Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

feat(providers): add DeepSeek provider (V4 Flash, V4 Pro) (#159)

ecournec764772db5e
diff --git a/.env.example b/.env.example
index 21f8fdec5..7c8b4c04f 100644
--- a/.env.example
+++ b/.env.example
@@ -12,6 +12,10 @@ ANTHROPIC_API_KEY=your_anthropic_api_key_here
# OpenAI
OPENAI_API_KEY=your_openai_api_key_here
+# DeepSeek
+DEEPSEEK_API_KEY=your_deepseek_api_key_here
+# DEEPSEEK_BASE_URL=https://api.deepseek.com
+
# Ollama — no API key needed, but set base URL if non-default
# OLLAMA_BASE_URL=http://localhost:11434
diff --git a/packages/cli/src/repowise/cli/commands/init_cmd.py b/packages/cli/src/repowise/cli/commands/init_cmd.py
index 858374287..7cdc821d1 100644
--- a/packages/cli/src/repowise/cli/commands/init_cmd.py
+++ b/packages/cli/src/repowise/cli/commands/init_cmd.py
@@ -819,7 +819,7 @@ def _workspace_init(
"--provider",
"provider_name",
default=None,
- help="LLM provider name (anthropic, openai, gemini, ollama, mock).",
+ help="LLM provider name (anthropic, openai, gemini, deepseek, ollama, litellm, mock).",
)
@click.option("--model", default=None, help="Model identifier override.")
@click.option(
diff --git a/packages/cli/src/repowise/cli/commands/mcp_cmd.py b/packages/cli/src/repowise/cli/commands/mcp_cmd.py
index a997a22a5..32ae65672 100644
--- a/packages/cli/src/repowise/cli/commands/mcp_cmd.py
+++ b/packages/cli/src/repowise/cli/commands/mcp_cmd.py
@@ -5,6 +5,7 @@ from __future__ import annotations
import click
from repowise.cli.helpers import console, resolve_repo_path
+from repowise.cli.ui import load_dotenv
@click.command("mcp")
@@ -27,6 +28,10 @@ def mcp_command(path: str | None, transport: str, port: int) -> None:
Exposes 16 tools for querying the repowise wiki via the MCP protocol.
Supports both stdio (for Claude Code, Cursor, Cline) and SSE transports.
+ Loads ``<repo>/.repowise/.env`` into the environment before starting so
+ that MCP tools (e.g. ``get_answer``) can resolve the configured LLM
+ provider and API keys.
+
Examples:
repowise mcp # stdio, current directory
@@ -34,6 +39,7 @@ def mcp_command(path: str | None, transport: str, port: int) -> None:
repowise mcp --transport sse # SSE on port 7338
"""
repo_path = resolve_repo_path(path)
+ load_dotenv(repo_path)
repowise_dir = repo_path / ".repowise"
if not repowise_dir.exists():
diff --git a/packages/cli/src/repowise/cli/helpers.py b/packages/cli/src/repowise/cli/helpers.py
index b717afe22..5da955232 100644
--- a/packages/cli/src/repowise/cli/helpers.py
+++ b/packages/cli/src/repowise/cli/helpers.py
@@ -316,6 +316,7 @@ def resolve_provider(
"anthropic": ["ANTHROPIC_BASE_URL"],
"openai": ["OPENAI_BASE_URL"],
"gemini": ["GEMINI_BASE_URL"],
+ "deepseek": ["DEEPSEEK_BASE_URL"],
"ollama": ["OLLAMA_BASE_URL"],
"litellm": ["LITELLM_BASE_URL", "LITELLM_API_BASE"],
}
@@ -357,6 +358,10 @@ def resolve_provider(
kwargs["api_key"] = os.environ.get("GEMINI_API_KEY") or os.environ.get("GOOGLE_API_KEY")
elif provider_name == "openrouter" and os.environ.get("OPENROUTER_API_KEY"):
kwargs["api_key"] = os.environ["OPENROUTER_API_KEY"]
+ elif provider_name == "deepseek" and os.environ.get("DEEPSEEK_API_KEY"):
+ kwargs["api_key"] = os.environ["DEEPSEEK_API_KEY"]
+ elif provider_name == "litellm" and os.environ.get("LITELLM_API_KEY"):
+ kwargs["api_key"] = os.environ["LITELLM_API_KEY"]
elif provider_name == "ollama" and os.environ.get("OLLAMA_BASE_URL"):
kwargs["base_url"] = os.environ["OLLAMA_BASE_URL"]
@@ -406,10 +411,20 @@ def resolve_provider(
if base_url:
kwargs["base_url"] = base_url
return get_provider("gemini", **kwargs)
+ if os.environ.get("DEEPSEEK_API_KEY") and os.environ["DEEPSEEK_API_KEY"].strip():
+ kwargs = (
+ {"model": model, "api_key": os.environ["DEEPSEEK_API_KEY"]}
+ if model
+ else {"api_key": os.environ["DEEPSEEK_API_KEY"]}
+ )
+ base_url = _resolve_base_url("deepseek")
+ if base_url:
+ kwargs["base_url"] = base_url
+ return get_provider("deepseek", **kwargs)
raise click.ClickException(
"No provider configured. Use --provider, set REPOWISE_PROVIDER, "
- "or set ANTHROPIC_API_KEY / OPENAI_API_KEY / OPENROUTER_API_KEY / OLLAMA_BASE_URL / GEMINI_API_KEY / GOOGLE_API_KEY."
+ "or set ANTHROPIC_API_KEY / OPENAI_API_KEY / OPENROUTER_API_KEY / OLLAMA_BASE_URL / GEMINI_API_KEY / GOOGLE_API_KEY / DEEPSEEK_API_KEY / LITELLM_API_KEY."
)
@@ -444,6 +459,7 @@ def validate_provider_config(provider_name: str | None = None) -> list[str]:
"anthropic": ["ANTHROPIC_API_KEY"],
"openai": ["OPENAI_API_KEY"],
"openrouter": ["OPENROUTER_API_KEY"],
+ "deepseek": ["DEEPSEEK_API_KEY"],
"gemini": ["GEMINI_API_KEY", "GOOGLE_API_KEY"], # Either one
"ollama": ["OLLAMA_BASE_URL"],
"litellm": ["LITELLM_API_KEY"], # May need others depending on backend
diff --git a/packages/cli/src/repowise/cli/ui.py b/packages/cli/src/repowise/cli/ui.py
index 0d5b22f4e..16c8a7f50 100644
--- a/packages/cli/src/repowise/cli/ui.py
+++ b/packages/cli/src/repowise/cli/ui.py
@@ -280,6 +280,7 @@ _PROVIDER_DEFAULTS: dict[str, str] = {
"gemini": "gemini-3.1-flash-lite-preview",
"openai": "gpt-4.1",
"anthropic": "claude-sonnet-4-6",
+ "deepseek": "deepseek-v4-flash",
"ollama": "llama3.2",
"openrouter": "anthropic/claude-sonnet-4.6",
"litellm": "groq/llama-3.1-70b-versatile",
@@ -289,6 +290,7 @@ _PROVIDER_ENV: dict[str, str] = {
"gemini": "GEMINI_API_KEY",
"openai": "OPENAI_API_KEY",
"anthropic": "ANTHROPIC_API_KEY",
+ "deepseek": "DEEPSEEK_API_KEY",
"ollama": "OLLAMA_BASE_URL",
"openrouter": "OPENROUTER_API_KEY",
}
@@ -297,6 +299,7 @@ _PROVIDER_SIGNUP: dict[str, str] = {
"gemini": "https://aistudio.google.com/apikey",
"openai": "https://platform.openai.com/api-keys",
"anthropic": "https://console.anthropic.com/settings/keys",
+ "deepseek": "https://platform.deepseek.com/api_keys",
"ollama": "https://ollama.com/download",
"openrouter": "https://openrouter.ai/keys",
}
@@ -308,21 +311,48 @@ _PROVIDER_SIGNUP: dict[str, str] = {
def load_dotenv(repo_path: Path) -> None:
- """Load ``<repo>/.repowise/.env`` into ``os.environ`` (without overwriting)."""
+ """Load ``<repo>/.repowise/.env`` into ``os.environ`` (without overwriting).
+
+ Supports ``export KEY=value``, quoted values (``KEY="value"``, ``KEY='value'``),
+ and inline comments (``KEY=value # comment``).
+ """
env_file = repo_path / ".repowise" / ".env"
if not env_file.exists():
return
for line in env_file.read_text(encoding="utf-8").splitlines():
line = line.strip()
- if not line or line.startswith("#") or "=" not in line:
+ if not line or line.startswith("#"):
+ continue
+ # Support `export KEY=value`
+ if line.startswith("export "):
+ line = line[len("export "):].lstrip()
+ if "=" not in line:
continue
- key, _, value = line.partition("=")
- key, value = key.strip(), value.strip()
+ key, _, raw_value = line.partition("=")
+ key = key.strip()
+ # Strip inline comments from the value (e.g. "sk-xxx # my key")
+ raw_value = raw_value.strip()
+ if "#" in raw_value:
+ # Only strip if # is preceded by whitespace (avoid stripping # in URLs)
+ hash_idx = raw_value.find(" #")
+ if hash_idx == -1:
+ hash_idx = raw_value.find("\t#")
+ if hash_idx >= 0:
+ raw_value = raw_value[:hash_idx].rstrip()
+ # Strip matching surrounding quotes
+ value = _strip_quotes(raw_value)
# Don't overwrite existing env vars (explicit env takes priority)
if key and value and key not in os.environ:
os.environ[key] = value
+def _strip_quotes(value: str) -> str:
+ """Strip one pair of matching surrounding single or double quotes."""
+ if len(value) >= 2 and value[0] == value[-1] and value[0] in ('"', "'"):
+ return value[1:-1]
+ return value
+
+
def _save_key_to_dotenv(repo_path: Path, env_var: str, value: str) -> None:
"""Append or update a key in ``<repo>/.repowise/.env``."""
env_dir = repo_path / ".repowise"
diff --git a/packages/core/src/repowise/core/generation/cost_tracker.py b/packages/core/src/repowise/core/generation/cost_tracker.py
index 542c131cb..234df4e63 100644
--- a/packages/core/src/repowise/core/generation/cost_tracker.py
+++ b/packages/core/src/repowise/core/generation/cost_tracker.py
@@ -36,6 +36,9 @@ _PRICING: dict[str, dict[str, float]] = {
# Gemini preview / experimental models
"gemini-3.1-flash-lite-preview": {"input": 0.075, "output": 0.30},
"gemini-3-flash-preview": {"input": 0.075, "output": 0.30},
+ # DeepSeek
+ "deepseek-v4-flash": {"input": 0.14, "output": 0.28},
+ "deepseek-v4-pro": {"input": 1.74, "output": 3.48},
}
_FALLBACK_PRICING: dict[str, float] = {"input": 3.0, "output": 15.0}
diff --git a/packages/core/src/repowise/core/providers/llm/__init__.py b/packages/core/src/repowise/core/providers/llm/__init__.py
index fa6aee0f8..7b1d272b5 100644
--- a/packages/core/src/repowise/core/providers/llm/__init__.py
+++ b/packages/core/src/repowise/core/providers/llm/__init__.py
@@ -13,6 +13,7 @@ Built-in providers:
openai — gpt-5.4-nano, gpt-5.4-mini, gpt-5.4
gemini — gemini-3.1-flash-lite-preview, gemini-3-flash-preview, gemini-3.1-pro-preview
openrouter — 200+ models via OpenRouter (anthropic/claude-sonnet-4.6, etc.)
+ deepseek — deepseek-v4-flash, deepseek-v4-pro via api.deepseek.com
ollama — local inference (llama3.2, codellama, etc.)
litellm — 100+ providers via LiteLLM proxy
mock — deterministic test provider
diff --git a/packages/core/src/repowise/core/providers/llm/deepseek.py b/packages/core/src/repowise/core/providers/llm/deepseek.py
new file mode 100644
index 000000000..be9095f45
--- /dev/null
+++ b/packages/core/src/repowise/core/providers/llm/deepseek.py
@@ -0,0 +1,284 @@
+"""DeepSeek provider for repowise.
+
+Access DeepSeek models (V4 Flash, V4 Pro) via the DeepSeek API at
+https://api.deepseek.com. The API is fully OpenAI-compatible — this provider
+uses the openai Python SDK with a custom base_url, following the same pattern
+as OpenRouterProvider.
+
+Models:
+ - deepseek-v4-flash — fast, economical (284B total / 13B active params) [default]
+ - deepseek-v4-pro — highest quality, full reasoning
+"""
+
+from __future__ import annotations
+
+import contextlib
+import os
+from collections.abc import AsyncIterator
+from typing import TYPE_CHECKING, Any
+
+import structlog
+from openai import APIStatusError as _OpenAIAPIStatusError
+from openai import AsyncOpenAI
+from openai import RateLimitError as _OpenAIRateLimitError
+from tenacity import (
+ RetryError,
+ retry,
+ retry_if_exception_type,
+ stop_after_attempt,
+ wait_exponential_jitter,
+)
+
+from repowise.core.providers.llm.base import (
+ BaseProvider,
+ ChatStreamEvent,
+ ChatToolCall,
+ GeneratedResponse,
+ ProviderError,
+ RateLimitError,
+)
+from repowise.core.rate_limiter import RateLimiter
+
+if TYPE_CHECKING:
+ from repowise.core.generation.cost_tracker import CostTracker
+
+log = structlog.get_logger(__name__)
+
+_MAX_RETRIES = 3
+_MIN_WAIT = 1.0
+_MAX_WAIT = 4.0
+
+_DEFAULT_BASE_URL = "https://api.deepseek.com"
+
+
+class DeepSeekProvider(BaseProvider):
+ """DeepSeek provider — access DeepSeek V4 models via OpenAI-compatible API.
+
+ Args:
+ api_key: DeepSeek API key. Falls back to DEEPSEEK_API_KEY env var.
+ model: Model identifier. Defaults to deepseek-v4-flash.
+ base_url: Override the DeepSeek API URL (rarely needed).
+ rate_limiter: Optional RateLimiter instance.
+ cost_tracker: Optional CostTracker instance for usage recording.
+ """
+
+ def __init__(
+ self,
+ api_key: str | None = None,
+ model: str = "deepseek-v4-flash",
+ base_url: str | None = None,
+ rate_limiter: RateLimiter | None = None,
+ cost_tracker: CostTracker | None = None,
+ ) -> None:
+ resolved_key = api_key or os.environ.get("DEEPSEEK_API_KEY")
+ if not resolved_key:
+ raise ProviderError(
+ "deepseek",
+ "No API key provided. Pass api_key= or set DEEPSEEK_API_KEY.",
+ )
+ resolved_base_url = base_url or os.environ.get("DEEPSEEK_BASE_URL") or _DEFAULT_BASE_URL
+ self._client = AsyncOpenAI(
+ api_key=resolved_key,
+ base_url=resolved_base_url,
+ )
+ self._model = model
+ self._rate_limiter = rate_limiter
+ self._cost_tracker = cost_tracker
+
+ @property
+ def provider_name(self) -> str:
+ return "deepseek"
+
+ @property
+ def model_name(self) -> str:
+ return self._model
+
+ async def generate(
+ self,
+ system_prompt: str,
+ user_prompt: str,
+ max_tokens: int = 4096,
+ temperature: float = 0.3,
+ request_id: str | None = None,
+ ) -> GeneratedResponse:
+ if self._rate_limiter:
+ await self._rate_limiter.acquire(estimated_tokens=max_tokens)
+
+ log.debug(
+ "deepseek.generate.start",
+ model=self._model,
+ max_tokens=max_tokens,
+ request_id=request_id,
+ )
+
+ try:
+ return await self._generate_with_retry(
+ system_prompt=system_prompt,
+ user_prompt=user_prompt,
+ max_tokens=max_tokens,
+ temperature=temperature,
+ request_id=request_id,
+ )
+ except RetryError as exc:
+ raise ProviderError(
+ "deepseek",
+ f"All {_MAX_RETRIES} retries exhausted: {exc}",
+ ) from exc
+
+ @retry(
+ retry=retry_if_exception_type(ProviderError),
+ stop=stop_after_attempt(_MAX_RETRIES),
+ wait=wait_exponential_jitter(initial=_MIN_WAIT, max=_MAX_WAIT),
+ reraise=True,
+ )
+ async def _generate_with_retry(
+ self,
+ system_prompt: str,
+ user_prompt: str,
+ max_tokens: int,
+ temperature: float,
+ request_id: str | None,
+ ) -> GeneratedResponse:
+ try:
+ response = await self._client.chat.completions.create(
+ model=self._model,
+ max_tokens=max_tokens,
+ temperature=temperature,
+ messages=[
+ {"role": "system", "content": system_prompt},
+ {"role": "user", "content": user_prompt},
+ ],
+ )
+ except _OpenAIRateLimitError as exc:
+ raise RateLimitError("deepseek", str(exc), status_code=429) from exc
+ except _OpenAIAPIStatusError as exc:
+ raise ProviderError(
+ "deepseek", str(exc), status_code=exc.status_code
+ ) from exc
+
+ usage = response.usage
+ result = GeneratedResponse(
+ content=response.choices[0].message.content or "",
+ input_tokens=usage.prompt_tokens if usage else 0,
+ output_tokens=usage.completion_tokens if usage else 0,
+ cached_tokens=0,
+ usage={
+ "prompt_tokens": usage.prompt_tokens if usage else 0,
+ "completion_tokens": usage.completion_tokens if usage else 0,
+ "total_tokens": usage.total_tokens if usage else 0,
+ },
+ )
+ log.debug(
+ "deepseek.generate.done",
+ input_tokens=result.input_tokens,
+ output_tokens=result.output_tokens,
+ request_id=request_id,
+ )
+
+ if self._cost_tracker is not None:
+ import asyncio
+
+ with contextlib.suppress(RuntimeError):
+ asyncio.get_event_loop().create_task(
+ self._cost_tracker.record(
+ model=self._model,
+ input_tokens=result.input_tokens,
+ output_tokens=result.output_tokens,
+ operation="doc_generation",
+ file_path=None,
+ )
+ )
+
+ return result
+
+ async def stream_chat(
+ self,
+ messages: list[dict[str, Any]],
+ tools: list[dict[str, Any]],
+ system_prompt: str,
+ max_tokens: int = 8192,
+ temperature: float = 0.7,
+ request_id: str | None = None,
+ tool_executor: Any | None = None,
+ ) -> AsyncIterator[ChatStreamEvent]:
+ import json as _json
+
+ full_messages = [{"role": "system", "content": system_prompt}, *messages]
+ kwargs: dict[str, Any] = {
+ "model": self._model,
+ "max_tokens": max_tokens,
+ "temperature": temperature,
+ "messages": full_messages,
+ "stream": True,
+ }
+ if tools:
+ kwargs["tools"] = tools
+
+ try:
+ stream = await self._client.chat.completions.create(**kwargs)
+ except _OpenAIRateLimitError as exc:
+ raise RateLimitError("deepseek", str(exc), status_code=429) from exc
+ except _OpenAIAPIStatusError as exc:
+ raise ProviderError("deepseek", str(exc), status_code=exc.status_code) from exc
+
+ tool_calls_acc: dict[int, dict[str, Any]] = {}
+
+ try:
+ async for chunk in stream:
+ choice = chunk.choices[0] if chunk.choices else None
+ if not choice:
+ if chunk.usage:
+ yield ChatStreamEvent(
+ type="usage",
+ input_tokens=chunk.usage.prompt_tokens or 0,
+ output_tokens=chunk.usage.completion_tokens or 0,
+ )
+ continue
+
+ delta = choice.delta
+ finish = choice.finish_reason
+
+ if delta and delta.content:
+ yield ChatStreamEvent(type="text_delta", text=delta.content)
+
+ if delta and delta.tool_calls:
+ for tc_delta in delta.tool_calls:
+ idx = tc_delta.index
+ if idx not in tool_calls_acc:
+ tool_calls_acc[idx] = {
+ "id": tc_delta.id or "",
+ "name": "",
+ "arguments": "",
+ }
+ acc = tool_calls_acc[idx]
+ if tc_delta.id:
+ acc["id"] = tc_delta.id
+ if tc_delta.function:
+ if tc_delta.function.name:
+ acc["name"] = tc_delta.function.name
+ if tc_delta.function.arguments:
+ acc["arguments"] += tc_delta.function.arguments
+
+ if finish:
+ for idx in sorted(tool_calls_acc.keys()):
+ acc = tool_calls_acc[idx]
+ try:
+ args = _json.loads(acc["arguments"]) if acc["arguments"] else {}
+ except Exception:
+ args = {}
+ yield ChatStreamEvent(
+ type="tool_start",
+ tool_call=ChatToolCall(
+ id=acc["id"],
+ name=acc["name"],
+ arguments=args,
+ ),
+ )
+ tool_calls_acc.clear()
+
+ stop_reason = "tool_use" if finish == "tool_calls" else "end_turn"
+ yield ChatStreamEvent(type="stop", stop_reason=stop_reason)
+ except _OpenAIRateLimitError as exc:
+ raise RateLimitError("deepseek", str(exc), status_code=429) from exc
+ except _OpenAIAPIStatusError as exc:
+ raise ProviderError("deepseek", str(exc), status_code=exc.status_code) from exc
diff --git a/packages/core/src/repowise/core/providers/llm/registry.py b/packages/core/src/repowise/core/providers/llm/registry.py
index 46c3612ef..ac9603f13 100644
--- a/packages/core/src/repowise/core/providers/llm/registry.py
+++ b/packages/core/src/repowise/core/providers/llm/registry.py
@@ -8,6 +8,7 @@ Built-in providers:
- anthropic → AnthropicProvider
- openai → OpenAIProvider
- openrouter → OpenRouterProvider
+ - deepseek → DeepSeekProvider
- ollama → OllamaProvider
- litellm → LiteLLMProvider
- mock → MockProvider (testing only)
@@ -41,6 +42,7 @@ _BUILTIN_PROVIDERS: dict[str, tuple[str, str]] = {
"gemini": ("repowise.core.providers.llm.gemini", "GeminiProvider"),
"ollama": ("repowise.core.providers.llm.ollama", "OllamaProvider"),
"litellm": ("repowise.core.providers.llm.litellm", "LiteLLMProvider"),
+ "deepseek": ("repowise.core.providers.llm.deepseek", "DeepSeekProvider"),
"mock": ("repowise.core.providers.llm.mock", "MockProvider"),
}
@@ -137,6 +139,7 @@ def get_provider(
"gemini": "google-genai",
"ollama": "openai", # ollama uses the openai package
"openrouter": "openai", # openrouter uses the openai package
+ "deepseek": "openai", # deepseek uses the openai package
"litellm": "litellm",
}
package = _missing.get(name, name)
diff --git a/packages/core/src/repowise/core/rate_limiter.py b/packages/core/src/repowise/core/rate_limiter.py
index 0e2c3065f..dfc5ba82a 100644
--- a/packages/core/src/repowise/core/rate_limiter.py
+++ b/packages/core/src/repowise/core/rate_limiter.py
@@ -50,6 +50,7 @@ PROVIDER_DEFAULTS: dict[str, RateLimitConfig] = {
# Ollama runs locally — effectively unlimited, but we cap to avoid OOM
"ollama": RateLimitConfig(requests_per_minute=1_000, tokens_per_minute=10_000_000),
"litellm": RateLimitConfig(requests_per_minute=60, tokens_per_minute=150_000),
+ "deepseek": RateLimitConfig(requests_per_minute=60, tokens_per_minute=200_000),
}
diff --git a/packages/server/src/repowise/server/mcp_server/tool_answer.py b/packages/server/src/repowise/server/mcp_server/tool_answer.py
index a38e4b6b9..a4ce5fef2 100644
--- a/packages/server/src/repowise/server/mcp_server/tool_answer.py
+++ b/packages/server/src/repowise/server/mcp_server/tool_answer.py
@@ -351,7 +351,7 @@ def _resolve_provider_for_answer(repo_path: Path | None = None):
try:
from repowise.core.providers.llm.registry import get_provider
except Exception:
- _log.debug("Provider registry import failed", exc_info=True)
+ _log.warning("Provider registry import failed", exc_info=True)
return None
persisted_name, persisted_model, env_overlay = _load_repo_provider_config(
@@ -374,7 +374,7 @@ def _resolve_provider_for_answer(repo_path: Path | None = None):
try:
return get_provider(provider_name, **kwargs)
except Exception:
- _log.debug("get_provider(%s) failed", provider_name, exc_info=True)
+ _log.warning("get_provider(%s) failed", provider_name, exc_info=True)
return None
def _resolve_base_url(provider_name: str) -> str | None:
@@ -382,6 +382,7 @@ def _resolve_provider_for_answer(repo_path: Path | None = None):
"openai": ["OPENAI_BASE_URL"],
"anthropic": ["ANTHROPIC_BASE_URL"],
"gemini": ["GEMINI_BASE_URL"],
+ "deepseek": ["DEEPSEEK_BASE_URL"],
"ollama": ["OLLAMA_BASE_URL"],
"litellm": ["LITELLM_BASE_URL", "LITELLM_API_BASE"],
}
@@ -400,6 +401,8 @@ def _resolve_provider_for_answer(repo_path: Path | None = None):
kw["api_key"] = _env("ANTHROPIC_API_KEY")
elif name == "openai" and _env("OPENAI_API_KEY"):
kw["api_key"] = _env("OPENAI_API_KEY")
+ elif name == "deepseek" and _env("DEEPSEEK_API_KEY"):
+ kw["api_key"] = _env("DEEPSEEK_API_KEY")
elif name == "gemini" and (
_env("GEMINI_API_KEY") or _env("GOOGLE_API_KEY")
):
@@ -439,6 +442,14 @@ def _resolve_provider_for_answer(repo_path: Path | None = None):
if model:
kw["model"] = model
return _try("ollama", **kw)
+ if _env("DEEPSEEK_API_KEY"):
+ kw = {"api_key": _env("DEEPSEEK_API_KEY")}
+ if model:
+ kw["model"] = model
+ base_url = _resolve_base_url("deepseek")
+ if base_url:
+ kw["base_url"] = base_url
+ return _try("deepseek", **kw)
return None
diff --git a/packages/server/src/repowise/server/provider_config.py b/packages/server/src/repowise/server/provider_config.py
index 6e2e14748..8cdfeb0b2 100644
--- a/packages/server/src/repowise/server/provider_config.py
+++ b/packages/server/src/repowise/server/provider_config.py
@@ -60,6 +60,14 @@ PROVIDER_CATALOG: list[dict[str, Any]] = [
"env_keys": ["OPENROUTER_API_KEY"],
"requires_key": True,
},
+ {
+ "id": "deepseek",
+ "name": "DeepSeek",
+ "default_model": "deepseek-v4-flash",
+ "models": ["deepseek-v4-flash", "deepseek-v4-pro"],
+ "env_keys": ["DEEPSEEK_API_KEY"],
+ "requires_key": True,
+ },
{
"id": "ollama",
"name": "Ollama (Local)",
@@ -139,6 +147,7 @@ def _get_base_url_for_provider(provider_id: str) -> str | None:
"openai": ["OPENAI_BASE_URL"],
"gemini": ["GEMINI_BASE_URL"],
"ollama": ["OLLAMA_BASE_URL"],
+ "deepseek": ["DEEPSEEK_BASE_URL"],
"litellm": ["LITELLM_BASE_URL", "LITELLM_API_BASE"],
}
for env_var in env_map.get(provider_id, []):
diff --git a/packages/ui/src/dashboard/quick-actions.tsx b/packages/ui/src/dashboard/quick-actions.tsx
index c1046d09f..10add9104 100644
--- a/packages/ui/src/dashboard/quick-actions.tsx
+++ b/packages/ui/src/dashboard/quick-actions.tsx
@@ -27,6 +27,8 @@ const COST_TABLE_EXACT: Record<string, [number, number]> = {
"claude-opus-4-6": [0.005, 0.025],
"claude-sonnet-4-6": [0.003, 0.015],
"claude-haiku-4-5": [0.001, 0.005],
+ "deepseek-v4-flash": [0.00014, 0.00028],
+ "deepseek-v4-pro": [0.00174, 0.00348],
};
const COST_TABLE_PREFIX: [string, [number, number]][] = [
@@ -37,6 +39,7 @@ const COST_TABLE_PREFIX: [string, [number, number]][] = [
["claude-sonnet", [0.003, 0.015]],
["claude-haiku", [0.001, 0.005]],
["claude", [0.003, 0.015]],
+ ["deepseek", [0.00014, 0.00028]],
["gemini", [0.00025, 0.0015]],
["llama", [0, 0]],
["mock", [0, 0]],
diff --git a/packages/web/src/components/repos/run-config-form.tsx b/packages/web/src/components/repos/run-config-form.tsx
index 168eaf0f7..51b5efc5f 100644
--- a/packages/web/src/components/repos/run-config-form.tsx
+++ b/packages/web/src/components/repos/run-config-form.tsx
@@ -27,7 +27,7 @@ interface Props {
onChange: (v: RunConfig) => void;
}
-const PROVIDERS = ["litellm", "openai", "anthropic", "ollama", "mock"] as const;
+const PROVIDERS = ["litellm", "openai", "anthropic", "gemini", "deepseek", "ollama", "mock"] as const;
export function RunConfigForm({ value, onChange }: Props) {
// Seed from saved settings on mount
diff --git a/packages/web/src/components/settings/provider-section.tsx b/packages/web/src/components/settings/provider-section.tsx
index cd6835460..b9ce0ed86 100644
--- a/packages/web/src/components/settings/provider-section.tsx
+++ b/packages/web/src/components/settings/provider-section.tsx
@@ -14,13 +14,14 @@ import {
SelectValue,
} from "@repowise-dev/ui/ui/select";
-const PROVIDERS = ["gemini", "openai", "anthropic", "ollama", "litellm", "mock"] as const;
+const PROVIDERS = ["gemini", "openai", "anthropic", "deepseek", "ollama", "litellm", "mock"] as const;
const EMBEDDERS = ["mock", "gemini", "openai"] as const;
const MODEL_PLACEHOLDERS: Record<string, string> = {
gemini: "gemini-3.1-flash-lite-preview",
openai: "gpt-4.1",
anthropic: "claude-sonnet-4-6",
+ deepseek: "deepseek-v4-flash",
ollama: "llama3.2",
litellm: "groq/llama-3.1-70b-versatile",
mock: "mock",
@@ -31,6 +32,7 @@ const PROVIDER_ENV_VARS: Record<string, { vars: string[]; installHint: string }>
openai: { vars: ["OPENAI_API_KEY"], installHint: "pip install openai" },
anthropic: { vars: ["ANTHROPIC_API_KEY"], installHint: "pip install anthropic" },
ollama: { vars: ["OLLAMA_BASE_URL"], installHint: "https://ollama.ai" },
+ deepseek: { vars: ["DEEPSEEK_API_KEY"], installHint: "pip install openai" },
litellm: { vars: ["LITELLM_*"], installHint: "pip install litellm" },
mock: { vars: [], installHint: "No key needed" },
};
diff --git a/tests/providers/test_registry.py b/tests/providers/test_registry.py
index 0f8608532..e9aa37077 100644
--- a/tests/providers/test_registry.py
+++ b/tests/providers/test_registry.py
@@ -115,5 +115,5 @@ class TestCustomProviderRegistration:
assert received.get("api_key") == "key-123"
def test_builtin_count(self) -> None:
- """Sanity check: we have exactly 7 built-in providers."""
- assert len(_BUILTIN_PROVIDERS) == 7
+ """Sanity check: we have exactly 8 built-in providers."""
+ assert len(_BUILTIN_PROVIDERS) == 8
diff --git a/tests/unit/test_providers/test_deepseek_provider.py b/tests/unit/test_providers/test_deepseek_provider.py
new file mode 100644
index 000000000..f46748360
--- /dev/null
+++ b/tests/unit/test_providers/test_deepseek_provider.py
@@ -0,0 +1,225 @@
+"""Unit tests for DeepSeekProvider.
+
+All tests mock the AsyncOpenAI client — no real API calls are made.
+"""
+
+from __future__ import annotations
+
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+pytest.importorskip("openai", reason="openai SDK not installed")
+
+from repowise.core.providers.llm.base import ChatStreamEvent, GeneratedResponse, ProviderError, RateLimitError
+from repowise.core.providers.llm.deepseek import DeepSeekProvider
+
+
+def test_provider_name():
+ p = DeepSeekProvider(api_key="sk-test")
+ assert p.provider_name == "deepseek"
+
+
+def test_default_model_is_flash():
+ p = DeepSeekProvider(api_key="sk-test")
+ assert p.model_name == "deepseek-v4-flash"
+
+
+def test_api_key_from_env(monkeypatch):
+ monkeypatch.setenv("DEEPSEEK_API_KEY", "sk-env-test")
+ p = DeepSeekProvider()
+ assert p.provider_name == "deepseek"
+
+
+def test_missing_api_key_raises(monkeypatch):
+ monkeypatch.delenv("DEEPSEEK_API_KEY", raising=False)
+ with pytest.raises(ProviderError):
+ DeepSeekProvider()
+
+
+def test_custom_model():
+ p = DeepSeekProvider(api_key="sk-test", model="deepseek-v4-pro")
+ assert p.model_name == "deepseek-v4-pro"
+
+
+def test_custom_base_url():
+ p = DeepSeekProvider(api_key="sk-test", base_url="https://custom.deepseek.com")
+ assert p.provider_name == "deepseek"
+
+
+def _make_mock_chat_response(text: str = "# Doc\nContent.") -> MagicMock:
+ usage = MagicMock()
+ usage.prompt_tokens = 120
+ usage.completion_tokens = 60
+ usage.total_tokens = 180
+
+ choice = MagicMock()
+ choice.message.content = text
+
+ response = MagicMock()
+ response.choices = [choice]
+ response.usage = usage
+ return response
+
+
+def _make_mock_stream_chunks(text: str) -> list[MagicMock]:
+ chunks = []
+ for char in text:
+ delta = MagicMock()
+ delta.content = char
+ delta.tool_calls = None
+ choice = MagicMock()
+ choice.delta = delta
+ choice.finish_reason = None
+ chunk = MagicMock()
+ chunk.choices = [choice]
+ chunk.usage = None
+ chunks.append(chunk)
+
+ finish_delta = MagicMock()
+ finish_delta.content = None
+ finish_delta.tool_calls = None
+ finish_choice = MagicMock()
+ finish_choice.delta = finish_delta
+ finish_choice.finish_reason = "stop"
+ finish_chunk = MagicMock()
+ finish_chunk.choices = [finish_choice]
+ finish_chunk.usage = None
+ chunks.append(finish_chunk)
+
+ return chunks
+
+
+async def test_generate_returns_generated_response():
+ provider = DeepSeekProvider(api_key="sk-test")
+ mock_response = _make_mock_chat_response("Hello from DeepSeek")
+
+ with patch("openai.AsyncOpenAI") as mock_client:
+ mock_client.return_value.chat.completions.create = AsyncMock(return_value=mock_response)
+ provider._client = mock_client.return_value
+
+ result = await provider.generate(
+ system_prompt="You are a test assistant",
+ user_prompt="Say hello",
+ )
+
+ assert isinstance(result, GeneratedResponse)
+ assert result.content == "Hello from DeepSeek"
+ assert result.input_tokens == 120
+ assert result.output_tokens == 60
+
+
+async def test_generate_uses_correct_model_name():
+ provider = DeepSeekProvider(api_key="sk-test", model="deepseek-v4-flash")
+ mock_response = _make_mock_chat_response()
+
+ with patch("openai.AsyncOpenAI") as mock_client:
+ mock_client.return_value.chat.completions.create = AsyncMock(return_value=mock_response)
+ provider._client = mock_client.return_value
+
+ await provider.generate(
+ system_prompt="system",
+ user_prompt="user",
+ )
+
+ mock_client.return_value.chat.completions.create.assert_called_once()
+ kwargs = mock_client.return_value.chat.completions.create.call_args.kwargs
+ assert kwargs["model"] == "deepseek-v4-flash"
+
+
+async def test_generate_rate_limit_retry():
+ from openai import RateLimitError as _OpenAIRateLimitError
+
+ provider = DeepSeekProvider(api_key="sk-test")
+
+ with patch("openai.AsyncOpenAI") as mock_client:
+ mock_client.return_value.chat.completions.create = AsyncMock(
+ side_effect=_OpenAIRateLimitError(
+ message="Rate limited",
+ body={},
+ response=MagicMock(status_code=429),
+ )
+ )
+ provider._client = mock_client.return_value
+
+ with pytest.raises(RateLimitError):
+ await provider.generate(
+ system_prompt="system",
+ user_prompt="user",
+ )
+
+
+async def test_generate_api_error():
+ from openai import APIStatusError as _OpenAIAPIStatusError
+
+ provider = DeepSeekProvider(api_key="sk-test")
+
+ with patch("openai.AsyncOpenAI") as mock_client:
+ mock_client.return_value.chat.completions.create = AsyncMock(
+ side_effect=_OpenAIAPIStatusError(
+ message="Internal error",
+ body={},
+ response=MagicMock(status_code=500),
+ )
+ )
+ provider._client = mock_client.return_value
+
+ with pytest.raises(ProviderError) as excinfo:
+ await provider.generate(
+ system_prompt="system",
+ user_prompt="user",
+ )
+ assert excinfo.value.status_code == 500
+
+
+async def test_cost_tracker_called():
+ from repowise.core.generation.cost_tracker import CostTracker
+
+ mock_tracker = MagicMock(spec=CostTracker)
+ mock_tracker.record = AsyncMock(return_value=0.0)
+
+ provider = DeepSeekProvider(api_key="sk-test", cost_tracker=mock_tracker)
+ mock_response = _make_mock_chat_response()
+
+ with patch("openai.AsyncOpenAI") as mock_client:
+ mock_client.return_value.chat.completions.create = AsyncMock(return_value=mock_response)
+ provider._client = mock_client.return_value
+
+ await provider.generate(
+ system_prompt="system",
+ user_prompt="user",
+ )
+
+ mock_tracker.record.assert_called_once()
+ call_kwargs = mock_tracker.record.call_args.kwargs
+ assert call_kwargs["model"] == "deepseek-v4-flash"
+ assert call_kwargs["input_tokens"] == 120
+ assert call_kwargs["output_tokens"] == 60
+
+
+async def test_stream_chat_emits_text_delta_and_stop():
+ provider = DeepSeekProvider(api_key="sk-test")
+
+ async def _async_gen():
+ for chunk in _make_mock_stream_chunks("Hi"):
+ yield chunk
+
+ with patch("openai.AsyncOpenAI") as mock_client:
+ mock_client.return_value.chat.completions.create = AsyncMock(return_value=_async_gen())
+ provider._client = mock_client.return_value
+
+ events = []
+ async for event in provider.stream_chat(
+ messages=[{"role": "user", "content": "Hi"}],
+ tools=[],
+ system_prompt="You are helpful",
+ ):
+ events.append(event)
+
+ text_deltas = [e for e in events if e.type == "text_delta"]
+ stops = [e for e in events if e.type == "stop"]
+ assert len(text_deltas) == 2
+ assert text_deltas[0].text == "H"
+ assert text_deltas[1].text == "i"
+ assert len(stops) == 1
+ assert stops[0].stop_reason == "end_turn"