mirrors/archon
1
0
Fork 0
mirror of https://github.com/coleam00/Archon.git synced 2025-12-24 02:39:17 -05:00
Code Issues Packages Projects Releases Wiki Activity

Add OpenRouter embeddings support

Browse Source 
Implements OpenRouter as an embedding provider option, enabling access to multiple
embedding models (OpenAI, Google Gemini, Qwen3, Mistral) through a single API key.

Backend changes:
- Add validate_openrouter_api_key() for API key validation (sk-or-v1- format)
- Add OpenRouterErrorAdapter for error sanitization
- Add openrouter to valid providers in llm_provider_service
- Create openrouter_discovery_service with hardcoded model list
- Create /api/openrouter/models endpoint for model discovery
- Register OpenRouter router in FastAPI main app

Frontend changes:
- Create openrouterService.ts for model discovery API client
- Add OpenRouter to RAGSettings.tsx provider options
- Configure default models with provider prefix (openai/text-embedding-3-small)
- Add OpenRouter to embedding-capable providers list

Documentation:
- Update .env.example with OPENROUTER_API_KEY documentation

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
leex279
2025-11-08 23:06:31 +01:00
parent 4ebdeda2f5
commit 44bbcfcafb
10 changed files with 1185 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
python/src/server/api_routes/openrouter_api.py Normal file
View File

@@ -0,0 +1,27 @@
"""
OpenRouter API routes.
Endpoints for OpenRouter model discovery and configuration.
"""
from fastapi import APIRouter
from ..services.openrouter_discovery_service import OpenRouterModelListResponse, openrouter_discovery_service
router = APIRouter(prefix="/api/openrouter", tags=["openrouter"])
@router.get("/models", response_model=OpenRouterModelListResponse)
async def get_openrouter_models() -> OpenRouterModelListResponse:
"""
Get available OpenRouter embedding models.
Returns a list of embedding models available through OpenRouter,
including models from OpenAI, Google, Qwen, and Mistral providers.
Returns:
OpenRouterModelListResponse: List of embedding models with metadata
"""
models = await openrouter_discovery_service.discover_embedding_models()
return OpenRouterModelListResponse(embedding_models=models, total_count=len(models))

13
python/src/server/config/config.py
View File

@@ -66,6 +66,19 @@ def validate_openai_api_key(api_key: str) -> bool:
return True
def validate_openrouter_api_key(api_key: str) -> bool:
"""Validate OpenRouter API key format."""
if not api_key:
raise ConfigurationError("OpenRouter API key cannot be empty")
if not api_key.startswith("sk-or-v1-"):
raise ConfigurationError(
"OpenRouter API key must start with 'sk-or-v1-'. " "Get your key at https://openrouter.ai/keys"
)
return True
def validate_supabase_key(supabase_key: str) -> tuple[bool, str]:
"""Validate Supabase key type and return validation result.

2
python/src/server/main.py
View File

@@ -26,6 +26,7 @@ from .api_routes.knowledge_api import router as knowledge_router
from .api_routes.mcp_api import router as mcp_router
from .api_routes.migration_api import router as migration_router
from .api_routes.ollama_api import router as ollama_router
from .api_routes.openrouter_api import router as openrouter_router
from .api_routes.pages_api import router as pages_router
from .api_routes.progress_api import router as progress_router
from .api_routes.projects_api import router as projects_router
@@ -187,6 +188,7 @@ app.include_router(mcp_router)
app.include_router(knowledge_router)
app.include_router(pages_router)
app.include_router(ollama_router)
app.include_router(openrouter_router)
app.include_router(projects_router)
app.include_router(progress_router)
app.include_router(agent_chat_router)

64
python/src/server/services/embeddings/provider_error_adapters.py
View File

@@ -8,13 +8,6 @@ with unified error handling and sanitization patterns.
import re
from abc import ABC, abstractmethod
from .embedding_exceptions import (
EmbeddingAPIError,
EmbeddingAuthenticationError,
EmbeddingQuotaExhaustedError,
EmbeddingRateLimitError,
)
class ProviderErrorAdapter(ABC):
"""Abstract base class for provider-specific error handling."""
@@ -37,7 +30,7 @@ class OpenAIErrorAdapter(ProviderErrorAdapter):
return "OpenAI API encountered an error. Please verify your API key and quota."
sanitized = message
# Comprehensive OpenAI patterns with case-insensitive matching
patterns = [
(r'sk-[a-zA-Z0-9]{48}', '[REDACTED_KEY]'), # OpenAI API keys
@@ -68,7 +61,7 @@ class GoogleAIErrorAdapter(ProviderErrorAdapter):
return "Google AI API encountered an error. Please verify your API key."
sanitized = message
# Comprehensive Google AI patterns
patterns = [
(r'AIza[a-zA-Z0-9_-]{35}', '[REDACTED_KEY]'), # Google AI API keys
@@ -99,7 +92,7 @@ class AnthropicErrorAdapter(ProviderErrorAdapter):
return "Anthropic API encountered an error. Please verify your API key."
sanitized = message
# Comprehensive Anthropic patterns
patterns = [
(r'sk-ant-[a-zA-Z0-9_-]{10,}', '[REDACTED_KEY]'), # Anthropic API keys
@@ -118,6 +111,34 @@ class AnthropicErrorAdapter(ProviderErrorAdapter):
return sanitized
class OpenRouterErrorAdapter(ProviderErrorAdapter):
def get_provider_name(self) -> str:
return "openrouter"
def sanitize_error_message(self, message: str) -> str:
if not isinstance(message, str) or not message.strip() or len(message) > 2000:
return "OpenRouter API encountered an error. Please verify your API key and quota."
sanitized = message
# Comprehensive OpenRouter patterns
patterns = [
(r'sk-or-v1-[a-zA-Z0-9_-]{10,}', '[REDACTED_KEY]'), # OpenRouter API keys
(r'https?://[^\s]*openrouter\.ai[^\s]*', '[REDACTED_URL]'), # OpenRouter URLs
(r'Bearer\s+[a-zA-Z0-9._-]+', 'Bearer [REDACTED_TOKEN]'), # Bearer tokens
]
for pattern, replacement in patterns:
sanitized = re.sub(pattern, replacement, sanitized, flags=re.IGNORECASE)
# Check for sensitive words
sensitive_words = ['internal', 'server', 'endpoint']
if any(word in sanitized.lower() for word in sensitive_words):
return "OpenRouter API encountered an error. Please verify your API key and quota."
return sanitized
class ProviderErrorFactory:
"""Factory for provider-agnostic error handling."""
@@ -125,6 +146,7 @@ class ProviderErrorFactory:
"openai": OpenAIErrorAdapter(),
"google": GoogleAIErrorAdapter(),
"anthropic": AnthropicErrorAdapter(),
"openrouter": OpenRouterErrorAdapter(),
}
@classmethod
@@ -141,22 +163,18 @@ class ProviderErrorFactory:
"""Detect provider from error message with comprehensive pattern matching."""
if not error_str:
return "openai"
error_lower = error_str.lower()
# Case-insensitive provider detection with multiple patterns
if ("anthropic" in error_lower or
re.search(r'sk-ant-[a-zA-Z0-9_-]+', error_str, re.IGNORECASE) or
"claude" in error_lower):
# Check OpenRouter first since it may contain "openai" in model names
if ("openrouter" in error_lower or re.search(r'sk-or-v1-[a-zA-Z0-9_-]+', error_str, re.IGNORECASE)):
return "openrouter"
elif ("anthropic" in error_lower or re.search(r'sk-ant-[a-zA-Z0-9_-]+', error_str, re.IGNORECASE) or "claude" in error_lower):
return "anthropic"
elif ("google" in error_lower or
re.search(r'AIza[a-zA-Z0-9_-]+', error_str, re.IGNORECASE) or
"googleapis" in error_lower or
"vertex" in error_lower):
elif ("google" in error_lower or re.search(r'AIza[a-zA-Z0-9_-]+', error_str, re.IGNORECASE) or "googleapis" in error_lower or "vertex" in error_lower):
return "google"
elif ("openai" in error_lower or
re.search(r'sk-[a-zA-Z0-9]{48}', error_str, re.IGNORECASE) or
"gpt" in error_lower):
elif ("openai" in error_lower or re.search(r'sk-[a-zA-Z0-9]{48}', error_str, re.IGNORECASE) or "gpt" in error_lower):
return "openai"
else:
return "openai" # Safe default
return "openai" # Safe default

18
python/src/server/services/llm_provider_service.py
View File

@@ -554,12 +554,12 @@ async def _get_optimal_ollama_instance(instance_type: str | None = None,
base_url_override: str | None = None) -> str:
"""
Get the optimal Ollama instance URL based on configuration and health status.
Args:
instance_type: Preferred instance type ('chat', 'embedding', 'both', or None)
use_embedding_provider: Whether this is for embedding operations
base_url_override: Override URL if specified
Returns:
Best available Ollama instance URL
"""
@@ -655,8 +655,8 @@ async def get_embedding_model(provider: str | None = None) -> str:
return "text-embedding-004"
elif provider_name == "openrouter":
# OpenRouter supports both OpenAI and Google embedding models
# Default to OpenAI's latest for compatibility
return "text-embedding-3-small"
# Model names MUST include provider prefix for OpenRouter API
return "openai/text-embedding-3-small"
elif provider_name == "anthropic":
# Anthropic supports OpenAI and Google embedding models through their API
# Default to OpenAI's latest for compatibility
@@ -846,7 +846,7 @@ def _extract_reasoning_strings(value: Any) -> list[str]:
text = value.strip()
return [text] if text else []
if isinstance(value, (list, tuple, set)):
if isinstance(value, list | tuple | set):
collected: list[str] = []
for item in value:
collected.extend(_extract_reasoning_strings(item))
@@ -1135,11 +1135,11 @@ def prepare_chat_completion_params(model: str, params: dict) -> dict:
async def get_embedding_model_with_routing(provider: str | None = None, instance_url: str | None = None) -> tuple[str, str]:
"""
Get the embedding model with intelligent routing for multi-instance setups.
Args:
provider: Override provider selection
instance_url: Specific instance URL to use
Returns:
Tuple of (model_name, instance_url) for embedding operations
"""
@@ -1171,11 +1171,11 @@ async def get_embedding_model_with_routing(provider: str | None = None, instance
async def validate_provider_instance(provider: str, instance_url: str | None = None) -> dict[str, any]:
"""
Validate a provider instance and return health information.
Args:
provider: Provider name (openai, ollama, google, etc.)
instance_url: Instance URL for providers that support multiple instances
Returns:
Dictionary with validation results and health status
"""

137
python/src/server/services/openrouter_discovery_service.py Normal file
View File

@@ -0,0 +1,137 @@
"""
OpenRouter model discovery service.
Provides discovery and metadata for OpenRouter embedding models.
"""
from pydantic import BaseModel, Field, field_validator
class OpenRouterEmbeddingModel(BaseModel):
"""OpenRouter embedding model metadata."""
id: str = Field(..., description="Full model ID with provider prefix (e.g., openai/text-embedding-3-large)")
provider: str = Field(..., description="Provider name (openai, google, qwen, mistralai)")
name: str = Field(..., description="Display name without prefix")
dimensions: int = Field(..., description="Embedding dimensions")
context_length: int = Field(..., description="Maximum context window in tokens")
pricing_per_1m_tokens: float = Field(..., description="Cost per 1M tokens in USD")
supports_dimension_reduction: bool = Field(default=False, description="Whether model supports dimension parameter")
@field_validator("id")
@classmethod
def validate_model_id_has_prefix(cls, v: str) -> str:
"""Ensure model ID includes provider prefix."""
if "/" not in v:
raise ValueError("OpenRouter model IDs must include provider prefix (e.g., openai/model-name)")
return v
class OpenRouterModelListResponse(BaseModel):
"""Response from OpenRouter model discovery."""
embedding_models: list[OpenRouterEmbeddingModel] = Field(default_factory=list)
total_count: int = Field(..., description="Total number of embedding models")
class OpenRouterDiscoveryService:
"""Discover and manage OpenRouter embedding models."""
async def discover_embedding_models(self) -> list[OpenRouterEmbeddingModel]:
"""
Get available OpenRouter embedding models.
Returns hardcoded list of supported embedding models with metadata.
Future enhancement: Could fetch from OpenRouter API if they provide a models endpoint.
"""
return [
# OpenAI models via OpenRouter
OpenRouterEmbeddingModel(
id="openai/text-embedding-3-small",
provider="openai",
name="text-embedding-3-small",
dimensions=1536,
context_length=8191,
pricing_per_1m_tokens=0.02,
supports_dimension_reduction=True,
),
OpenRouterEmbeddingModel(
id="openai/text-embedding-3-large",
provider="openai",
name="text-embedding-3-large",
dimensions=3072,
context_length=8191,
pricing_per_1m_tokens=0.13,
supports_dimension_reduction=True,
),
OpenRouterEmbeddingModel(
id="openai/text-embedding-ada-002",
provider="openai",
name="text-embedding-ada-002",
dimensions=1536,
context_length=8191,
pricing_per_1m_tokens=0.10,
supports_dimension_reduction=False,
),
# Google models via OpenRouter
OpenRouterEmbeddingModel(
id="google/gemini-embedding-001",
provider="google",
name="gemini-embedding-001",
dimensions=768,
context_length=20000,
pricing_per_1m_tokens=0.00, # Free tier available
supports_dimension_reduction=True,
),
OpenRouterEmbeddingModel(
id="google/text-embedding-004",
provider="google",
name="text-embedding-004",
dimensions=768,
context_length=20000,
pricing_per_1m_tokens=0.00, # Free tier available
supports_dimension_reduction=True,
),
# Qwen models via OpenRouter
OpenRouterEmbeddingModel(
id="qwen/qwen3-embedding-0.6b",
provider="qwen",
name="qwen3-embedding-0.6b",
dimensions=1024,
context_length=32768,
pricing_per_1m_tokens=0.01,
supports_dimension_reduction=False,
),
OpenRouterEmbeddingModel(
id="qwen/qwen3-embedding-4b",
provider="qwen",
name="qwen3-embedding-4b",
dimensions=1024,
context_length=32768,
pricing_per_1m_tokens=0.01,
supports_dimension_reduction=False,
),
OpenRouterEmbeddingModel(
id="qwen/qwen3-embedding-8b",
provider="qwen",
name="qwen3-embedding-8b",
dimensions=1024,
context_length=32768,
pricing_per_1m_tokens=0.01,
supports_dimension_reduction=False,
),
# Mistral models via OpenRouter
OpenRouterEmbeddingModel(
id="mistralai/mistral-embed",
provider="mistralai",
name="mistral-embed",
dimensions=1024,
context_length=8192,
pricing_per_1m_tokens=0.10,
supports_dimension_reduction=False,
),
]
# Create singleton instance
openrouter_discovery_service = OpenRouterDiscoveryService()