks)),
+ ("/mi/v1/embeddings", embed),
+ )
+ repeated_kinds: Final = tuple(chain.from_iterable(repeat(kind, _PER_KIND) for kind in kinds))
+ calls: Final = tuple((target, call, new_marker()) for target, call in repeated_kinds)
+
+ async def burst() -> list[Answer]:
+ return await asyncio.gather(*(call(marker) for _, call, marker in calls))
+
+ answers: Final = asyncio.run(burst())
+ assert [text for text, _ in answers] == [reply_text(marker) for *_, marker in calls]
+ received: Final = wire.drain()
+ assert len(received) == len(calls), sorted(request.target for request in received)
+ for target, _, marker in calls:
+ assert_one_request(of_marker(received, marker), target, marker)
+ logged: Final = {
+ identity: f"{wire.url}{target}"
+ for (target, _, _), (_, identity) in zip(calls, answers, strict=True)
+ if identity is not None
+ }
+ assert _spend_api_bases(tuple(logged)) == logged
+
+
+def test_upstream_dropping_doomed_connections_mid_burst_fails_only_those_calls(gateway: Gateway) -> None:
+ markers: Final = tuple(new_marker() for _ in range(16))
+ doomed: Final = frozenset(markers[::2])
+
+ def respond(request: Request) -> Reply:
+ return Reply(drop_connection=True) if marker_of(request) in doomed else join_peer(request)
+
+ with wire_server(respond) as wire, gateway.scenario() as scenario:
+ claude: Final = _deploy(scenario, "anthropic/claude-haiku-5-5", f"{wire.url}/an/")
+ gemini: Final = _deploy(scenario, "gemini/gemini-3.8-flash", f"{wire.url}/gm/v1beta/")
+ routed: Final = {
+ marker: (claude, "/an/v1/messages") if index % 4 < 2 else (gemini, _GENERATE)
+ for index, marker in enumerate(markers)
+ }
+
+ async def call(marker: str) -> tuple[int, str]:
+ async with httpx.AsyncClient(base_url=str(gateway.client.base_url), timeout=60, trust_env=False) as client:
+ response: Final = await client.post(
+ "/v1/chat/completions",
+ json={"model": routed[marker][0], "messages": [{"role": "user", "content": marker}]},
+ headers={"Authorization": f"Bearer {gateway.key}"},
+ )
+ return response.status_code, response.text
+
+ async def burst() -> list[tuple[int, str]]:
+ return await asyncio.gather(*(call(marker) for marker in markers))
+
+ answers: Final = dict(zip(markers, asyncio.run(burst()), strict=True))
+ received: Final = wire.drain()
+ assert_control_answers(gateway, scenario)
+ for marker in doomed:
+ assert answers[marker][0] >= 500 and "error" in answers[marker][1], answers[marker]
+ assert {request.target for request in of_marker(received, marker)} == {routed[marker][1]}, marker
+ for marker in frozenset(markers) - doomed:
+ assert answers[marker][0] == 200 and reply_text(marker) in answers[marker][1], answers[marker]
+ assert_one_request(of_marker(received, marker), routed[marker][1], marker)
diff --git a/tests/integration/providers/test_api_base_url_join_host_owned_proxy.py b/tests/integration/providers/test_api_base_url_join_host_owned_proxy.py
new file mode 100644
index 00000000000..4a075138305
--- /dev/null
+++ b/tests/integration/providers/test_api_base_url_join_host_owned_proxy.py
@@ -0,0 +1,140 @@
+import os
+from collections.abc import Callable, Iterator
+from pathlib import Path
+from typing import Final
+
+import pytest
+import yaml
+from integration._support.client import JSON_OBJECT, Gateway, gateway_from_environment
+from integration._support.forward_proxy import ForwardProxy, refusing_forward_proxy
+from integration._support.process import graceful_stop_seconds, owned_proxy
+from integration.providers._api_base_trailing_slash_support import KEY, new_marker
+from pydantic import JsonValue
+
+_AZURE_HOST: Final = "anthropic-prod.services.ai.azure.com"
+_EMBEDDINGS: Final = "host-embeddings"
+_RERANK: Final = "host-rerank"
+_AZURE_CLAUDE: Final = "host-azure-claude"
+_AZURE_COUNT: Final = "host-azure-count"
+_EMPTY_BASE: Final = "host-empty-base"
+_SCRUBBED_PREFIXES: Final = ("ANTHROPIC_", "AZURE_AI_", "INFINITY_")
+_PROXY_VARIABLES: Final = frozenset({"HTTPS_PROXY", "HTTP_PROXY", "ALL_PROXY", "NO_PROXY"})
+_load_yaml: Final[Callable[[str], object]] = yaml.safe_load
+
+pytestmark: Final = pytest.mark.timeout(2 * graceful_stop_seconds() + 120)
+
+
+def _deployment(name: str, model: str, api_base: str) -> dict[str, JsonValue]:
+ return {"model_name": name, "litellm_params": {"model": model, "api_base": api_base, "api_key": KEY}}
+
+
+def _config(directory: Path) -> Path:
+ base: Final = JSON_OBJECT.validate_python(_load_yaml(Path("tests/integration/proxy_config.yaml").read_text()))
+ configuration: Final = {
+ **base,
+ "model_list": [
+ _deployment(_EMBEDDINGS, "infinity/BAAI/bge-small-en-v1.5", "http://embeddings"),
+ _deployment(_RERANK, "infinity/BAAI/bge-reranker-base", "http://rerank"),
+ _deployment(_AZURE_CLAUDE, "azure_ai/claude-haiku-5-5", f"http://{_AZURE_HOST}/anthropic"),
+ _deployment(_AZURE_COUNT, "azure_ai/claude-haiku-5-5", f"http://{_AZURE_HOST}"),
+ _deployment(_EMPTY_BASE, "anthropic/claude-haiku-5-5", ""),
+ ],
+ "router_settings": {"disable_cooldowns": True, "num_retries": 0},
+ }
+ path: Final = directory / "host-named-like-the-route.yaml"
+ path.write_text(yaml.safe_dump(configuration))
+ return path
+
+
+def _scrubbed() -> tuple[str, ...]:
+ return tuple(
+ name
+ for name in os.environ
+ if name.upper() in _PROXY_VARIABLES or name.upper().startswith(_SCRUBBED_PREFIXES)
+ )
+
+
[email protected](scope="module")
+def forward() -> Iterator[ForwardProxy]:
+ with refusing_forward_proxy() as proxy:
+ yield proxy
+
+
[email protected](scope="module")
+def isolated(forward: ForwardProxy, tmp_path_factory: pytest.TempPathFactory) -> Iterator[Gateway]:
+ directory: Final = tmp_path_factory.mktemp("url-join-host")
+ with gateway_from_environment() as shared, owned_proxy(
+ shared,
+ directory,
+ {
+ "HTTP_PROXY": forward.url,
+ "HTTPS_PROXY": forward.url,
+ "NO_PROXY": "127.0.0.1,localhost",
+ "PROXY_BATCH_POLLING_ENABLED": "false",
+ },
+ config=_config(directory),
+ remove_environment=_scrubbed(),
+ ) as candidate:
+ yield candidate
+
+
+def _targets_after(forward: ForwardProxy, call: Callable[[], object]) -> frozenset[str]:
+ forward.drain()
+ call()
+ targets: Final = frozenset(forward.targets())
+ assert targets, "the forward proxy saw no outbound request"
+ return targets
+
+
+def _chat(isolated: Gateway, model: str) -> object:
+ return isolated.request(
+ "POST", "/v1/chat/completions", {"model": model, "messages": [{"role": "user", "content": new_marker()}]}
+ )
+
+
+def _messages(isolated: Gateway, model: str, path: str) -> object:
+ return isolated.request(
+ "POST",
+ path,
+ {"model": model, "max_tokens": 16, "messages": [{"role": "user", "content": new_marker()}]},
+ headers={"anthropic-version": "2023-06-01"},
+ )
+
+
+def test_embedding_host_named_embeddings_gets_the_route_appended(isolated: Gateway, forward: ForwardProxy) -> None:
+ targets: Final = _targets_after(
+ forward, lambda: isolated.request("POST", "/v1/embeddings", {"model": _EMBEDDINGS, "input": new_marker()})
+ )
+ assert targets == {"http://embeddings/embeddings"}
+
+
+def test_rerank_host_named_rerank_gets_the_route_appended(isolated: Gateway, forward: ForwardProxy) -> None:
+ body: Final[dict[str, JsonValue]] = {"model": _RERANK, "query": new_marker(), "documents": ["only document"]}
+ targets: Final = _targets_after(forward, lambda: isolated.request("POST", "/v1/rerank", body))
+ assert targets == {"http://rerank/rerank"}
+
+
+def test_azure_claude_chat_on_a_host_starting_with_anthropic_keeps_the_host(
+ isolated: Gateway, forward: ForwardProxy
+) -> None:
+ targets: Final = _targets_after(forward, lambda: _chat(isolated, _AZURE_CLAUDE))
+ assert targets == {f"http://{_AZURE_HOST}/anthropic/v1/messages"}
+
+
+def test_azure_claude_messages_on_a_host_starting_with_anthropic_keeps_the_host(
+ isolated: Gateway, forward: ForwardProxy
+) -> None:
+ targets: Final = _targets_after(forward, lambda: _messages(isolated, _AZURE_CLAUDE, "/v1/messages"))
+ assert targets == {f"http://{_AZURE_HOST}/anthropic/v1/messages"}
+
+
+def test_azure_claude_count_tokens_on_a_bare_host_starting_with_anthropic_adds_the_anthropic_path(
+ isolated: Gateway, forward: ForwardProxy
+) -> None:
+ targets: Final = _targets_after(forward, lambda: _messages(isolated, _AZURE_COUNT, "/v1/messages/count_tokens"))
+ assert targets == {f"http://{_AZURE_HOST}/anthropic/v1/messages/count_tokens"}
+
+
+def test_empty_api_base_falls_back_to_the_provider_default_host(isolated: Gateway, forward: ForwardProxy) -> None:
+ targets: Final = _targets_after(forward, lambda: _chat(isolated, _EMPTY_BASE))
+ assert targets == {"api.anthropic.com:443"}
diff --git a/tests/integration/providers/test_api_base_url_join_wire.py b/tests/integration/providers/test_api_base_url_join_wire.py
new file mode 100644
index 00000000000..42e53ff9f95
--- /dev/null
+++ b/tests/integration/providers/test_api_base_url_join_wire.py
@@ -0,0 +1,373 @@
+import json
+from typing import Final
+from urllib.parse import urlsplit
+
+import anthropic
+import openai
+import pytest
+from integration._support.client import Gateway, Scenario, eventually, list_value, object_value, string_value
+from integration._support.database import read_rows
+from integration._support.wire import Reply, Request, wire_server
+from integration.providers._api_base_trailing_slash_support import (
+ KEY,
+ TRAILING,
+ TRAILING_IDS,
+ VECTOR,
+ Row,
+ assert_control_answers,
+ assert_one_request,
+ body_of,
+ first_item,
+ new_marker,
+ reply_text,
+ request_json,
+)
+from integration.providers._api_base_url_join_support import COUNTED_TOKENS, RELEVANCE, join_peer
+from pydantic import JsonValue
+
+_HAIKU: Final = "anthropic.claude-haiku-5-5"
+_ANTHROPIC: Final = Row("anthropic/claude-haiku-5-5", "/an", "/v1/messages")
+_GEMINI: Final = Row("gemini/gemini-3.8-flash", "/gm/v1beta", "/models/gemini-3.8-flash:generateContent")
+_MISTRAL: Final = Row("mistral/mistral-small-latest", "/mi/v1", "/chat/completions")
+_BEDROCK_INVOKE: Final = Row(f"bedrock/invoke/{_HAIKU}", "/bi", f"/model/{_HAIKU}/invoke")
+_CHAT_ROWS: Final = (
+ _ANTHROPIC,
+ _GEMINI,
+ _MISTRAL,
+ Row("databricks/databricks-meta-llama-3-3-70b-instruct", "/db/serving-endpoints", "/chat/completions"),
+ _BEDROCK_INVOKE,
+ Row(f"bedrock/converse/{_HAIKU}", "/bc", f"/model/{_HAIKU}/converse"),
+)
+_STREAM_ROWS: Final = (
+ _ANTHROPIC,
+ Row("gemini/gemini-3.8-flash", "/gm/v1beta", "/models/gemini-3.8-flash:streamGenerateContent?alt=sse"),
+ _MISTRAL,
+ Row(f"bedrock/converse/{_HAIKU}", "/bc", f"/model/{_HAIKU}/converse-stream"),
+)
+_MESSAGES_ROWS: Final = (_ANTHROPIC, _BEDROCK_INVOKE, _GEMINI, _MISTRAL)
+_RESPONSES_ROWS: Final = (_ANTHROPIC, _GEMINI)
+_EMBEDDING_ROWS: Final = (
+ Row("mistral/mistral-embed", "/mi/v1", "/embeddings"),
+ Row("gemini/gemini-embedding-2", "/gm/v1beta", "/models/gemini-embedding-2:batchEmbedContents"),
+)
+_TRANSCRIPTION: Final = Row("mistral/voxtral-mini-latest", "/mi/v1", "/audio/transcriptions")
+_FULL_ROUTE_SUFFIXES: Final = ("", "/", "?beta=true")
+_FULL_ROUTE_IDS: Final = ("exact", "slash", "query")
+
+
+def _ids(rows: tuple[Row, ...]) -> tuple[str, ...]:
+ return tuple(row.model for row in rows)
+
+
+def _deploy(scenario: Scenario, row: Row, api_base: str) -> str:
+ if row.model.startswith("bedrock/"):
+ return scenario.model(model=row.model, api_base=api_base, api_key=KEY, aws_region_name="us-east-1")
+ return scenario.model(model=row.model, api_base=api_base, api_key=KEY)
+
+
+def _openai(gateway: Gateway) -> openai.OpenAI:
+ return openai.OpenAI(base_url=str(gateway.client.base_url) + "/v1", api_key=gateway.key, max_retries=0)
+
+
+def _anthropic(gateway: Gateway) -> anthropic.Anthropic:
+ return anthropic.Anthropic(base_url=str(gateway.client.base_url), api_key=gateway.key, max_retries=0)
+
+
+def _chat(gateway: Gateway, model: str, marker: str) -> str | None:
+ completion: Final = _openai(gateway).chat.completions.create(
+ model=model, messages=[{"role": "user", "content": marker}]
+ )
+ return completion.choices[0].message.content
+
+
+def _message_text(gateway: Gateway, model: str, marker: str) -> list[str]:
+ message: Final = _anthropic(gateway).messages.create(
+ model=model, max_tokens=16, messages=[{"role": "user", "content": marker}]
+ )
+ return [block.text for block in message.content if block.type == "text"]
+
+
+def _spend_rows(pattern: str) -> list[dict[str, JsonValue]]:
+ return read_rows(
+ 'SELECT request_id, api_base, cache_hit FROM "LiteLLM_SpendLogs" WHERE request_id LIKE %s', (pattern,)
+ )
+
+
+def _one_spend_row(pattern: str) -> dict[str, JsonValue]:
+ (row,) = eventually(lambda: _spend_rows(pattern), lambda rows: len(rows) == 1, seconds=70)
+ return row
+
+
+def _gemini_contents(marker: str) -> dict[str, JsonValue]:
+ return {"contents": [{"role": "user", "parts": [{"text": marker}]}]}
+
+
+def _gemini_text(payload: dict[str, JsonValue]) -> JsonValue:
+ content: Final = object_value(first_item(payload["candidates"])["content"])
+ return first_item(content["parts"])["text"]
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
[email protected]("row", _CHAT_ROWS, ids=_ids(_CHAT_ROWS))
+def test_chat_completion_reaches_one_single_slash_url(gateway: Gateway, row: Row, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, row, f"{wire.url}{row.base}{trailing}")
+ assert _chat(gateway, model, marker) == reply_text(marker)
+ assert_one_request(wire.drain(), row.target, marker)
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
[email protected]("row", _STREAM_ROWS, ids=_ids(_STREAM_ROWS))
+def test_streamed_chat_reaches_one_single_slash_url(gateway: Gateway, row: Row, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, row, f"{wire.url}{row.base}{trailing}")
+ stream: Final = _openai(gateway).chat.completions.create(
+ model=model, messages=[{"role": "user", "content": marker}], stream=True
+ )
+ text: Final = "".join(chunk.choices[0].delta.content or "" for chunk in stream if chunk.choices)
+ assert text == reply_text(marker)
+ assert_one_request(wire.drain(), row.target, marker)
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
[email protected]("row", _MESSAGES_ROWS, ids=_ids(_MESSAGES_ROWS))
+def test_messages_reaches_one_single_slash_url(gateway: Gateway, row: Row, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, row, f"{wire.url}{row.base}{trailing}")
+ assert _message_text(gateway, model, marker) == [reply_text(marker)]
+ assert_one_request(wire.drain(), row.target, marker)
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
+def test_streamed_messages_reach_one_single_slash_url(gateway: Gateway, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _ANTHROPIC, f"{wire.url}/an{trailing}")
+ with _anthropic(gateway).messages.stream(
+ model=model, max_tokens=16, messages=[{"role": "user", "content": marker}]
+ ) as stream:
+ text: Final = "".join(stream.text_stream)
+ assert text == reply_text(marker)
+ assert request_json(assert_one_request(wire.drain(), "/an/v1/messages", marker))["stream"] is True
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
[email protected]("row", _RESPONSES_ROWS, ids=_ids(_RESPONSES_ROWS))
+def test_responses_bridge_reaches_one_single_slash_url(gateway: Gateway, row: Row, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, row, f"{wire.url}{row.base}{trailing}")
+ assert _openai(gateway).responses.create(model=model, input=marker).output_text == reply_text(marker)
+ assert_one_request(wire.drain(), row.target, marker)
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
[email protected]("row", _EMBEDDING_ROWS, ids=_ids(_EMBEDDING_ROWS))
+def test_embedding_reaches_one_single_slash_url(gateway: Gateway, row: Row, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, row, f"{wire.url}{row.base}{trailing}")
+ response: Final = gateway.request("POST", "/v1/embeddings", {"model": model, "input": marker})
+ assert first_item(body_of(response)["data"])["embedding"] == list(VECTOR)
+ assert_one_request(wire.drain(), row.target, marker)
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
+def test_mistral_transcription_reaches_one_single_slash_url(gateway: Gateway, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _TRANSCRIPTION, f"{wire.url}/mi/v1{trailing}")
+ response: Final = gateway.request_multipart(
+ "/v1/audio/transcriptions",
+ {"model": model},
+ {"file": ("speech.wav", f"RIFF {marker}".encode(), "audio/wav")},
+ )
+ assert body_of(response)["text"] == reply_text(marker)
+ assert_one_request(wire.drain(), _TRANSCRIPTION.target, marker)
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
+def test_google_genai_generate_content_reaches_one_single_slash_url(gateway: Gateway, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _GEMINI, f"{wire.url}/gm/v1beta{trailing}")
+ response: Final = gateway.request(
+ "POST", f"/v1beta/models/{model}:generateContent", _gemini_contents(marker)
+ )
+ assert _gemini_text(body_of(response)) == reply_text(marker)
+ assert_one_request(wire.drain(), _GEMINI.target, marker)
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
+def test_google_genai_stream_generate_content_reaches_one_single_slash_url(gateway: Gateway, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _GEMINI, f"{wire.url}/gm/v1beta{trailing}")
+ response: Final = gateway.request(
+ "POST",
+ f"/v1beta/models/{model}:streamGenerateContent",
+ _gemini_contents(marker),
+ params={"alt": "sse"},
+ )
+ assert response.status_code == 200, response.text
+ assert marker in response.text, response.text
+ assert_one_request(
+ wire.drain(), "/gm/v1beta/models/gemini-3.8-flash:streamGenerateContent?alt=sse", marker
+ )
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
+def test_count_tokens_reaches_one_single_slash_url(gateway: Gateway, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _ANTHROPIC, f"{wire.url}/an{trailing}")
+ response: Final = gateway.request(
+ "POST", "/v1/messages/count_tokens", {"model": model, "messages": [{"role": "user", "content": marker}]}
+ )
+ assert body_of(response)["input_tokens"] == COUNTED_TOKENS
+ assert_one_request(wire.drain(), "/an/v1/messages/count_tokens", marker)
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
+def test_health_check_reaches_one_single_slash_url(gateway: Gateway, trailing: str) -> None:
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _ANTHROPIC, f"{wire.url}/an{trailing}")
+ report: Final = body_of(gateway.request("GET", "/health", params={"model": model}))
+ assert (report["healthy_count"], report["unhealthy_count"]) == (1, 0), report
+ assert [(request.method, request.target) for request in wire.drain()] == [("POST", "/an/v1/messages")]
+
+
[email protected]("trailing", TRAILING, ids=TRAILING_IDS)
+def test_spend_log_records_the_single_slash_url_it_called(gateway: Gateway, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _ANTHROPIC, f"{wire.url}/an{trailing}")
+ response: Final = gateway.request(
+ "POST", "/v1/chat/completions", {"model": model, "messages": [{"role": "user", "content": marker}]}
+ )
+ identity: Final = string_value(body_of(response)["id"])
+ assert_one_request(wire.drain(), "/an/v1/messages", marker)
+ assert _one_spend_row(identity)["api_base"] == f"{wire.url}/an/v1/messages"
+
+
+def test_cache_hit_on_a_slash_base_skips_the_upstream_and_logs_its_own_row(gateway: Gateway) -> None:
+ pytest.skip("BUG: a response-cache hit writes two cache_hit spend rows for one chat completion")
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _ANTHROPIC, f"{wire.url}/an/")
+ body: Final[dict[str, JsonValue]] = {"model": model, "messages": [{"role": "user", "content": marker}]}
+ first: Final = body_of(gateway.request("POST", "/v1/chat/completions", body))
+ second: Final = body_of(gateway.request("POST", "/v1/chat/completions", body))
+ assert first_item(second["choices"])["message"] == first_item(first["choices"])["message"]
+ assert_one_request(wire.drain(), "/an/v1/messages", marker)
+ identity: Final = string_value(first["id"])
+ assert _one_spend_row(identity)["api_base"] == f"{wire.url}/an/v1/messages"
+ hit: Final = _one_spend_row(f"{identity}_cache_hit%")
+ assert hit["cache_hit"] == "True", hit
+ assert "//" not in urlsplit(string_value(hit["api_base"])).path, hit
+
+
[email protected]("suffix", _FULL_ROUTE_SUFFIXES, ids=_FULL_ROUTE_IDS)
+def test_chat_base_that_already_names_the_route_keeps_its_path_and_query(gateway: Gateway, suffix: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _ANTHROPIC, f"{wire.url}/an/v1/messages{suffix}")
+ assert _chat(gateway, model, marker) == reply_text(marker)
+ assert_one_request(wire.drain(), f"/an/v1/messages{suffix.rstrip('/')}", marker)
+
+
[email protected]("suffix", _FULL_ROUTE_SUFFIXES, ids=_FULL_ROUTE_IDS)
+def test_messages_base_that_already_names_the_route_keeps_its_path_and_query(gateway: Gateway, suffix: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _ANTHROPIC, f"{wire.url}/an/v1/messages{suffix}")
+ assert _message_text(gateway, model, marker) == [reply_text(marker)]
+ assert_one_request(wire.drain(), f"/an/v1/messages{suffix.rstrip('/')}", marker)
+
+
[email protected]("suffix", ("", "/", "?tier=a"), ids=("exact", "slash", "query"))
+def test_rerank_base_that_already_names_the_route_keeps_its_path_and_query(gateway: Gateway, suffix: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, Row("cohere/rerank-v3.5", "/co", ""), f"{wire.url}/co/v2/rerank{suffix}")
+ response: Final = gateway.request(
+ "POST", "/v1/rerank", {"model": model, "query": marker, "documents": ["only document"]}
+ )
+ assert [object_value(result)["relevance_score"] for result in list_value(body_of(response)["results"])] == [
+ RELEVANCE
+ ]
+ assert_one_request(wire.drain(), f"/co/v2/rerank{suffix.rstrip('/')}", marker)
+
+
+ ("row", "trailing"),
+ (
+ (_ANTHROPIC, "//"),
+ (Row("anthropic/claude-haiku-5-5", f"/an/{'p' * 5000}", "/v1/messages"), "/"),
+ (_GEMINI, "//"),
+ (Row("gemini/gemini-3.8-flash", f"/gm/{'p' * 5000}", "/models/gemini-3.8-flash:generateContent"), "/"),
+ ),
+ ids=("anthropic-two-slashes", "anthropic-5kb-prefix", "gemini-two-slashes", "gemini-5kb-prefix"),
+)
+def test_base_with_repeated_slashes_or_a_long_prefix_reaches_one_url(gateway: Gateway, row: Row, trailing: str) -> None:
+ marker: Final = new_marker()
+ with wire_server(join_peer) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, row, f"{wire.url}{row.base}{trailing}")
+ assert _chat(gateway, model, marker) == reply_text(marker)
+ assert_one_request(wire.drain(), row.target, marker)
+
+
+def _anthropic_error(status: int, marker: str) -> bytes:
+ return json.dumps(
+ {"type": "error", "error": {"type": "authentication_error", "message": f"scripted {status} {marker}"}}
+ ).encode()
+
+
+def _gemini_error(status: int, marker: str) -> bytes:
+ error: Final = {"code": status, "message": f"scripted {status} {marker}", "status": "NOT_FOUND"}
+ return json.dumps({"error": error}).encode()
+
+
+ ("row", "status"),
+ ((_ANTHROPIC, 401), (_GEMINI, 404)),
+ ids=("anthropic-401", "gemini-404"),
+)
+def test_upstream_error_on_a_slash_base_reaches_the_caller_from_one_url(
+ gateway: Gateway, row: Row, status: int
+) -> None:
+ marker: Final = new_marker()
+ error: Final = _anthropic_error(status, marker) if row is _ANTHROPIC else _gemini_error(status, marker)
+
+ def respond(_: Request) -> Reply:
+ return Reply(status=status, body=error)
+
+ with wire_server(respond) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, row, f"{wire.url}{row.base}/")
+ response: Final = gateway.request(
+ "POST", "/v1/chat/completions", {"model": model, "messages": [{"role": "user", "content": marker}]}
+ )
+ assert response.status_code == status, response.text
+ assert marker in response.text, response.text
+ assert_one_request(wire.drain(), row.target, marker)
+ assert_control_answers(gateway, scenario)
+
+
+def test_streamed_chat_upstream_error_on_a_slash_base_reaches_the_caller(gateway: Gateway) -> None:
+ marker: Final = new_marker()
+
+ def respond(_: Request) -> Reply:
+ return Reply(status=401, body=_anthropic_error(401, marker))
+
+ with wire_server(respond) as wire, gateway.scenario() as scenario:
+ model: Final = _deploy(scenario, _ANTHROPIC, f"{wire.url}/an/")
+ with pytest.raises(openai.AuthenticationError) as raised:
+ _openai(gateway).chat.completions.create(
+ model=model, messages=[{"role": "user", "content": marker}], stream=True
+ )
+ assert marker in str(raised.value)
+ assert_one_request(wire.drain(), "/an/v1/messages", marker)
+ assert_control_answers(gateway, scenario)
diff --git a/tests/unit/litellm_core_utils/test_url_utils.py b/tests/unit/litellm_core_utils/test_url_utils.py
index 46bb91959c0..9b29eac208e 100644
--- a/tests/unit/litellm_core_utils/test_url_utils.py
+++ b/tests/unit/litellm_core_utils/test_url_utils.py
@@ -1,6 +1,8 @@
import asyncio
import socket
import threading
+from collections.abc import Callable
+from urllib.parse import urlsplit
import httpx
import pytest
@@ -13,6 +15,8 @@ from litellm.litellm_core_utils.url_utils import (
assert_same_origin,
encode_url_path_segment,
encode_url_path_segments,
+ ensure_api_route,
+ join_api_base,
validate_url,
)
@@ -571,3 +575,80 @@ async def test_async_safe_get_resolves_dns_off_the_event_loop(monkeypatch):
assert response.status_code == 200
assert resolver_threads and all(thread is not loop_thread for thread in resolver_threads)
assert released_by_the_loop and all(released_by_the_loop)
+
+
+ ("api_base", "route", "expected"),
+ [
+ ("https://gw.test/p/", "v1/messages", "https://gw.test/p/v1/messages"),
+ ("https://gw.test/p//", "v1/messages", "https://gw.test/p/v1/messages"),
+ ("https://gw.test/p", "v1/messages", "https://gw.test/p/v1/messages"),
+ ("https://gw.test/p", "/v1/messages", "https://gw.test/p/v1/messages"),
+ ("https://gw.test/p/", "/v1/messages", "https://gw.test/p/v1/messages"),
+ ("https://gw.test/p/v1/messages", "v1/messages", "https://gw.test/p/v1/messages/v1/messages"),
+ ("https://api.test/v1/evals/cancel", "cancel", "https://api.test/v1/evals/cancel/cancel"),
+ ("http://embeddings", "embeddings", "http://embeddings/embeddings"),
+ ("http:/", "v1/files", "http://v1/files"),
+ ("https://gw.test/p/?x=1", "v1/messages", "https://gw.test/p?x=1/v1/messages"),
+ ],
+ ids=(
+ "trailing-slash-base",
+ "repeated-trailing-slashes",
+ "base-without-slash",
+ "route-with-leading-slash",
+ "slashes-on-both-sides",
+ "base-already-ends-in-route",
+ "id-equal-to-the-route",
+ "host-named-like-the-route",
+ "host-stripped-as-a-suffix-leaves-only-the-scheme",
+ "slash-before-the-query-matches-the-base-without-it",
+ ),
+)
+def test_join_api_base_always_appends_the_route_after_exactly_one_slash(
+ api_base: str, route: str, expected: str
+) -> None:
+ assert join_api_base(api_base, route) == expected
+
+
[email protected]("api_base", ["https://", "http://", "https:///"])
[email protected]("route", ["v1/messages", "models/gemini-flash:generateContent"])
[email protected]("join", [join_api_base, ensure_api_route], ids=["join_api_base", "ensure_api_route"])
+def test_a_base_that_is_only_a_scheme_never_turns_the_route_into_the_host(
+ api_base: str, route: str, join: Callable[[str, str], str]
+) -> None:
+ assert urlsplit(join(api_base, route)).netloc == ""
+
+
+ ("api_base", "route", "expected"),
+ [
+ ("https://gw.test/p/v1/messages", "v1/messages", "https://gw.test/p/v1/messages"),
+ ("https://gw.test/p/v1/messages/", "/v1/messages", "https://gw.test/p/v1/messages"),
+ ("https://gw.test/x/listen", "listen", "https://gw.test/x/listen"),
+ ("https://gw.test/p/", "v1/messages", "https://gw.test/p/v1/messages"),
+ ("https://gw.test/p", "/v1/messages", "https://gw.test/p/v1/messages"),
+ ("https://gw.test/research", "search", "https://gw.test/research/search"),
+ ("http://embeddings", "embeddings", "http://embeddings/embeddings"),
+ ("http://embeddings:8080/", "embeddings", "http://embeddings:8080/embeddings"),
+ ("http:/", "anthropic/v1/messages", "http://anthropic/v1/messages"),
+ ("https://gw.test/p/v1/messages/?beta=true", "v1/messages", "https://gw.test/p/v1/messages?beta=true"),
+ ("https://gw.test/p/v1/messages?beta=true", "v1/messages", "https://gw.test/p/v1/messages?beta=true"),
+ ],
+ ids=(
+ "path-already-ends-in-route",
+ "path-ends-in-route-before-trailing-slash",
+ "last-segment-is-the-route",
+ "trailing-slash-base",
+ "route-with-leading-slash",
+ "path-ends-in-lookalike-suffix",
+ "host-named-like-the-route",
+ "host-with-port-named-like-the-route",
+ "host-stripped-as-a-suffix-leaves-only-the-scheme",
+ "route-path-with-a-slash-before-the-query",
+ "route-path-before-the-query",
+ ),
+)
+def test_ensure_api_route_appends_the_route_only_when_the_url_path_lacks_it(
+ api_base: str, route: str, expected: str
+) -> None:
+ assert ensure_api_route(api_base, route) == expected
diff --git a/tests/unit/llms/azure/realtime/test_azure_realtime_handler.py b/tests/unit/llms/azure/realtime/test_azure_realtime_handler.py
index 5c1ff1b46a3..f1db76c1e21 100644
--- a/tests/unit/llms/azure/realtime/test_azure_realtime_handler.py
+++ b/tests/unit/llms/azure/realtime/test_azure_realtime_handler.py
@@ -740,3 +740,15 @@ async def test_arealtime_forwards_deployment_azure_ad_token(monkeypatch, no_ambi
)
assert mock_async_realtime.call_args.kwargs["azure_ad_token"] == "deployment-entra-token"
+
+
[email protected]("protocol", [None, "GA"])
+def test_construct_url_with_a_trailing_slash_api_base_matches_the_url_without_it(protocol):
+ from litellm.llms.azure.realtime.handler import AzureOpenAIRealtime
+
+ def url_for(api_base: str) -> str:
+ return AzureOpenAIRealtime().construct_url(
+ api_base=api_base, model="gpt-realtime", api_version="2024-10-01-preview", realtime_protocol=protocol
+ )
+
+ assert url_for("https://gw.example.com/azure/") == url_for("https://gw.example.com/azure")
diff --git a/tests/unit/llms/azure/test_azure_common_utils.py b/tests/unit/llms/azure/test_azure_common_utils.py
index b3bd1b8743a..1dc7ff74323 100644
--- a/tests/unit/llms/azure/test_azure_common_utils.py
+++ b/tests/unit/llms/azure/test_azure_common_utils.py
@@ -1702,6 +1702,24 @@ def test_azure_v1_api_uses_openai_client(api_version):
), f"base_url should contain /openai/v1/, got {async_client.base_url}"
+ "api_base",
+ [
+ "https://test.openai.azure.com",
+ "https://test.openai.azure.com/",
+ "https://test.openai.azure.com/openai/v1",
+ "https://test.openai.azure.com/openai/v1/",
+ ],
+)
[email protected]("is_async", [False, True])
+def test_azure_v1_api_client_targets_openai_v1_once_whatever_the_api_base_ends_in(api_base: str, is_async: bool):
+ client: Final = BaseAzureLLM().get_azure_openai_client(
+ api_key="test-key", api_base=api_base, api_version="v1", _is_async=is_async
+ )
+
+ assert str(client.base_url) == "https://test.openai.azure.com/openai/v1/"
+
+
@pytest.mark.parametrize("api_version", ["v1", "latest", "preview"])
def test_azure_v1_api_with_azure_ad_token_provider(api_version):
"""
diff --git a/tests/unit/llms/azure_ai/agents/test_handler.py b/tests/unit/llms/azure_ai/agents/test_handler.py
index f92f5359c49..71466bb172c 100644
--- a/tests/unit/llms/azure_ai/agents/test_handler.py
+++ b/tests/unit/llms/azure_ai/agents/test_handler.py
@@ -22,10 +22,10 @@ def _httpx_only_transport(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setenv("DISABLE_AIOHTTP_TRANSPORT", "True")
-def _assistants_api(thread_id: str, answer: str) -> respx.MockRouter:
- thread_path: Final = f"/api/projects/litellm-test/threads/{thread_id}"
+def _assistants_api(thread_id: str, answer: str, project_path: str = "/api/projects/litellm-test") -> respx.MockRouter:
+ thread_path: Final = f"{project_path}/threads/{thread_id}"
router: Final = respx.mock(assert_all_called=False)
- router.post(path="/api/projects/litellm-test/threads").respond(
+ router.post(path=f"{project_path}/threads").respond(
200, json={"id": thread_id, "object": "thread", "created_at": 1767225600, "metadata": {}}
)
router.post(path=f"{thread_path}/messages").respond(
@@ -116,6 +116,23 @@ async def test_acompletion_creates_thread_posts_message_runs_and_returns_the_rep
assert response._hidden_params["thread_id"] == "thread_new"
+async def test_acompletion_creates_the_thread_under_a_project_named_threads() -> None:
+ router: Final = _assistants_api("thread_new", "hello", project_path="/api/projects/threads")
+
+ with router:
+ await litellm.acompletion(
+ model=f"azure_ai/agents/{AGENT_ID}",
+ messages=[{"role": "user", "content": "hi"}],
+ api_base="https://agents-test.services.ai.azure.com/api/projects/threads",
+ api_key="azure-ad-token",
+ stream=False,
+ )
+ recorded: Final = _recorded(router)
+
+ assert recorded[0][:2] == ("POST", f"/api/projects/threads/threads?api-version={API_VERSION}")
+
+
@pytest.mark.asyncio
async def test_handler_acompletion_with_thread_id_reuses_that_thread() -> None:
router: Final = _assistants_api("thread_existing", "Your name is Alice.")
diff --git a/tests/unit/llms/azure_ai/claude/test_azure_anthropic_count_tokens_transformation.py b/tests/unit/llms/azure_ai/claude/test_azure_anthropic_count_tokens_transformation.py
index 4b317cff975..8971bd31a60 100644
--- a/tests/unit/llms/azure_ai/claude/test_azure_anthropic_count_tokens_transformation.py
+++ b/tests/unit/llms/azure_ai/claude/test_azure_anthropic_count_tokens_transformation.py
@@ -4,8 +4,9 @@ Tests for Azure AI Anthropic CountTokens transformation.
Verifies that the CountTokens API uses the correct authentication headers.
"""
+from typing import Final
-
+import pytest
from litellm.llms.azure_ai.anthropic.count_tokens.transformation import (
AzureAIAnthropicCountTokensConfig,
@@ -105,3 +106,22 @@ class TestAzureAIAnthropicCountTokensConfig:
endpoint
== "https://my-resource.services.ai.azure.com/anthropic/v1/messages/count_tokens"
)
+
+ def test_get_count_tokens_endpoint_for_a_resource_named_anthropic(self):
+ config: Final = AzureAIAnthropicCountTokensConfig()
+
+ endpoint: Final = config.get_count_tokens_endpoint("https://anthropic-res.services.ai.azure.com")
+
+ assert endpoint == "https://anthropic-res.services.ai.azure.com/anthropic/v1/messages/count_tokens"
+
+ @pytest.mark.parametrize(
+ "api_base",
+ [
+ "https://my-resource.services.ai.azure.com/anthropic/v1/messages",
+ "https://my-resource.services.ai.azure.com/anthropic/v1/messages/",
+ ],
+ )
+ def test_get_count_tokens_endpoint_sits_next_to_the_messages_url(self, api_base: str):
+ endpoint: Final = AzureAIAnthropicCountTokensConfig().get_count_tokens_endpoint(api_base)
+
+ assert endpoint == "https://my-resource.services.ai.azure.com/anthropic/v1/messages/count_tokens"
diff --git a/tests/unit/llms/azure_ai/claude/test_azure_anthropic_messages_transformation.py b/tests/unit/llms/azure_ai/claude/test_azure_anthropic_messages_transformation.py
index 38483b8a1fa..5e0395f84dc 100644
--- a/tests/unit/llms/azure_ai/claude/test_azure_anthropic_messages_transformation.py
+++ b/tests/unit/llms/azure_ai/claude/test_azure_anthropic_messages_transformation.py
@@ -198,6 +198,17 @@ class TestAzureAnthropicMessagesConfig:
assert url == "https://test.services.ai.azure.com/anthropic/v1/messages"
+ def test_get_complete_url_adds_the_anthropic_path_for_a_host_named_anthropic(self):
+ url: Final = AzureAnthropicMessagesConfig().get_complete_url(
+ api_base="https://anthropic-res.services.ai.azure.com",
+ api_key="test-api-key",
+ model="claude-sonnet-4-5",
+ optional_params={},
+ litellm_params={},
+ )
+
+ assert url == "https://anthropic-res.services.ai.azure.com/anthropic/v1/messages"
+
def test_get_complete_url_raises_error_when_api_base_missing(self):
"""Test get_complete_url raises error when api_base is None"""
config = AzureAnthropicMessagesConfig()
diff --git a/tests/unit/llms/bedrock/chat/agentcore/test_agentcore_transformation.py b/tests/unit/llms/bedrock/chat/agentcore/test_agentcore_transformation.py
index 053ba477212..413061a732d 100644
--- a/tests/unit/llms/bedrock/chat/agentcore/test_agentcore_transformation.py
+++ b/tests/unit/llms/bedrock/chat/agentcore/test_agentcore_transformation.py
@@ -1269,3 +1269,17 @@ def test_agentcore_runtime_response_becomes_assistant_message(
assert json.loads(route.calls[0].request.content)["prompt"] == "Explain machine learning in simple terms"
assert response.choices[0].message.content == "Machine learning lets computers learn from data."
assert response.choices[0].finish_reason == "stop"
+
+
+def test_get_complete_url_sends_a_non_string_runtime_endpoint_to_that_endpoint():
+ vpc_endpoint: Final = "https://vpce-1.bedrock-agentcore.us-west-2.vpce.amazonaws.com"
+
+ url: Final = AmazonAgentCoreConfig().get_complete_url(
+ api_base=None,
+ api_key=None,
+ model="agentcore/arn:aws:bedrock-agentcore:us-west-2:111122223333:runtime/my-agent",
+ optional_params={"aws_bedrock_runtime_endpoint": httpx.URL(f"{vpc_endpoint}/")},
+ litellm_params={},
+ )
+
+ assert url.startswith(f"{vpc_endpoint}/runtimes/")
diff --git a/tests/unit/llms/openai/evals/test_openai_evals_transformation.py b/tests/unit/llms/openai/evals/test_openai_evals_transformation.py
index f359f214f97..ed8f86d4727 100644
--- a/tests/unit/llms/openai/evals/test_openai_evals_transformation.py
+++ b/tests/unit/llms/openai/evals/test_openai_evals_transformation.py
@@ -264,6 +264,17 @@ def test_transform_cancel_eval_request(config: OpenAIEvalsConfig):
assert request_body == {}
+def test_transform_cancel_eval_request_appends_cancel_to_an_eval_id_named_cancel(config: OpenAIEvalsConfig) -> None:
+ eval_url, _ = config.transform_get_eval_request(
+ eval_id="cancel", api_base="https://api.openai.com", litellm_params=GenericLiteLLMParams(), headers={}
+ )
+ cancel_url, _, _ = config.transform_cancel_eval_request(
+ eval_id="cancel", api_base="https://api.openai.com", litellm_params=GenericLiteLLMParams(), headers={}
+ )
+
+ assert cancel_url == f"{eval_url}/cancel"
+
+
def test_transform_cancel_eval_response(config: OpenAIEvalsConfig):
"""Test transformation of cancel eval response"""
response = httpx.Response(
diff --git a/tests/unit/llms/test_api_base_trailing_slash_routes.py b/tests/unit/llms/test_api_base_trailing_slash_routes.py
new file mode 100644
index 00000000000..ce8cecf11a5
--- /dev/null
+++ b/tests/unit/llms/test_api_base_trailing_slash_routes.py
@@ -0,0 +1,119 @@
+import asyncio
+import contextlib
+from collections.abc import Callable, Iterator, Mapping
+from types import MappingProxyType
+from typing import Final
+
+import pytest
+import respx
+
+import litellm
+from litellm.evals.main import cancel_run, create_run, get_run
+
+FAKE_KEY: Final = "sk-fake"
+FAKE_AWS: Final[Mapping[str, str]] = MappingProxyType(
+ {"aws_access_key_id": "AKIAFAKE", "aws_secret_access_key": "fake-secret", "aws_region_name": "us-east-1"}
+)
+MESSAGES: Final = ({"role": "user", "content": "hi"},)
+GATEWAY_PATH: Final = "/gw/"
+
+
+GATEWAY_HOST: Final = "gateway.test"
+
+
+def received_paths() -> Iterator[Callable[[], tuple[str, ...]]]:
+ with respx.mock(assert_all_called=False) as router:
+ route: Final = router.route(host=GATEWAY_HOST).respond(
+ 404, json={"error": {"message": "not found", "code": 404}}
+ )
+ yield lambda: tuple(call.request.url.raw_path.decode().split("?", 1)[0] for call in route.calls)
+
+
+def _drain(result: object) -> None:
+ if isinstance(result, litellm.CustomStreamWrapper):
+ for _ in result:
+ pass
+
+
+def _completion(model: str, stream: bool = False, **kwargs: object) -> Callable[[str], object]:
+ return lambda api_base: _drain(
+ litellm.completion(model=model, messages=list(MESSAGES), api_base=api_base, stream=stream, timeout=5, **kwargs)
+ )
+
+
+def _embedding(model: str, **kwargs: object) -> Callable[[str], object]:
+ return lambda api_base: litellm.embedding(model=model, input=["hi"], api_base=api_base, timeout=5, **kwargs)
+
+
+ROUTES: Final[Mapping[str, Callable[[str], object]]] = MappingProxyType(
+ {
+ "anthropic-chat": _completion("anthropic/claude-opus-5-5", api_key=FAKE_KEY),
+ "bedrock-converse": _completion("bedrock/converse/anthropic.claude-sonnet-5-5", **FAKE_AWS),
+ "bedrock-converse-stream": _completion("bedrock/converse/anthropic.claude-sonnet-5-5", stream=True, **FAKE_AWS),
+ "gemini-chat": _completion("gemini/gemini-pro-latest", api_key=FAKE_KEY),
+ "gemini-chat-stream": _completion("gemini/gemini-pro-latest", stream=True, api_key=FAKE_KEY),
+ "mistral-chat": _completion("mistral/mistral-large-latest", api_key=FAKE_KEY),
+ "gigachat-chat": _completion("gigachat/GigaChat-2-Max", api_key=FAKE_KEY),
+ "bedrock-titan-embedding": _embedding("bedrock/amazon.titan-embed-text-v2:0", **FAKE_AWS),
+ "gemini-embedding": _embedding("gemini/gemini-embedding-2", api_key=FAKE_KEY),
+ "gigachat-embedding": _embedding("gigachat/Embeddings-2", api_key=FAKE_KEY),
+ "volcengine-embedding": _embedding("volcengine/doubao-embedding-text-240715", api_key=FAKE_KEY),
+ "ollama-embedding": _embedding("ollama/nomic-embed-text"),
+ "oobabooga-chat": _completion("oobabooga/local-model"),
+ "oobabooga-embedding": _embedding("oobabooga/local-model"),
+ "bedrock-rerank": lambda api_base: litellm.rerank(
+ model="bedrock/arn:aws:bedrock:us-west-2::foundation-model/amazon.rerank-v1:0",
+ query="hi",
+ documents=["a"],
+ api_base=api_base,
+ **FAKE_AWS,
+ ),
+ "bedrock-image-generation": lambda api_base: litellm.image_generation(
+ model="bedrock/amazon.nova-canvas-v1:0", prompt="a cat", api_base=api_base, timeout=5, **FAKE_AWS
+ ),
+ "gemini-interaction-get": lambda api_base: litellm.interactions.get(
+ interaction_id="interaction-1", api_base=api_base, api_key=FAKE_KEY
+ ),
+ "gemini-interaction-delete": lambda api_base: litellm.interactions.delete(
+ interaction_id="interaction-1", api_base=api_base, api_key=FAKE_KEY
+ ),
+ "gemini-interaction-cancel": lambda api_base: litellm.interactions.cancel(
+ interaction_id="interaction-1", api_base=api_base, api_key=FAKE_KEY
+ ),
+ "openai-eval-run-create": lambda api_base: create_run(
+ eval_id="eval-1", data_source={"type": "jsonl"}, api_base=api_base, api_key=FAKE_KEY
+ ),
+ "openai-eval-run-get": lambda api_base: get_run(
+ eval_id="eval-1", run_id="run-1", api_base=api_base, api_key=FAKE_KEY
+ ),
+ "openai-eval-run-cancel": lambda api_base: cancel_run(
+ eval_id="eval-1", run_id="run-1", api_base=api_base, api_key=FAKE_KEY
+ ),
+ "edenai-model-list": lambda api_base: litellm.get_valid_models(
+ check_provider_endpoint=True, custom_llm_provider="edenai", api_key=FAKE_KEY, api_base=api_base
+ ),
+ "gemini-rag-ingest": lambda api_base: asyncio.run(
+ litellm.aingest(
+ ingest_options={
+ "vector_store": {"custom_llm_provider": "gemini", "api_base": api_base, "api_key": FAKE_KEY}
+ },
+ file_data=("notes.txt", b"hello", "text/plain"),
+ )
+ ),
+ }
+)
+
+
[email protected]("route", ROUTES.values(), ids=ROUTES.keys())
+def test_api_base_with_a_trailing_slash_reaches_the_gateway_path_with_single_slashes(
+ received_paths: Callable[[], tuple[str, ...]], route: Callable[[str], object], monkeypatch: pytest.MonkeyPatch
+) -> None:
+ monkeypatch.setenv("GIGACHAT_ACCESS_TOKEN", FAKE_KEY)
+ monkeypatch.setattr(litellm, "disable_aiohttp_transport", True)
+ with contextlib.suppress(Exception):
+ route(f"http://{GATEWAY_HOST}{GATEWAY_PATH}")
+
+ paths: Final = received_paths()
+ assert paths, "the route never reached the api_base"
+ assert all(path.startswith(GATEWAY_PATH) and "//" not in path for path in paths), paths
diff --git a/tests/unit/llms/test_get_complete_url_trailing_slash.py b/tests/unit/llms/test_get_complete_url_trailing_slash.py
new file mode 100644
index 00000000000..9118498c2c2
--- /dev/null
+++ b/tests/unit/llms/test_get_complete_url_trailing_slash.py
@@ -0,0 +1,278 @@
+import copy
+import importlib
+import inspect
+import itertools
+from collections.abc import Mapping
+from dataclasses import dataclass, field
+from pathlib import Path
+from types import MappingProxyType, ModuleType
+from typing import Final
+from urllib.parse import urlsplit, urlunsplit
+
+import pytest
+from _pytest.mark.structures import ParameterSet
+
+import litellm
+from litellm.llms.elevenlabs.text_to_speech.transformation import ElevenLabsTextToSpeechConfig
+from litellm.llms.meta.realtime.transformation import MUSE_MODEL
+from litellm.llms.openai_like.dynamic_config import create_config_class, create_responses_config_class
+from litellm.llms.openai_like.json_loader import SimpleProviderConfig
+
+API_BASE_WITH_TRAILING_SLASH: Final = "https://gw.example.com/p/"
+API_BASE_WITHOUT_TRAILING_SLASH: Final = API_BASE_WITH_TRAILING_SLASH.rstrip("/")
+LLMS_ROOT: Final = Path(litellm.__file__).parent / "llms"
+PACKAGE_ROOT: Final = LLMS_ROOT.parent.parent
+
+ARGUMENTS_BY_NAME: Final[Mapping[str, object]] = MappingProxyType(
+ {
+ "api_key": "sk-test",
+ "model": "test-model",
+ "optional_params": {},
+ "litellm_params": {"api_key": "sk-test", "vertex_project": "project", "vertex_location": "us-central1"},
+ "stream": False,
+ "endpoint": "v1/messages",
+ "request_query_params": None,
+ "data": None,
+ "api_version": "2024-01-01",
+ "vector_store_id": "vs_test",
+ }
+)
+
+
+SAMPLE_JSON_PROVIDER: Final = SimpleProviderConfig(
+ "json-provider", {"base_url": "https://json-provider.example.com", "api_key_env": "UNSET_KEY"}
+)
+FACTORY_BUILDERS: Final[tuple[type, ...]] = (
+ create_config_class(SAMPLE_JSON_PROVIDER),
+ create_responses_config_class(SAMPLE_JSON_PROVIDER),
+)
+
+
+async def _resolve_access_token() -> str:
+ return "token"
+
+
+@dataclass(frozen=True, slots=True)
+class BuilderInputs:
+ constructor: Mapping[str, object] = field(default_factory=lambda: MappingProxyType({}))
+ call: Mapping[str, object] = field(default_factory=lambda: MappingProxyType({}))
+ api_base_option: str | None = None
+
+
+INPUTS_BY_BUILDER: Final[Mapping[str, BuilderInputs]] = MappingProxyType(
+ {
+ "anthropic.skills.transformation.AnthropicSkillsConfig": BuilderInputs(call={"endpoint": "skills"}),
+ "azure_ai.image_edit.transformation.AzureFoundryFluxImageEditConfig": BuilderInputs(
+ call={"litellm_params": {"api_version": "2025-04-01-preview"}}
+ ),
+ "bedrock.chat.agentcore.transformation.AmazonAgentCoreConfig": BuilderInputs(
+ call={"model": "agentcore/arn:aws:bedrock-agentcore:us-west-2:123456789012:runtime/test-runtime"},
+ api_base_option="aws_bedrock_runtime_endpoint",
+ ),
+ "bedrock.chat.invoke_agent.transformation.AmazonInvokeAgentConfig": BuilderInputs(
+ call={"model": "agent/AGENT_ID/ALIAS_ID", "optional_params": {"sessionID": "session-id"}}
+ ),
+ "black_forest_labs.image_edit.transformation.BlackForestLabsImageEditConfig": BuilderInputs(
+ call={"model": "flux-kontext-pro"}
+ ),
+ "black_forest_labs.image_generation.transformation.BlackForestLabsImageGenerationConfig": BuilderInputs(
+ call={"model": "flux-pro-1.1"}
+ ),
+ "elevenlabs.text_to_speech.transformation.ElevenLabsTextToSpeechConfig": BuilderInputs(
+ call={"litellm_params": {ElevenLabsTextToSpeechConfig.ELEVENLABS_VOICE_ID_KEY: "voice-id"}}
+ ),
+ "meta.realtime.transformation.MetaRealtimeConfig": BuilderInputs(call={"model": MUSE_MODEL}),
+ "openai.evals.transformation.OpenAIEvalsConfig": BuilderInputs(call={"endpoint": "evals"}),
+ "openai_like.messages.transformation.JSONProviderAnthropicMessagesConfig": BuilderInputs(
+ constructor={"provider": SAMPLE_JSON_PROVIDER}
+ ),
+ "predibase.chat.transformation.PredibaseConfig": BuilderInputs(
+ call={"litellm_params": {"predibase_tenant_id": "tenant"}}
+ ),
+ "ragflow.chat.transformation.RAGFlowConfig": BuilderInputs(call={"model": "ragflow/chat/chat-id/model"}),
+ "scaledown.chat.transformation.ScaleDownChatConfig": BuilderInputs(call={"model": "classify"}),
+ "vertex_ai.audio_transcription.realtime_transformation.VertexChirpRealtimeConfig": BuilderInputs(
+ constructor={"resolve_access_token": _resolve_access_token, "project": "project", "location": None},
+ call={"model": "chirp_3"},
+ ),
+ "vertex_ai.interactions.transformation.VertexAIInteractionsConfig": BuilderInputs(
+ constructor={"mint_access_token": lambda _credentials, _project: ("token", "project")}
+ ),
+ "vertex_ai.realtime.transformation.VertexAIRealtimeConfig": BuilderInputs(
+ constructor={"access_token": "token", "project": "project", "location": "us-central1"}
+ ),
+ "vertex_ai.text_to_speech.transformation.VertexAILyriaTextToSpeechConfig": BuilderInputs(
+ call={"model": "lyria-002"}
+ ),
+ }
+)
+
+UNCALLABLE_BUILDERS: Final[Mapping[str, str]] = MappingProxyType(
+ {
+ "base_llm.search.transformation.BaseSearchConfig": "raises-NotImplementedError-for-providers-to-override",
+ "bedrock.image_edit.amazon_nova_canvas_image_edit_transformation.BedrockAmazonNovaCanvasImageEditConfig": (
+ "raises-because-BedrockImageEdit-builds-the-url"
+ ),
+ "sap.chat.transformation.GenAIHubOrchestrationConfig": "url-comes-from-a-live-SAP-deployment-lookup",
+ "sap.embed.transformation.GenAIHubEmbeddingConfig": "url-comes-from-a-live-SAP-deployment-lookup",
+ "triton.completion.transformation.TritonConfig": "api_base-must-be-the-full-generate-or-infer-url",
+ "vertex_ai.rerank.transformation.VertexAIRerankConfig": "ignores-api_base-and-needs-Vertex-credentials",
+ }
+)
+
+
+API_BASE_IS_THE_FULL_URL: Final[Mapping[str, str]] = MappingProxyType(
+ {
+ builder_id: "sends-the-configured-api_base-as-the-request-url"
+ for builder_id in (
+ "base_llm.audio_transcription.transformation.BaseAudioTranscriptionConfig",
+ "base_llm.chat.transformation.BaseConfig",
+ "base_llm.completion.transformation.BaseTextCompletionConfig",
+ "base_llm.embedding.transformation.BaseEmbeddingConfig",
+ "base_llm.image_generation.transformation.BaseImageGenerationConfig",
+ "base_llm.image_variations.transformation.BaseImageVariationConfig",
+ "base_llm.vector_store.transformation.BaseDirectVectorStoreConfig",
+ "bedrock.realtime.transformation.BedrockRealtimeConfig",
+ "bedrock.search.transformation.AgentCoreSearchConfig",
+ "brave.search.transformation.BraveSearchConfig",
+ "cohere.chat.v2_transformation.CohereV2ChatConfig",
+ "cohere.embed.transformation.CohereEmbeddingConfig",
+ "dashscope.image_generation.transformation.DashScopeImageGenerationConfig",
+ "dataforseo.search.transformation.DataForSEOSearchConfig",
+ "datarobot.chat.transformation.DataRobotConfig",
+ "duckduckgo.search.transformation.DuckDuckGoSearchConfig",
+ "google_pse.search.transformation.GooglePSESearchConfig",
+ "searchapi.search.transformation.SearchAPIConfig",
+ "tinyfish.search.transformation.TinyfishSearchConfig",
+ "vertex_ai.text_to_speech.transformation.VertexAITextToSpeechConfig",
+ "vertex_ai.vertex_ai_partner_models.anthropic.experimental_pass_through.transformation."
+ "VertexAIPartnerModelsAnthropicMessagesConfig",
+ )
+ }
+)
+
+
+HOST_NAMED_LIKE_A_STRIPPED_SUFFIX_IS_NOT_KEPT: Final[Mapping[str, str]] = MappingProxyType(
+ {
+ "deepinfra.rerank.transformation.DeepinfraRerankConfig": "removes-every-openai-substring-on-main-too",
+ "minimax.videos.transformation.MinimaxVideoConfig": "strips-surface-suffixes-to-the-scheme-on-main-too",
+ "volcengine.videos.transformation.VolcEngineVideoConfig": "strips-surface-suffixes-to-the-scheme-on-main-too",
+ "vertex_ai.audio_transcription.realtime_transformation.VertexChirpRealtimeConfig": "returns-a-grpc-host",
+ "azure.realtime.http_transformation.AzureRealtimeHTTPConfig": "replaces-the-host-with-a-route-segment-on-main-too",
+ "bedrock.responses.transformation.BedrockOpenAIResponsesConfig": "replaces-the-host-with-a-route-segment-on-main-too",
+ "bedrock_mantle.chat.claude_transformation.BedrockMantleClaudeChatConfig": "replaces-the-host-with-a-route-segment-on-main-too",
+ "bedrock_mantle.messages.transformation.BedrockMantleAnthropicMessagesConfig": "replaces-the-host-with-a-route-segment-on-main-too",
+ "bedrock_mantle.responses.transformation.BedrockMantleResponsesAPIConfig": "replaces-the-host-with-a-route-segment-on-main-too",
+ "deepseek.messages.transformation.DeepSeekAnthropicMessagesConfig": "replaces-the-host-with-a-route-segment-on-main-too",
+ "ragflow.chat.transformation.RAGFlowConfig": "replaces-the-host-with-a-route-segment-on-main-too",
+ "scaledown.chat.transformation.ScaleDownChatConfig": "replaces-the-host-with-a-route-segment-on-main-too",
+ }
+)
+
+
+def _builder_id(cls: type) -> str:
+ return f"{cls.__module__.removeprefix('litellm.llms.')}.{cls.__qualname__}"
+
+
+def _module_name(path: Path) -> str:
+ return ".".join(path.relative_to(PACKAGE_ROOT).with_suffix("").parts)
+
+
+def _builder_modules() -> tuple[ModuleType, ...]:
+ builder_paths: Final = tuple(path for path in LLMS_ROOT.rglob("*.py") if "def get_complete_url" in path.read_text())
+ return tuple(importlib.import_module(_module_name(path)) for path in sorted(builder_paths))
+
+
+def _defined_builders(module: ModuleType) -> tuple[type, ...]:
+ classes: Final = (*(cls for _, cls in inspect.getmembers(module, inspect.isclass)), *FACTORY_BUILDERS)
+ return tuple(cls for cls in classes if cls.__module__ == module.__name__ and "get_complete_url" in vars(cls))
+
+
+def _all_builders() -> tuple[type, ...]:
+ builders: Final = itertools.chain.from_iterable(_defined_builders(module) for module in _builder_modules())
+ return tuple(sorted(builders, key=_builder_id))
+
+
+def _skip_reason(cls: type) -> str | None:
+ if getattr(vars(cls)["get_complete_url"], "__isabstractmethod__", False):
+ return "abstract-get_complete_url"
+ return UNCALLABLE_BUILDERS.get(_builder_id(cls))
+
+
+def _builder_param(cls: type) -> ParameterSet:
+ reason: Final = _skip_reason(cls)
+ if reason is None:
+ return pytest.param(cls, id=_builder_id(cls))
+ return pytest.param(cls, id=f"{_builder_id(cls)}-skipped-{reason}", marks=pytest.mark.skip(reason=reason))
+
+
+def _unimplemented(*_args: object, **_kwargs: object) -> object:
+ raise NotImplementedError
+
+
+def _instantiable(cls: type) -> type:
+ abstract_methods: Final[frozenset[str]] = getattr(cls, "__abstractmethods__", frozenset())
+ if not abstract_methods:
+ return cls
+ return type(cls.__name__, (cls,), dict.fromkeys(abstract_methods, _unimplemented))
+
+
+def _call_arguments(builder: object, inputs: BuilderInputs, api_base: str) -> dict[str, object]:
+ parameters: Final = inspect.signature(getattr(builder, "get_complete_url")).parameters.values()
+ named: Final = tuple(p for p in parameters if p.kind not in (p.VAR_POSITIONAL, p.VAR_KEYWORD))
+ api_base_options: Final = (
+ {} if inputs.api_base_option is None else {"optional_params": {inputs.api_base_option: api_base}}
+ )
+ values: Final = {**ARGUMENTS_BY_NAME, "api_base": api_base, **inputs.call, **api_base_options}
+ missing: Final = tuple(p.name for p in named if p.default is p.empty and p.name not in values)
+ assert not missing, f"add a value for {missing} to ARGUMENTS_BY_NAME or INPUTS_BY_BUILDER"
+ return {p.name: copy.deepcopy(values[p.name]) for p in named if p.name in values}
+
+
+def _built_url(cls: type, api_base: str) -> str:
+ inputs: Final = INPUTS_BY_BUILDER.get(_builder_id(cls), BuilderInputs())
+ builder: Final = _instantiable(cls)(**inputs.constructor)
+ result: Final = getattr(builder, "get_complete_url")(**_call_arguments(builder, inputs, api_base))
+ return str(result[0] if isinstance(result, tuple) else result)
+
+
+def _without_one_trailing_path_slash(url: str) -> str:
+ parts: Final = urlsplit(url)
+ return urlunsplit(parts._replace(path=parts.path.removesuffix("/")))
+
+
[email protected]("builder_class", [_builder_param(cls) for cls in _all_builders()])
+def test_get_complete_url_keeps_one_slash_after_a_trailing_slash_api_base(builder_class: type) -> None:
+ url: Final = _built_url(builder_class, API_BASE_WITH_TRAILING_SLASH)
+
+ assert "//" not in urlsplit(url).path, url
+
+
[email protected]("builder_class", [_builder_param(cls) for cls in _all_builders()])
+def test_get_complete_url_sends_the_same_url_with_or_without_a_trailing_slash(builder_class: type) -> None:
+ with_slash: Final = _built_url(builder_class, API_BASE_WITH_TRAILING_SLASH)
+ without_slash: Final = _built_url(builder_class, API_BASE_WITHOUT_TRAILING_SLASH)
+
+ if _builder_id(builder_class) in API_BASE_IS_THE_FULL_URL:
+ assert _without_one_trailing_path_slash(with_slash) == _without_one_trailing_path_slash(without_slash)
+ else:
+ assert with_slash == without_slash
+
+
+def _host_preserving_builder_param(cls: type) -> ParameterSet:
+ reason: Final = HOST_NAMED_LIKE_A_STRIPPED_SUFFIX_IS_NOT_KEPT.get(_builder_id(cls))
+ if reason is None:
+ return _builder_param(cls)
+ return pytest.param(cls, id=f"{_builder_id(cls)}-skipped-{reason}", marks=pytest.mark.skip(reason=reason))
+
+
[email protected]("api_base", ["https://v1", "https://api/v1", "https://openai/v1/", "https://anthropic/v1"])
[email protected]("builder_class", [_host_preserving_builder_param(cls) for cls in _all_builders()])
+def test_get_complete_url_keeps_a_host_named_like_the_suffix_it_strips(builder_class: type, api_base: str) -> None:
+ url: Final = _built_url(builder_class, api_base)
+ follows_the_api_base_host: Final = (
+ urlsplit(_built_url(builder_class, "https://gateway.test/v1")).netloc == "gateway.test"
+ )
+
+ assert urlsplit(url).netloc, url
+ assert not follows_the_api_base_host or urlsplit(url).netloc == urlsplit(api_base).netloc, url
diff --git a/tests/unit/llms/test_predibase_transformation.py b/tests/unit/llms/test_predibase_transformation.py
index 20d547c3b63..a02840ed4ef 100644
--- a/tests/unit/llms/test_predibase_transformation.py
+++ b/tests/unit/llms/test_predibase_transformation.py
@@ -1,3 +1,4 @@
+from typing import Final
from unittest.mock import AsyncMock, Mock
import httpx
@@ -84,6 +85,19 @@ def test_predibase_get_complete_url_stream_and_non_stream():
assert stream_url.endswith("/generate_stream")
[email protected](("model", "stream"), [("generate", False), ("generate_stream", True)])
+def test_predibase_get_complete_url_appends_the_route_to_a_model_named_like_it(model: str, stream: bool) -> None:
+ url: Final = PredibaseConfig().get_complete_url(
+ api_base="https://serving.example.com",
+ api_key="test-key",
+ model=model,
+ optional_params={"stream": stream},
+ litellm_params={"predibase_tenant_id": "tenant-123"},
+ )
+
+ assert url.endswith(f"/llms/{model}/{model}")
+
+
def test_predibase_get_complete_url_missing_tenant_id():
config = PredibaseConfig()
diff --git a/tests/unit/llms/vertex_ai/test_vertex_llm_base.py b/tests/unit/llms/vertex_ai/test_vertex_llm_base.py
index 17a92b31284..dd2b86b6f81 100644
--- a/tests/unit/llms/vertex_ai/test_vertex_llm_base.py
+++ b/tests/unit/llms/vertex_ai/test_vertex_llm_base.py
@@ -977,6 +977,22 @@ class TestVertexBase:
assert result_url == f"{gateway_api_base}:embedContent"
+ def test_check_custom_proxy_vertex_api_base_with_path_and_trailing_slash_appends_endpoint_once(self):
+ gateway_api_base: Final = "https://gateway.example.com/vertex/v1/projects/p/locations/us-central1/endpoints/123"
+
+ _, result_url = VertexBase()._check_custom_proxy(
+ api_base=f"{gateway_api_base}/",
+ custom_llm_provider="vertex_ai",
+ gemini_api_key=None,
+ endpoint="predict",
+ stream=None,
+ auth_header="Bearer token123",
+ url="https://us-central1-aiplatform.googleapis.com/v1/projects/p/locations/us-central1/endpoints/123:predict",
+ model="123",
+ )
+
+ assert result_url == f"{gateway_api_base}:predict"
+
def test_check_custom_proxy_vertex_api_base_with_version_path_grafts_default_path(self):
vertex_base = VertexBase()