Skip to content

Use Claude Code, Codex, VSCode, Pi, and OpenCode (and 6 other harnesses) for free (1.3B+ free tokens) from your terminal, app, IDE, or phone, and now from the browser with native browser sessions (multi-harness + multi-model) like OpenClaw (voice supported + ToS friendly)

Python

Add Gemini Provider

Alishahryar11324c36da52f
diff --git a/.env.example b/.env.example
index cbb4e332..c0c2c8a6 100644
--- a/.env.example
+++ b/.env.example
@@ -34,6 +34,10 @@ ZAI_API_KEY=""
FIREWORKS_API_KEY=""
+# Gemini / Google AI Studio (OpenAI-compatible Chat Completions; see https://ai.google.dev/gemini-api/docs/openai)
+GEMINI_API_KEY=""
+
+
# LM Studio Config (local provider, no API key required)
LM_STUDIO_BASE_URL="http://localhost:1234/v1"
@@ -48,7 +52,7 @@ OLLAMA_BASE_URL="http://localhost:11434"
# All Claude model requests are mapped to these models, plain model is fallback
# Format: provider_type/model/name
-# Valid providers: "nvidia_nim" | "open_router" | "mistral" | "deepseek" | "kimi" | "wafer" | "lmstudio" | "llamacpp" | "ollama" | "opencode" | "opencode_go" | "zai" | "fireworks"
+# Valid providers: "nvidia_nim" | "open_router" | "mistral" | "deepseek" | "kimi" | "wafer" | "lmstudio" | "llamacpp" | "ollama" | "opencode" | "opencode_go" | "zai" | "fireworks" | "gemini"
MODEL_OPUS=
MODEL_SONNET=
MODEL_HAIKU=
@@ -70,6 +74,7 @@ FCC_SMOKE_MODEL_OPENCODE=
FCC_SMOKE_MODEL_OPENCODE_GO=
FCC_SMOKE_MODEL_ZAI=
FCC_SMOKE_MODEL_FIREWORKS=
+FCC_SMOKE_MODEL_GEMINI=
FCC_SMOKE_NIM_MODELS=
FCC_SMOKE_NIM_EXTRA_MODELS=
FCC_SMOKE_OPENROUTER_FREE_MODELS=
@@ -98,6 +103,7 @@ OPENCODE_PROXY=""
OPENCODE_GO_PROXY=""
ZAI_PROXY=""
FIREWORKS_PROXY=""
+GEMINI_PROXY=""
PROVIDER_RATE_LIMIT=1
PROVIDER_RATE_WINDOW=3
diff --git a/README.md b/README.md
index 7e5d589e..cc38a642 100644
--- a/README.md
+++ b/README.md
@@ -37,7 +37,7 @@ Free Claude Code routes Anthropic Messages API traffic from Claude Code to any p
## What You Get
- Drop-in proxy for Claude Code's Anthropic API calls.
-- Twelve provider backends: NVIDIA NIM, OpenRouter, Mistral La Plateforme, DeepSeek, Kimi, Wafer, LM Studio, llama.cpp, Ollama, OpenCode Zen, OpenCode Go, and Z.ai.
+- Fourteen provider backends: NVIDIA NIM, OpenRouter, Mistral La Plateforme, DeepSeek, Kimi, Wafer, LM Studio, llama.cpp, Ollama, OpenCode Zen, OpenCode Go, Z.ai, Fireworks AI, and Google AI Studio (Gemini).
- Per-model routing: send Opus, Sonnet, Haiku, and fallback traffic to different providers.
- Native Claude Code `/model` picker support through the proxy's `/v1/models` endpoint (Claude Code must opt in to Gateway model discovery; see [Model Picker](#model-picker)).
- Streaming, tool use, reasoning/thinking block handling, and local request optimizations.
@@ -248,7 +248,30 @@ Popular examples:
Browse models at [Z.ai](https://z.ai).
-### 13. Mix Providers By Model Tier
+### 13. [Fireworks AI](https://fireworks.ai/)
+
+Get an API key at [fireworks.ai/account/api-keys](https://fireworks.ai/account/api-keys).
+
+In the Admin UI, paste it into `FIREWORKS_API_KEY`, then set `MODEL` to a Fireworks model slug such as `fireworks/accounts/fireworks/models/llama-v3p3-70b-instruct`.
+
+Fireworks exposes an OpenAI-compatible Chat Completions API at `https://api.fireworks.ai/inference/v1`.
+
+Browse models at [fireworks.ai/models](https://fireworks.ai/models).
+
+### 14. [Google AI Studio (Gemini)](https://aistudio.google.com/)
+
+Get a Gemini API key at [Google AI Studio](https://aistudio.google.com/apikey) (see Google's [Gemini OpenAI compatibility](https://ai.google.dev/gemini-api/docs/openai) docs).
+
+In the Admin UI, paste it into `GEMINI_API_KEY`, then set `MODEL` to a Gemini model slug such as `gemini/gemini-2.5-flash` or `gemini/gemini-3.1-flash-lite`.
+
+The Gemini API exposes an OpenAI-compatible endpoint at `https://generativelanguage.googleapis.com/v1beta/openai/`. Free tier quotas are per-model; prompts may be used to improve Google's products outside the UK/CH/EEA/EU unless your account region says otherwise—see Google's terms.
+
+Popular examples:
+
+- `gemini/gemini-2.5-flash`
+- `gemini/gemini-3.1-flash-lite`
+
+### 15. Mix Providers By Model Tier
Each model tier can use a different provider by setting `MODEL_OPUS`, `MODEL_SONNET`, and `MODEL_HAIKU` in the Admin UI. Leave a tier blank to inherit `MODEL`.
diff --git a/api/admin_config.py b/api/admin_config.py
index dcb03352..16266b3f 100644
--- a/api/admin_config.py
+++ b/api/admin_config.py
@@ -199,6 +199,20 @@ FIELDS: tuple[ConfigFieldSpec, ...] = (
secret=True,
description="Fireworks AI inference API key.",
),
+ ConfigFieldSpec(
+ "GEMINI_API_KEY",
+ "Gemini API Key",
+ "providers",
+ "secret",
+ settings_attr="gemini_api_key",
+ secret=True,
+ description=(
+ "Google AI Studio Gemini API key (Google AI Studio / Gemini API "
+ "[OpenAI-compatible](https://ai.google.dev/gemini-api/docs/openai)); "
+ "free tier has per-model rate limits and data may be used for improvement "
+ "outside the UK/CH/EEA/EU."
+ ),
+ ),
ConfigFieldSpec(
"LM_STUDIO_BASE_URL",
"LM Studio Base URL",
@@ -319,6 +333,15 @@ FIELDS: tuple[ConfigFieldSpec, ...] = (
secret=True,
advanced=True,
),
+ ConfigFieldSpec(
+ "GEMINI_PROXY",
+ "Gemini Proxy",
+ "providers",
+ "secret",
+ settings_attr="gemini_proxy",
+ secret=True,
+ advanced=True,
+ ),
ConfigFieldSpec(
"MODEL",
"Default Model",
@@ -779,6 +802,18 @@ FIELDS: tuple[ConfigFieldSpec, ...] = (
"smoke",
advanced=True,
),
+ ConfigFieldSpec(
+ "FCC_SMOKE_MODEL_FIREWORKS",
+ "Smoke Fireworks Model",
+ "smoke",
+ advanced=True,
+ ),
+ ConfigFieldSpec(
+ "FCC_SMOKE_MODEL_GEMINI",
+ "Smoke Gemini Model",
+ "smoke",
+ advanced=True,
+ ),
ConfigFieldSpec(
"FCC_SMOKE_NIM_MODELS",
"Smoke NIM Models",
diff --git a/config/provider_catalog.py b/config/provider_catalog.py
index e6e6084f..e5a06069 100644
--- a/config/provider_catalog.py
+++ b/config/provider_catalog.py
@@ -28,6 +28,8 @@ OLLAMA_DEFAULT_BASE = "http://localhost:11434"
OPENCODE_DEFAULT_BASE = "https://opencode.ai/zen/v1"
OPENCODE_GO_DEFAULT_BASE = "https://opencode.ai/zen/go/v1"
ZAI_DEFAULT_BASE = "https://api.z.ai/api/coding/paas/v4"
+# Google AI Studio Gemini API OpenAI-compat layer (not Vertex AI).
+GEMINI_DEFAULT_BASE = "https://generativelanguage.googleapis.com/v1beta/openai/"
@dataclass(frozen=True, slots=True)
@@ -178,6 +180,16 @@ PROVIDER_CATALOG: dict[str, ProviderDescriptor] = {
proxy_attr="fireworks_proxy",
capabilities=("chat", "streaming", "tools", "thinking", "rate_limit"),
),
+ "gemini": ProviderDescriptor(
+ provider_id="gemini",
+ transport_type="openai_chat",
+ credential_env="GEMINI_API_KEY",
+ credential_url="https://aistudio.google.com/apikey",
+ credential_attr="gemini_api_key",
+ default_base_url=GEMINI_DEFAULT_BASE,
+ proxy_attr="gemini_proxy",
+ capabilities=("chat", "streaming", "tools", "thinking", "rate_limit"),
+ ),
}
# Order matches docs / historical error text; must match PROVIDER_CATALOG keys.
diff --git a/config/settings.py b/config/settings.py
index ed070757..44a36699 100644
--- a/config/settings.py
+++ b/config/settings.py
@@ -132,6 +132,9 @@ class Settings(BaseSettings):
# ==================== Fireworks AI Config ====================
fireworks_api_key: str = Field(default="", validation_alias="FIREWORKS_API_KEY")
+ # ==================== Google Gemini (Google AI Studio) ====================
+ gemini_api_key: str = Field(default="", validation_alias="GEMINI_API_KEY")
+
# ==================== Messaging Platform Selection ====================
# Valid: "telegram" | "discord" | "none"
messaging_platform: str = Field(
@@ -188,6 +191,7 @@ class Settings(BaseSettings):
opencode_go_proxy: str = Field(default="", validation_alias="OPENCODE_GO_PROXY")
zai_proxy: str = Field(default="", validation_alias="ZAI_PROXY")
fireworks_proxy: str = Field(default="", validation_alias="FIREWORKS_PROXY")
+ gemini_proxy: str = Field(default="", validation_alias="GEMINI_PROXY")
# ==================== Provider Rate Limiting ====================
provider_rate_limit: int = Field(default=40, validation_alias="PROVIDER_RATE_LIMIT")
diff --git a/providers/defaults.py b/providers/defaults.py
index baed6ecc..7bc76353 100644
--- a/providers/defaults.py
+++ b/providers/defaults.py
@@ -3,6 +3,7 @@
from config.provider_catalog import (
DEEPSEEK_ANTHROPIC_DEFAULT_BASE,
DEEPSEEK_DEFAULT_BASE,
+ GEMINI_DEFAULT_BASE,
KIMI_DEFAULT_BASE,
LLAMACPP_DEFAULT_BASE,
LMSTUDIO_DEFAULT_BASE,
@@ -19,6 +20,7 @@ from config.provider_catalog import (
__all__ = (
"DEEPSEEK_ANTHROPIC_DEFAULT_BASE",
"DEEPSEEK_DEFAULT_BASE",
+ "GEMINI_DEFAULT_BASE",
"KIMI_DEFAULT_BASE",
"LLAMACPP_DEFAULT_BASE",
"LMSTUDIO_DEFAULT_BASE",
diff --git a/providers/gemini/__init__.py b/providers/gemini/__init__.py
new file mode 100644
index 00000000..2321819b
--- /dev/null
+++ b/providers/gemini/__init__.py
@@ -0,0 +1,7 @@
+"""Google AI Studio Gemini (OpenAI-compat) adapter."""
+
+from providers.defaults import GEMINI_DEFAULT_BASE
+
+from .client import GeminiProvider
+
+__all__ = ["GEMINI_DEFAULT_BASE", "GeminiProvider"]
diff --git a/providers/gemini/client.py b/providers/gemini/client.py
new file mode 100644
index 00000000..3f682563
--- /dev/null
+++ b/providers/gemini/client.py
@@ -0,0 +1,31 @@
+"""Google AI Studio Gemini provider (OpenAI-compatible chat completions)."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from providers.base import ProviderConfig
+from providers.defaults import GEMINI_DEFAULT_BASE
+from providers.openai_compat import OpenAIChatTransport
+
+from .request import build_request_body
+
+
+class GeminiProvider(OpenAIChatTransport):
+ """Gemini API using ``https://generativelanguage.googleapis.com/v1beta/openai/``."""
+
+ def __init__(self, config: ProviderConfig):
+ super().__init__(
+ config,
+ provider_name="GEMINI",
+ base_url=config.base_url or GEMINI_DEFAULT_BASE,
+ api_key=config.api_key,
+ )
+
+ def _build_request_body(
+ self, request: Any, thinking_enabled: bool | None = None
+ ) -> dict:
+ return build_request_body(
+ request,
+ thinking_enabled=self._is_thinking_enabled(request, thinking_enabled),
+ )
diff --git a/providers/gemini/request.py b/providers/gemini/request.py
new file mode 100644
index 00000000..129ed642
--- /dev/null
+++ b/providers/gemini/request.py
@@ -0,0 +1,55 @@
+"""Request builder for Google Gemini API (AI Studio OpenAI-compatible chat completions)."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from loguru import logger
+
+from core.anthropic import ReasoningReplayMode, build_base_request_body
+from core.anthropic.conversion import OpenAIConversionError
+from providers.exceptions import InvalidRequestError
+
+
+def build_request_body(request_data: Any, *, thinking_enabled: bool) -> dict:
+ """Build OpenAI-format request body from an Anthropic request for Gemini."""
+ logger.debug(
+ "GEMINI_REQUEST: conversion start model={} msgs={}",
+ getattr(request_data, "model", "?"),
+ len(getattr(request_data, "messages", [])),
+ )
+ try:
+ body = build_base_request_body(
+ request_data,
+ reasoning_replay=ReasoningReplayMode.REASONING_CONTENT
+ if thinking_enabled
+ else ReasoningReplayMode.DISABLED,
+ )
+ except OpenAIConversionError as exc:
+ raise InvalidRequestError(str(exc)) from exc
+
+ extra_body: dict[str, Any] = {}
+ request_extra = getattr(request_data, "extra_body", None)
+ if request_extra:
+ extra_body.update(request_extra)
+
+ if thinking_enabled:
+ body["reasoning_effort"] = "high"
+ google_section = extra_body.setdefault("google", {})
+ if isinstance(google_section, dict):
+ thinking_cfg = google_section.setdefault("thinking_config", {})
+ if isinstance(thinking_cfg, dict):
+ thinking_cfg.setdefault("include_thoughts", True)
+ else:
+ body["reasoning_effort"] = "none"
+
+ if extra_body:
+ body["extra_body"] = extra_body
+
+ logger.debug(
+ "GEMINI_REQUEST: conversion done model={} msgs={} tools={}",
+ body.get("model"),
+ len(body.get("messages", [])),
+ len(body.get("tools", [])),
+ )
+ return body
diff --git a/providers/registry.py b/providers/registry.py
index cf06b3a7..466e3558 100644
--- a/providers/registry.py
+++ b/providers/registry.py
@@ -110,6 +110,12 @@ def _create_fireworks(config: ProviderConfig, _settings: Settings) -> BaseProvid
return FireworksProvider(config)
+def _create_gemini(config: ProviderConfig, _settings: Settings) -> BaseProvider:
+ from providers.gemini import GeminiProvider
+
+ return GeminiProvider(config)
+
+
PROVIDER_FACTORIES: dict[str, ProviderFactory] = {
"nvidia_nim": _create_nvidia_nim,
"open_router": _create_open_router,
@@ -124,6 +130,7 @@ PROVIDER_FACTORIES: dict[str, ProviderFactory] = {
"opencode_go": _create_opencode_go,
"zai": _create_zai,
"fireworks": _create_fireworks,
+ "gemini": _create_gemini,
}
if set(PROVIDER_DESCRIPTORS) != set(SUPPORTED_PROVIDER_IDS) or set(
diff --git a/smoke/README.md b/smoke/README.md
index 87e22fd8..857ad5bf 100644
--- a/smoke/README.md
+++ b/smoke/README.md
@@ -118,7 +118,10 @@ uv run pytest smoke/product -n 0 -s --tb=short
- `FCC_SMOKE_TARGETS`: comma-separated targets, or `all`.
- `FCC_SMOKE_PROVIDER_MATRIX`: comma-separated provider prefixes to require.
- `FCC_SMOKE_MODEL_NVIDIA_NIM`, `FCC_SMOKE_MODEL_OPEN_ROUTER`,
- `FCC_SMOKE_MODEL_MISTRAL`, `FCC_SMOKE_MODEL_DEEPSEEK`, `FCC_SMOKE_MODEL_LMSTUDIO`,
+ `FCC_SMOKE_MODEL_MISTRAL`, `FCC_SMOKE_MODEL_DEEPSEEK`, `FCC_SMOKE_MODEL_KIMI`,
+ `FCC_SMOKE_MODEL_WAFER`, `FCC_SMOKE_MODEL_OPENCODE`, `FCC_SMOKE_MODEL_OPENCODE_GO`,
+ `FCC_SMOKE_MODEL_ZAI`, `FCC_SMOKE_MODEL_FIREWORKS`, `FCC_SMOKE_MODEL_GEMINI`,
+ `FCC_SMOKE_MODEL_LMSTUDIO`,
`FCC_SMOKE_MODEL_LLAMACPP`, `FCC_SMOKE_MODEL_OLLAMA`: optional per-provider
smoke model overrides. Values may include the provider prefix or just the model
name for that provider.
diff --git a/smoke/lib/config.py b/smoke/lib/config.py
index e47a10d7..de1932dd 100644
--- a/smoke/lib/config.py
+++ b/smoke/lib/config.py
@@ -53,6 +53,7 @@ PROVIDER_SMOKE_DEFAULT_MODELS: dict[str, str] = {
"opencode": "opencode/gpt-5.3-codex",
"opencode_go": "opencode_go/minimax-m2.7",
"zai": "zai/glm-5.1",
+ "gemini": "gemini/gemini-2.5-flash",
}
NVIDIA_NIM_CLI_DEFAULT_MODELS: tuple[str, ...] = (
@@ -229,6 +230,8 @@ class SmokeConfig:
return bool(self.settings.mistral_api_key.strip())
if provider == "deepseek":
return bool(self.settings.deepseek_api_key.strip())
+ if provider == "kimi":
+ return bool(self.settings.kimi_api_key.strip())
if provider == "lmstudio":
return bool(self.settings.lm_studio_base_url.strip())
if provider == "llamacpp":
@@ -237,10 +240,16 @@ class SmokeConfig:
return bool(self.settings.ollama_base_url.strip())
if provider == "wafer":
return bool(self.settings.wafer_api_key.strip())
+ if provider == "fireworks":
+ return bool(self.settings.fireworks_api_key.strip())
if provider == "opencode":
return bool(self.settings.opencode_api_key.strip())
+ if provider == "opencode_go":
+ return bool(self.settings.opencode_api_key.strip())
if provider == "zai":
return bool(self.settings.zai_api_key.strip())
+ if provider == "gemini":
+ return bool(self.settings.gemini_api_key.strip())
return False
diff --git a/tests/api/test_admin.py b/tests/api/test_admin.py
index 21c41e0b..2e67d6e7 100644
--- a/tests/api/test_admin.py
+++ b/tests/api/test_admin.py
@@ -103,6 +103,7 @@ def test_admin_config_masks_secrets_and_exposes_manifest(monkeypatch, tmp_path):
assert "ANTHROPIC_AUTH_TOKEN" in keys
assert "OPENROUTER_API_KEY" in keys
assert "FIREWORKS_API_KEY" in keys
+ assert "GEMINI_API_KEY" in keys
assert "ZAI_BASE_URL" not in keys
assert "CLAUDE_WORKSPACE" not in keys
assert "CLAUDE_CLI_BIN" not in keys
@@ -207,6 +208,31 @@ def test_admin_apply_writes_fireworks_key_and_masks_preview(monkeypatch, tmp_pat
assert "FIREWORKS_API_KEY=fw-secret" in text
+def test_admin_apply_writes_gemini_key_and_masks_preview(monkeypatch, tmp_path):
+ _set_home(monkeypatch, tmp_path)
+ _clear_process_config(monkeypatch)
+ app = create_app(lifespan_enabled=False)
+
+ response = _local_client(app).post(
+ "/admin/api/config/apply",
+ json={
+ "values": {
+ "MODEL": "gemini/gemini-2.5-flash",
+ "GEMINI_API_KEY": "gm-secret",
+ }
+ },
+ )
+
+ assert response.status_code == 200
+ body = response.json()
+ assert body["applied"] is True
+ assert "GEMINI_API_KEY=********" in body["env_preview"]
+ env_file = tmp_path / ".fcc" / ".env"
+ text = env_file.read_text(encoding="utf-8")
+ assert "MODEL=gemini/gemini-2.5-flash" in text
+ assert "GEMINI_API_KEY=gm-secret" in text
+
+
def test_admin_apply_preserves_hidden_diagnostics_and_smoke_values(
monkeypatch, tmp_path
):
diff --git a/tests/api/test_dependencies.py b/tests/api/test_dependencies.py
index 240803a8..6cc3bd10 100644
--- a/tests/api/test_dependencies.py
+++ b/tests/api/test_dependencies.py
@@ -17,6 +17,7 @@ from api.dependencies import (
from config.nim import NimSettings
from providers.deepseek import DeepSeekProvider
from providers.exceptions import ServiceUnavailableError, UnknownProviderTypeError
+from providers.gemini import GeminiProvider
from providers.lmstudio import LMStudioProvider
from providers.mistral import MistralProvider
from providers.nvidia_nim import NvidiaNimProvider
@@ -54,6 +55,10 @@ def _make_mock_settings(**overrides):
mock.opencode_proxy = ""
mock.opencode_go_proxy = ""
mock.zai_proxy = ""
+ mock.fireworks_api_key = ""
+ mock.fireworks_proxy = ""
+ mock.gemini_api_key = ""
+ mock.gemini_proxy = ""
mock.nim = NimSettings()
mock.http_read_timeout = 300.0
mock.http_write_timeout = 10.0
@@ -218,6 +223,41 @@ async def test_get_provider_mistral():
assert provider._api_key == "test_mistral_key"
+async def test_get_provider_gemini():
+ """Test that provider_type=gemini returns GeminiProvider."""
+ with patch("api.dependencies.get_settings") as mock_settings:
+ mock_settings.return_value = _make_mock_settings(
+ provider_type="gemini",
+ gemini_api_key="secret",
+ )
+
+ provider = get_provider()
+
+ assert isinstance(provider, GeminiProvider)
+ assert provider._base_url == (
+ "https://generativelanguage.googleapis.com/v1beta/openai"
+ )
+ assert provider._api_key == "secret"
+
+
+async def test_get_provider_gemini_missing_api_key():
+ """Gemini with empty API key raises HTTPException 503."""
+ with patch("api.dependencies.get_settings") as mock_settings:
+ mock_settings.return_value = _make_mock_settings(
+ provider_type="gemini",
+ gemini_api_key="",
+ )
+
+ with pytest.raises(HTTPException) as exc_info:
+ get_provider()
+
+ assert exc_info.value.status_code == 503
+ assert "GEMINI_API_KEY" in exc_info.value.detail
+ assert "aistudio.google.com" in exc_info.value.detail
+
+
@pytest.mark.asyncio
async def test_get_provider_wafer():
"""Test that provider_type=wafer returns WaferProvider."""
diff --git a/tests/config/test_config.py b/tests/config/test_config.py
index e41c8689..72a302d8 100644
--- a/tests/config/test_config.py
+++ b/tests/config/test_config.py
@@ -766,6 +766,7 @@ class TestPerModelMapping:
assert Settings.parse_provider_type("llamacpp/model") == "llamacpp"
assert Settings.parse_provider_type("ollama/llama3.1") == "ollama"
assert Settings.parse_provider_type("wafer/DeepSeek-V4-Pro") == "wafer"
+ assert Settings.parse_provider_type("gemini/gemini-2.5-flash") == "gemini"
def test_parse_model_name(self):
"""parse_model_name extracts model name from model string."""
@@ -781,6 +782,9 @@ class TestPerModelMapping:
assert Settings.parse_model_name("llamacpp/model") == "model"
assert Settings.parse_model_name("ollama/llama3.1") == "llama3.1"
assert Settings.parse_model_name("wafer/DeepSeek-V4-Pro") == "DeepSeek-V4-Pro"
+ assert (
+ Settings.parse_model_name("gemini/gemini-2.5-flash") == "gemini-2.5-flash"
+ )
def test_configured_chat_model_refs_collects_unique_models_with_sources(
self, monkeypatch
diff --git a/tests/contracts/test_feature_manifest.py b/tests/contracts/test_feature_manifest.py
index 49192199..82486924 100644
--- a/tests/contracts/test_feature_manifest.py
+++ b/tests/contracts/test_feature_manifest.py
@@ -6,6 +6,7 @@ from pathlib import Path
from messaging.platforms.factory import create_messaging_platform
from providers.base import BaseProvider
from providers.deepseek import DeepSeekProvider
+from providers.gemini import GeminiProvider
from providers.llamacpp import LlamaCppProvider
from providers.lmstudio import LMStudioProvider
from providers.mistral import MistralProvider
@@ -82,6 +83,7 @@ def test_provider_and_platform_registries_include_advertised_builtins() -> None:
"opencode": OpenCodeProvider,
"opencode_go": OpenCodeProvider,
"zai": ZaiProvider,
+ "gemini": GeminiProvider,
}
for provider_class in provider_classes.values():
assert issubclass(provider_class, BaseProvider)
diff --git a/tests/contracts/test_smoke_config.py b/tests/contracts/test_smoke_config.py
index 8bbfce67..41235439 100644
--- a/tests/contracts/test_smoke_config.py
+++ b/tests/contracts/test_smoke_config.py
@@ -27,9 +27,12 @@ def _settings(**overrides):
"open_router_api_key": "",
"mistral_api_key": "",
"deepseek_api_key": "",
+ "kimi_api_key": "",
"wafer_api_key": "",
"opencode_api_key": "",
"zai_api_key": "",
+ "gemini_api_key": "",
+ "fireworks_api_key": "",
"lm_studio_base_url": "",
"llamacpp_base_url": "",
"ollama_base_url": "http://localhost:11434",
diff --git a/tests/providers/test_gemini.py b/tests/providers/test_gemini.py
new file mode 100644
index 00000000..fd085926
--- /dev/null
+++ b/tests/providers/test_gemini.py
@@ -0,0 +1,203 @@
+"""Tests for Google AI Studio Gemini (OpenAI-compatible) provider."""
+
+from contextlib import asynccontextmanager
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+from providers.base import ProviderConfig
+from providers.gemini import GEMINI_DEFAULT_BASE, GeminiProvider
+
+
+class MockMessage:
+ def __init__(self, role, content):
+ self.role = role
+ self.content = content
+
+
+class MockRequest:
+ def __init__(self, **kwargs):
+ self.model = "gemini-2.5-flash"
+ self.messages = [MockMessage("user", "Hello")]
+ self.max_tokens = 100
+ self.temperature = 0.5
+ self.top_p = 0.9
+ self.system = "System prompt"
+ self.stop_sequences = None
+ self.tools = []
+ self.thinking = MagicMock()
+ self.thinking.enabled = True
+ for key, value in kwargs.items():
+ setattr(self, key, value)
+
+
+def gemini_config():
+ return ProviderConfig(
+ api_key="test_gemini_key",
+ base_url=GEMINI_DEFAULT_BASE,
+ rate_limit=10,
+ rate_window=60,
+ enable_thinking=True,
+ )
+
+
[email protected](autouse=True)
+def mock_rate_limiter():
+ """Mock the global rate limiter to prevent waiting."""
+
+ @asynccontextmanager
+ async def _slot():
+ yield
+
+ with patch("providers.openai_compat.GlobalRateLimiter") as mock:
+ instance = mock.get_scoped_instance.return_value
+
+ async def _passthrough(fn, *args, **kwargs):
+ return await fn(*args, **kwargs)
+
+ instance.execute_with_retry = AsyncMock(side_effect=_passthrough)
+ instance.concurrency_slot.side_effect = _slot
+ yield instance
+
+
+def gemini_provider(gemini_config):
+ return GeminiProvider(gemini_config)
+
+
+def test_init(gemini_config):
+ """Test provider initialization."""
+ with patch("providers.openai_compat.AsyncOpenAI") as mock_openai:
+ provider = GeminiProvider(gemini_config)
+ assert provider._api_key == "test_gemini_key"
+ assert (
+ provider._base_url
+ == "https://generativelanguage.googleapis.com/v1beta/openai"
+ )
+ mock_openai.assert_called_once()
+
+
+def test_default_base_url_constant():
+ assert GEMINI_DEFAULT_BASE == (
+ "https://generativelanguage.googleapis.com/v1beta/openai/"
+ )
+
+
+def test_build_request_body_basic(gemini_provider):
+ """Basic body conversion attaches Gemini thinking fields when thinking is on."""
+ req = MockRequest()
+ body = gemini_provider._build_request_body(req)
+
+ assert body["model"] == "gemini-2.5-flash"
+ assert body["messages"][0]["role"] == "system"
+ assert body["reasoning_effort"] == "high"
+ eb = body.get("extra_body")
+ assert isinstance(eb, dict)
+ gc = eb.get("google")
+ assert isinstance(gc, dict)
+ tc = gc.get("thinking_config")
+ assert isinstance(tc, dict)
+ assert tc.get("include_thoughts") is True
+
+
+def test_build_request_body_global_disable_sets_reasoning_none():
+ """When thinking is off, Gemini uses reasoning_effort none (Gemini 2.5 convention)."""
+ provider = GeminiProvider(
+ ProviderConfig(
+ api_key="test_gemini_key",
+ base_url=GEMINI_DEFAULT_BASE,
+ rate_limit=10,
+ rate_window=60,
+ enable_thinking=False,
+ )
+ )
+ req = MockRequest()
+ body = provider._build_request_body(req)
+
+ assert body["reasoning_effort"] == "none"
+ roles = [m.get("role") for m in body.get("messages", [])]
+ assert "assistant_reasoning_content" not in roles
+
+
+def test_build_request_body_preserves_caller_extra_body(gemini_provider):
+ req = MockRequest(extra_body={"metadata": {"user": "u1"}})
+
+ body = gemini_provider._build_request_body(req)
+
+ eb = body.get("extra_body")
+ assert isinstance(eb, dict)
+ assert eb.get("metadata") == {"user": "u1"}
+
+
+async def test_stream_response_text(gemini_provider):
+ req = MockRequest()
+
+ mock_chunk = MagicMock()
+ mock_chunk.choices = [
+ MagicMock(
+ delta=MagicMock(
+ content="Hello back!",
+ reasoning_content=None,
+ tool_calls=None,
+ ),
+ finish_reason="stop",
+ )
+ ]
+ mock_chunk.usage = MagicMock(completion_tokens=5, prompt_tokens=10)
+
+ async def mock_stream():
+ yield mock_chunk
+
+ with patch.object(
+ gemini_provider._client.chat.completions, "create", new_callable=AsyncMock
+ ) as mock_create:
+ mock_create.return_value = mock_stream()
+
+ events = [event async for event in gemini_provider.stream_response(req)]
+
+ assert any(
+ '"text_delta"' in event and "Hello back!" in event for event in events
+ )
+
+
+async def test_stream_response_reasoning_content(gemini_provider):
+ req = MockRequest()
+
+ mock_chunk = MagicMock()
+ mock_chunk.choices = [
+ MagicMock(
+ delta=MagicMock(
+ content=None,
+ reasoning_content="Thinking...",
+ tool_calls=None,
+ ),
+ finish_reason="stop",
+ )
+ ]
+ mock_chunk.usage = MagicMock(completion_tokens=2, prompt_tokens=10)
+
+ async def mock_stream():
+ yield mock_chunk
+
+ with patch.object(
+ gemini_provider._client.chat.completions, "create", new_callable=AsyncMock
+ ) as mock_create:
+ mock_create.return_value = mock_stream()
+
+ events = [event async for event in gemini_provider.stream_response(req)]
+
+ assert any(
+ '"thinking_delta"' in event and "Thinking..." in event for event in events
+ )
+
+
+async def test_cleanup(gemini_provider):
+ gemini_provider._client = AsyncMock()
+
+ await gemini_provider.cleanup()
+
+ gemini_provider._client.close.assert_called_once()
diff --git a/tests/providers/test_registry.py b/tests/providers/test_registry.py
index 7a998ef9..0c1374d8 100644
--- a/tests/providers/test_registry.py
+++ b/tests/providers/test_registry.py
@@ -9,6 +9,7 @@ from config.provider_catalog import PROVIDER_CATALOG, ZAI_DEFAULT_BASE
from config.provider_ids import SUPPORTED_PROVIDER_IDS
from providers.deepseek import DeepSeekProvider
from providers.exceptions import UnknownProviderTypeError
+from providers.gemini import GeminiProvider
from providers.llamacpp import LlamaCppProvider
from providers.lmstudio import LMStudioProvider
from providers.mistral import MistralProvider
@@ -50,6 +51,10 @@ def _make_settings(**overrides):
mock.opencode_proxy = ""
mock.opencode_go_proxy = ""
mock.zai_proxy = ""
+ mock.fireworks_proxy = ""
+ mock.fireworks_api_key = ""
+ mock.gemini_api_key = ""
+ mock.gemini_proxy = ""
mock.provider_rate_limit = 40
mock.provider_rate_window = 60
mock.provider_max_concurrency = 5
@@ -147,7 +152,7 @@ def test_create_provider_uses_native_openrouter_by_default():
def test_create_provider_instantiates_each_builtin():
- settings = _make_settings()
+ settings = _make_settings(gemini_api_key="test_gemini_key")
cases = {
"nvidia_nim": NvidiaNimProvider,
"mistral": MistralProvider,
@@ -159,6 +164,7 @@ def test_create_provider_instantiates_each_builtin():
"opencode": OpenCodeProvider,
"opencode_go": OpenCodeProvider,
"zai": ZaiProvider,
+ "gemini": GeminiProvider,
}
with (