fix(docker): Fix LLM API key handling for multi-provider support

Previously, the system incorrectly used OPENAI_API_KEY for all LLM providers
due to a hardcoded api_key_env fallback in config.yml. This caused authentication
errors when using non-OpenAI providers like Gemini.

Changes:
- Remove api_key_env from config.yml to let litellm handle provider-specific env vars
- Simplify get_llm_api_key() to return None, allowing litellm to auto-detect keys
- Update validate_llm_provider() to trust litellm's built-in key detection
- Update documentation to reflect the new automatic key handling

The fix leverages litellm's existing capability to automatically find the correct
environment variable for each provider (OPENAI_API_KEY, GEMINI_API_TOKEN, etc.)
without manual configuration.

ref #1291

README.md

@@ -373,7 +373,7 @@ async def main():
     
     async with AsyncWebCrawler(config=browser_config) as crawler:
         result = await crawler.arun(
-            url="https://docs.micronaut.io/4.7.6/guide/",
+            url="https://docs.micronaut.io/4.9.9/guide/",
             config=run_config
         )
         print(len(result.markdown.raw_markdown))
@@ -425,7 +425,7 @@ async def main():
             "type": "attribute",
             "attribute": "src"
         }
-    }
+    ]
 }
 
     extraction_strategy = JsonCssExtractionStrategy(schema, verbose=True)

deploy/docker/README.md

@@ -692,8 +692,7 @@ app:
 # Default LLM Configuration
 llm:
   provider: "openai/gpt-4o-mini"  # Can be overridden by LLM_PROVIDER env var
-  api_key_env: "OPENAI_API_KEY"
+  # api_key: sk-...  # If you pass the API key directly (not recommended)
-  # api_key: sk-...  # If you pass the API key directly then api_key_env will be ignored
 
 # Redis Configuration (Used by internal Redis server managed by supervisord)
 redis:

deploy/docker/api.py

@@ -4,7 +4,7 @@ import asyncio
 from typing import List, Tuple, Dict
 from functools import partial
 from uuid import uuid4
-from datetime import datetime
+from datetime import datetime, timezone
 from base64 import b64encode
 
 import logging
@@ -96,7 +96,7 @@ async def handle_llm_qa(
         response = perform_completion_with_backoff(
             provider=config["llm"]["provider"],
             prompt_with_variables=prompt,
-            api_token=get_llm_api_key(config)
+            api_token=get_llm_api_key(config)  # Returns None to let litellm handle it
         )
 
         return response.choices[0].message.content
@@ -127,7 +127,7 @@ async def process_llm_extraction(
                 "error": error_msg
             })
             return
-        api_key = get_llm_api_key(config, provider)
+        api_key = get_llm_api_key(config, provider)  # Returns None to let litellm handle it
         llm_strategy = LLMExtractionStrategy(
             llm_config=LLMConfig(
                 provider=provider or config["llm"]["provider"],
@@ -203,7 +203,7 @@ async def handle_markdown_request(
                 FilterType.LLM: LLMContentFilter(
                     llm_config=LLMConfig(
                         provider=provider or config["llm"]["provider"],
-                        api_token=get_llm_api_key(config, provider),
+                        api_token=get_llm_api_key(config, provider),  # Returns None to let litellm handle it
                     ),
                     instruction=query or "Extract main content"
                 )
@@ -576,7 +576,7 @@ async def handle_crawl_job(
     task_id = f"crawl_{uuid4().hex[:8]}"
     await redis.hset(f"task:{task_id}", mapping={
         "status": TaskStatus.PROCESSING,         # <-- keep enum values consistent
-        "created_at": datetime.utcnow().isoformat(),
+        "created_at": datetime.now(timezone.utc).replace(tzinfo=None).isoformat(),
         "url": json.dumps(urls),                 # store list as JSON string
         "result": "",
         "error": "",

deploy/docker/config.yml

@@ -11,8 +11,7 @@ app:
 # Default LLM Configuration
 llm:
   provider: "openai/gpt-4o-mini"
-  api_key_env: "OPENAI_API_KEY"
+  # api_key: sk-...  # If you pass the API key directly (not recommended)
-  # api_key: sk-...  # If you pass the API key directly then api_key_env will be ignored
 
 # Redis Configuration
 redis:

deploy/docker/utils.py

@@ -71,7 +71,7 @@ def decode_redis_hash(hash_data: Dict[bytes, bytes]) -> Dict[str, str]:
 
 
 
-def get_llm_api_key(config: Dict, provider: Optional[str] = None) -> str:
+def get_llm_api_key(config: Dict, provider: Optional[str] = None) -> Optional[str]:
     """Get the appropriate API key based on the LLM provider.
     
     Args:
@@ -79,19 +79,14 @@ def get_llm_api_key(config: Dict, provider: Optional[str] = None) -> str:
         provider: Optional provider override (e.g., "openai/gpt-4")
     
     Returns:
-        The API key for the provider, or empty string if not found
+        The API key if directly configured, otherwise None to let litellm handle it
     """
-        
+    # Check if direct API key is configured (for backward compatibility)
-    # Use provided provider or fall back to config
-    if not provider:
-        provider = config["llm"]["provider"]
-    
-    # Check if direct API key is configured
     if "api_key" in config["llm"]:
         return config["llm"]["api_key"]
     
-    # Fall back to the configured api_key_env if no match
+    # Return None - litellm will automatically find the right environment variable
-    return os.environ.get(config["llm"].get("api_key_env", ""), "")
+    return None
 
 
 def validate_llm_provider(config: Dict, provider: Optional[str] = None) -> tuple[bool, str]:
@@ -104,16 +99,12 @@ def validate_llm_provider(config: Dict, provider: Optional[str] = None) -> tuple
     Returns:
         Tuple of (is_valid, error_message)
     """
-    # Use provided provider or fall back to config
+    # If a direct API key is configured, validation passes
-    if not provider:
+    if "api_key" in config["llm"]:
-        provider = config["llm"]["provider"]
+        return True, ""
-    
-    # Get the API key for this provider
-    api_key = get_llm_api_key(config, provider)
-    
-    if not api_key:
-        return False, f"No API key found for provider '{provider}'. Please set the appropriate environment variable."
     
+    # Otherwise, trust that litellm will find the appropriate environment variable
+    # We can't easily validate this without reimplementing litellm's logic
     return True, ""
 
 

docs/md_v2/core/docker-deployment.md

@@ -176,7 +176,7 @@ The Docker setup now supports flexible LLM provider configuration through three
 
 3. **Config File Default**: Falls back to `config.yml` (default: `openai/gpt-4o-mini`)
 
-The system automatically selects the appropriate API key based on the configured `api_key_env` in the config file.
+The system automatically selects the appropriate API key based on the provider. LiteLLM handles finding the correct environment variable for each provider (e.g., OPENAI_API_KEY for OpenAI, GEMINI_API_TOKEN for Google Gemini, etc.).
 
 #### 3. Build and Run with Compose
 
@@ -693,8 +693,7 @@ app:
 # Default LLM Configuration
 llm:
   provider: "openai/gpt-4o-mini"  # Can be overridden by LLM_PROVIDER env var
-  api_key_env: "OPENAI_API_KEY"
+  # api_key: sk-...  # If you pass the API key directly (not recommended)
-  # api_key: sk-...  # If you pass the API key directly then api_key_env will be ignored
 
 # Redis Configuration (Used by internal Redis server managed by supervisord)
 redis:

docs/md_v2/core/url-seeding.md

@@ -102,16 +102,16 @@ async def smart_blog_crawler():
     
     # Step 2: Configure discovery - let's find all blog posts
     config = SeedingConfig(
-        source="sitemap",           # Use the website's sitemap
+        source="sitemap+cc",      # Use the website's sitemap+cc
-        pattern="*/blog/*.html",    # Only blog posts
+        pattern="*/courses/*",    # Only courses related posts
         extract_head=True,          # Get page metadata
         max_urls=100               # Limit for this example
     )
     
     # Step 3: Discover URLs from the Python blog
-    print("🔍 Discovering blog posts...")
+    print("🔍 Discovering course posts...")
     urls = await seeder.urls("realpython.com", config)
-    print(f"✅ Found {len(urls)} blog posts")
+    print(f"✅ Found {len(urls)} course posts")
     
     # Step 4: Filter for Python tutorials (using metadata!)
     tutorials = [
@@ -134,7 +134,8 @@ async def smart_blog_crawler():
     async with AsyncWebCrawler() as crawler:
         config = CrawlerRunConfig(
             only_text=True,
-            word_count_threshold=300  # Only substantial articles
+            word_count_threshold=300,  # Only substantial articles
+            stream=True
         )
         
         # Extract URLs and crawl them
@@ -155,7 +156,7 @@ asyncio.run(smart_blog_crawler())
 
 **What just happened?**
 
-1. We discovered all blog URLs from the sitemap
+1. We discovered all blog URLs from the sitemap+cc
 2. We filtered using metadata (no crawling needed!)
 3. We crawled only the relevant tutorials
 4. We saved tons of time and bandwidth
@@ -282,8 +283,8 @@ config = SeedingConfig(
     live_check=True,  # Verify each URL is accessible
     concurrency=20    # Check 20 URLs in parallel
 )
-
+async with AsyncUrlSeeder() as seeder:
-urls = await seeder.urls("example.com", config)
+    urls = await seeder.urls("example.com", config)
 
 # Now you can filter by status
 live_urls = [u for u in urls if u["status"] == "valid"]
@@ -311,8 +312,8 @@ This is where URL seeding gets really powerful. Instead of crawling entire pages
 config = SeedingConfig(
     extract_head=True  # Extract metadata from <head> section
 )
-
+async with AsyncUrlSeeder() as seeder:
-urls = await seeder.urls("example.com", config)
+    urls = await seeder.urls("example.com", config)
 
 # Now each URL has rich metadata
 for url in urls[:3]:
@@ -387,8 +388,8 @@ config = SeedingConfig(
     scoring_method="bm25",
     score_threshold=0.3
 )
-
+async with AsyncUrlSeeder() as seeder:
-urls = await seeder.urls("example.com", config)
+    urls = await seeder.urls("example.com", config)
 
 # URLs are scored based on:
 # 1. Domain parts matching (e.g., 'python' in python.example.com)
@@ -429,8 +430,8 @@ config = SeedingConfig(
     extract_head=True,
     live_check=True
 )
-
+async with AsyncUrlSeeder() as seeder:
-urls = await seeder.urls("blog.example.com", config)
+    urls = await seeder.urls("blog.example.com", config)
 
 # Analyze the results
 for url in urls[:5]:
@@ -488,8 +489,8 @@ config = SeedingConfig(
     scoring_method="bm25",       # Use BM25 algorithm
     score_threshold=0.3          # Minimum relevance score
 )
-
+async with AsyncUrlSeeder() as seeder:
-urls = await seeder.urls("realpython.com", config)
+    urls = await seeder.urls("realpython.com", config)
 
 # Results are automatically sorted by relevance!
 for url in urls[:5]:
@@ -511,8 +512,8 @@ config = SeedingConfig(
     score_threshold=0.5,
     max_urls=20
 )
-
+async with AsyncUrlSeeder() as seeder:
-urls = await seeder.urls("docs.example.com", config)
+    urls = await seeder.urls("docs.example.com", config)
 
 # The highest scoring URLs will be API docs!
 ```
@@ -529,8 +530,8 @@ config = SeedingConfig(
     score_threshold=0.4,
     pattern="*/product/*"  # Combine with pattern matching
 )
-
+async with AsyncUrlSeeder() as seeder:
-urls = await seeder.urls("shop.example.com", config)
+    urls = await seeder.urls("shop.example.com", config)
 
 # Filter further by price (from metadata)
 affordable = [
@@ -550,8 +551,8 @@ config = SeedingConfig(
     scoring_method="bm25",
     score_threshold=0.35
 )
-
+async with AsyncUrlSeeder() as seeder:
-urls = await seeder.urls("technews.com", config)
+    urls = await seeder.urls("technews.com", config)
 
 # Filter by date
 from datetime import datetime, timedelta
@@ -591,8 +592,8 @@ for query in queries:
         score_threshold=0.4,
         max_urls=10  # Top 10 per topic
     )
-    
+    async with AsyncUrlSeeder() as seeder:
-    urls = await seeder.urls("learning-platform.com", config)
+        urls = await seeder.urls("learning-platform.com", config)
     all_tutorials.extend(urls)
 
 # Remove duplicates while preserving order
@@ -625,7 +626,8 @@ config = SeedingConfig(
 )
 
 # Returns a dictionary: {domain: [urls]}
-results = await seeder.many_urls(domains, config)
+async with AsyncUrlSeeder() as seeder:
+    results = await seeder.many_urls(domains, config)
 
 # Process results
 for domain, urls in results.items():
@@ -654,8 +656,8 @@ config = SeedingConfig(
     pattern="*/blog/*",
     max_urls=100
 )
-
+async with AsyncUrlSeeder() as seeder:
-results = await seeder.many_urls(competitors, config)
+    results = await seeder.many_urls(competitors, config)
 
 # Analyze content types
 for domain, urls in results.items():
@@ -690,8 +692,8 @@ config = SeedingConfig(
     score_threshold=0.3,
     max_urls=20  # Per site
 )
-
+async with AsyncUrlSeeder() as seeder:
-results = await seeder.many_urls(educational_sites, config)
+    results = await seeder.many_urls(educational_sites, config)
 
 # Find the best beginner tutorials
 all_tutorials = []
@@ -731,8 +733,8 @@ config = SeedingConfig(
     score_threshold=0.5,  # High threshold for relevance
     max_urls=10
 )
-
+async with AsyncUrlSeeder() as seeder:
-results = await seeder.many_urls(news_sites, config)
+    results = await seeder.many_urls(news_sites, config)
 
 # Collect all mentions
 mentions = []