Fix Issue #362: Provider-agnostic error handling for all LLM providers

archon-ui-main/src/features/knowledge/hooks/useKnowledgeQueries.ts

@@ -11,6 +11,7 @@ import { useActiveOperations } from "../progress/hooks";
 import { progressKeys } from "../progress/hooks/useProgressQueries";
 import type { ActiveOperation, ActiveOperationsResponse } from "../progress/types";
 import { knowledgeService } from "../services";
+import { getProviderErrorMessage } from "../utils/providerErrorHandler";
 import type {
   CrawlRequest,
   CrawlStartResponse,
@@ -273,7 +274,7 @@ export function useCrawlUrl() {
         queryClient.setQueryData(progressKeys.list(), context.previousOperations);
       }
 
-      const errorMessage = error instanceof Error ? error.message : "Failed to start crawl";
+      const errorMessage = getProviderErrorMessage(error) || "Failed to start crawl";
       showToast(errorMessage, "error");
     },
   });

archon-ui-main/src/features/knowledge/utils/index.ts

@@ -1 +1,2 @@
 export * from "./knowledge-utils";
+export * from "./providerErrorHandler";

archon-ui-main/src/features/knowledge/utils/providerErrorHandler.ts

@@ -0,0 +1,71 @@
+/**
+ * Provider-agnostic error handler for LLM operations
+ * Supports OpenAI, Google AI, Anthropic, and other providers
+ */
+
+export interface ProviderError extends Error {
+  statusCode?: number;
+  provider?: string;
+  errorType?: string;
+  isProviderError?: boolean;
+}
+
+/**
+ * Parse backend error responses into provider-aware error objects
+ */
+export function parseProviderError(error: unknown): ProviderError {
+  const providerError = error as ProviderError;
+
+  // Check if this is a structured provider error from backend
+  if (error && typeof error === 'object') {
+    if (error.statusCode || error.status) {
+      providerError.statusCode = error.statusCode || error.status;
+    }
+
+    // Parse backend error structure
+    if (error.message && error.message.includes('detail')) {
+      try {
+        const parsed = JSON.parse(error.message);
+        if (parsed.detail && parsed.detail.error_type) {
+          providerError.isProviderError = true;
+          providerError.provider = parsed.detail.provider || 'LLM';
+          providerError.errorType = parsed.detail.error_type;
+          providerError.message = parsed.detail.message || error.message;
+        }
+      } catch {
+        // If parsing fails, use message as-is
+      }
+    }
+  }
+
+  return providerError;
+}
+
+/**
+ * Get user-friendly error message for any LLM provider
+ */
+export function getProviderErrorMessage(error: unknown): string {
+  const parsed = parseProviderError(error);
+
+  if (parsed.isProviderError) {
+    const provider = parsed.provider || 'LLM';
+
+    switch (parsed.errorType) {
+      case 'authentication_failed':
+        return `Please verify your ${provider} API key in Settings.`;
+      case 'quota_exhausted':
+        return `${provider} quota exhausted. Please check your billing settings.`;
+      case 'rate_limit':
+        return `${provider} rate limit exceeded. Please wait and try again.`;
+      default:
+        return `${provider} API error. Please check your configuration.`;
+    }
+  }
+
+  // Handle status codes for non-structured errors
+  if (parsed.statusCode === 401) {
+    return "Please verify your API key in Settings.";
+  }
+
+  return parsed.message || "An error occurred.";
+}

python/src/server/api_routes/knowledge_api.py

@@ -22,6 +22,8 @@
 from ..config.logfire_config import get_logger, safe_logfire_error, safe_logfire_info
 from ..services.crawler_manager import get_crawler
 from ..services.crawling import CrawlingService
+from ..services.credential_service import credential_service
+from ..services.embeddings.provider_error_adapters import ProviderErrorFactory
 from ..services.knowledge import DatabaseMetricsService, KnowledgeItemService, KnowledgeSummaryService
 from ..services.search.rag_service import RAGService
 from ..services.storage import DocumentStorageService
@@ -53,6 +55,59 @@
 active_crawl_tasks: dict[str, asyncio.Task] = {}
 
 
+
+
+async def _validate_provider_api_key(provider: str = None) -> None:
+    """Validate LLM provider API key before starting operations."""
+    logger.info("🔑 Starting API key validation...")
+
+    try:
+        if not provider:
+            provider = "openai"
+
+        logger.info(f"🔑 Testing {provider.title()} API key with minimal embedding request...")
+
+        # Test API key with minimal embedding request - this will fail if key is invalid
+        from ..services.embeddings.embedding_service import create_embedding
+        test_result = await create_embedding(text="test")
+
+        if not test_result:
+            logger.error(f"❌ {provider.title()} API key validation failed - no embedding returned")
+            raise HTTPException(
+                status_code=401,
+                detail={
+                    "error": f"Invalid {provider.title()} API key",
+                    "message": f"Please verify your {provider.title()} API key in Settings.",
+                    "error_type": "authentication_failed",
+                    "provider": provider
+                }
+            )
+
+        logger.info(f"✅ {provider.title()} API key validation successful")
+
+    except HTTPException:
+        # Re-raise our intended HTTP exceptions
+        logger.error("🚨 Re-raising HTTPException from validation")
+        raise
+    except Exception as e:
+        # Sanitize error before logging to prevent sensitive data exposure
+        error_str = str(e)
+        sanitized_error = ProviderErrorFactory.sanitize_provider_error(error_str, provider or "openai")
+        logger.error(f"❌ Caught exception during API key validation: {sanitized_error}")
+
+        # Always fail for any exception during validation - better safe than sorry
+        logger.error("🚨 API key validation failed - blocking crawl operation")
+        raise HTTPException(
+            status_code=401,
+            detail={
+                "error": "Invalid API key",
+                "message": f"Please verify your {(provider or 'openai').title()} API key in Settings before starting a crawl.",
+                "error_type": "authentication_failed",
+                "provider": provider or "openai"
+            }
+        ) from None
+
+
 # Request Models
 class KnowledgeItemRequest(BaseModel):
     url: str
@@ -479,6 +534,14 @@ async def get_knowledge_item_code_examples(
 @router.post("/knowledge-items/{source_id}/refresh")
 async def refresh_knowledge_item(source_id: str):
     """Refresh a knowledge item by re-crawling its URL with the same metadata."""
+
+    # Validate API key before starting expensive refresh operation
+    logger.info("🔍 About to validate API key for refresh...")
+    provider_config = await credential_service.get_active_provider("embedding")
+    provider = provider_config.get("provider", "openai")
+    await _validate_provider_api_key(provider)
+    logger.info("✅ API key validation completed successfully for refresh")
+
     try:
         safe_logfire_info(f"Starting knowledge item refresh | source_id={source_id}")
 
@@ -597,6 +660,13 @@ async def crawl_knowledge_item(request: KnowledgeItemRequest):
     if not request.url.startswith(("http://", "https://")):
         raise HTTPException(status_code=422, detail="URL must start with http:// or https://")
 
+    # Validate API key before starting expensive operation
+    logger.info("🔍 About to validate API key...")
+    provider_config = await credential_service.get_active_provider("embedding")
+    provider = provider_config.get("provider", "openai")
+    await _validate_provider_api_key(provider)
+    logger.info("✅ API key validation completed successfully")
+
     try:
         safe_logfire_info(
             f"Starting knowledge item crawl | url={str(request.url)} | knowledge_type={request.knowledge_type} | tags={request.tags}"
@@ -750,6 +820,14 @@ async def upload_document(
     knowledge_type: str = Form("technical"),
 ):
     """Upload and process a document with progress tracking."""
+
+    # Validate API key before starting expensive upload operation  
+    logger.info("🔍 About to validate API key for upload...")
+    provider_config = await credential_service.get_active_provider("embedding")
+    provider = provider_config.get("provider", "openai")
+    await _validate_provider_api_key(provider)
+    logger.info("✅ API key validation completed successfully for upload")
+
     try:
         # DETAILED LOGGING: Track knowledge_type parameter flow
         safe_logfire_info(

python/src/server/services/embeddings/embedding_exceptions.py

@@ -99,6 +99,22 @@ def __init__(self, message: str, original_error: Exception | None = None, **kwar
             self.metadata["original_error_message"] = str(original_error)
 
 
+class EmbeddingAuthenticationError(EmbeddingError):
+    """
+    Raised when API authentication fails (invalid or expired API key).
+
+    This is a CRITICAL error that should stop the entire process
+    as continuing would be pointless without valid API access.
+    """
+
+    def __init__(self, message: str, api_key_prefix: str | None = None, **kwargs):
+        super().__init__(message, **kwargs)
+        # Store masked API key prefix for debugging
+        self.api_key_prefix = api_key_prefix[:3] + "…" if api_key_prefix and len(api_key_prefix) >= 3 else None
+        if self.api_key_prefix:
+            self.metadata["api_key_prefix"] = self.api_key_prefix
+
+
 class EmbeddingValidationError(EmbeddingError):
     """
     Raised when embedding validation fails (e.g., zero vector detected).

python/src/server/services/embeddings/provider_error_adapters.py

@@ -0,0 +1,162 @@
+"""
+Provider-agnostic error handling for LLM embedding services.
+
+Supports OpenAI, Google AI, Anthropic, Ollama, and future providers
+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."""
+
+    @abstractmethod
+    def get_provider_name(self) -> str:
+        pass
+
+    @abstractmethod
+    def sanitize_error_message(self, message: str) -> str:
+        pass
+
+
+class OpenAIErrorAdapter(ProviderErrorAdapter):
+    def get_provider_name(self) -> str:
+        return "openai"
+
+    def sanitize_error_message(self, message: str) -> str:
+        if not isinstance(message, str) or not message.strip() or len(message) > 2000:
+            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
+            (r'https?://[^\s]*openai\.com[^\s]*', '[REDACTED_URL]'),   # OpenAI URLs
+            (r'org-[a-zA-Z0-9]{20,}', '[REDACTED_ORG]'),              # Organization IDs
+            (r'proj_[a-zA-Z0-9]{10,}', '[REDACTED_PROJECT]'),         # Project IDs
+            (r'req_[a-zA-Z0-9]{10,}', '[REDACTED_REQUEST]'),          # Request IDs
+            (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 after sanitization
+        sensitive_words = ['internal', 'server', 'endpoint']
+        if any(word in sanitized.lower() for word in sensitive_words):
+            return "OpenAI API encountered an error. Please verify your API key and quota."
+
+        return sanitized
+
+
+class GoogleAIErrorAdapter(ProviderErrorAdapter):
+    def get_provider_name(self) -> str:
+        return "google"
+
+    def sanitize_error_message(self, message: str) -> str:
+        if not isinstance(message, str) or not message.strip() or len(message) > 2000:
+            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
+            (r'https?://[^\s]*googleapis\.com[^\s]*', '[REDACTED_URL]'),      # Google API URLs
+            (r'https?://[^\s]*googleusercontent\.com[^\s]*', '[REDACTED_URL]'), # Google content URLs
+            (r'projects/[a-zA-Z0-9_-]+', 'projects/[REDACTED_PROJECT]'),      # GCP project paths
+            (r'ya29\.[a-zA-Z0-9_-]+', '[REDACTED_TOKEN]'),                   # OAuth tokens
+            (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', 'project']
+        if any(word in sanitized.lower() for word in sensitive_words):
+            return "Google AI API encountered an error. Please verify your API key."
+
+        return sanitized
+
+
+class AnthropicErrorAdapter(ProviderErrorAdapter):
+    def get_provider_name(self) -> str:
+        return "anthropic"
+
+    def sanitize_error_message(self, message: str) -> str:
+        if not isinstance(message, str) or not message.strip() or len(message) > 2000:
+            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
+            (r'https?://[^\s]*anthropic\.com[^\s]*', '[REDACTED_URL]'),        # Anthropic 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 "Anthropic API encountered an error. Please verify your API key."
+
+        return sanitized
+
+
+class ProviderErrorFactory:
+    """Factory for provider-agnostic error handling."""
+
+    _adapters = {
+        "openai": OpenAIErrorAdapter(),
+        "google": GoogleAIErrorAdapter(),
+        "anthropic": AnthropicErrorAdapter(),
+    }
+
+    @classmethod
+    def get_adapter(cls, provider: str) -> ProviderErrorAdapter:
+        return cls._adapters.get(provider.lower(), cls._adapters["openai"])
+
+    @classmethod
+    def sanitize_provider_error(cls, message: str, provider: str) -> str:
+        adapter = cls.get_adapter(provider)
+        return adapter.sanitize_error_message(message)
+
+    @classmethod
+    def detect_provider_from_error(cls, error_str: str) -> str:
+        """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):
+            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):
+            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):
+            return "openai"
+        else:
+            return "openai"  # Safe default