Skip to main content

​
Configuration Issues

​
Issue: “OpenAI API keys must be provided”

Solution: Make sure you’ve set OPENAI_API_KEYS or OPENAI_API_KEY environment variable: 
export OPENAI_API_KEYS="sk-key1,sk-key2"
Or in code: 
config = LangChatConfig(
    openai_api_keys=["sk-..."]  # At least one key required
)

​
Issue: “Supabase URL and key must be provided”

Solution: Set Supabase credentials: 
export SUPABASE_URL="https://xxxxx.supabase.co"
export SUPABASE_KEY="eyJ..."

​
Issue: “Pinecone API key must be provided”

Solution: Set Pinecone credentials: 
export PINECONE_API_KEY="pcsk-..."
export PINECONE_INDEX_NAME="your-index-name"
Make sure your Pinecone index is created before using LangChat.

​
Runtime Issues

​
Issue: All API Keys Exhausted

Error: All API keys exhausted Solutions:
  • Add more API keys: openai_api_keys=["sk-1", "sk-2", "sk-3"]
  • Increase retry count: max_llm_retries=3
  • Check API key validity
  • Wait for rate limit reset

​
Issue: Rate Limit Errors

Error: Rate limit exceeded Solutions:
  • Use multiple API keys from different organizations
  • Reduce request frequency
  • Implement request queuing
  • Upgrade OpenAI plan

​
Issue: Model Not Found

Error: Model not found Solutions:
  • Verify model name: "gpt-4o-mini", "gpt-4o", etc.
  • Check if model is available in your region
  • Ensure you have access to the model

​
Vector Search Issues

​
Issue: No Relevant Results

Solutions:
  • Increase retrieval_k: retrieval_k=10
  • Check Pinecone index has documents
  • Verify embeddings are generated correctly
  • Try different embedding model

​
Issue: Index Not Found

Error: Index 'xxx' not found Solutions:
  • Check index name spelling
  • Verify index exists in Pinecone console
  • Check API key permissions
  • Create index first

​
Issue: Dimension Mismatch

Error: Dimension mismatch Solutions:
  • Ensure embedding model dimension matches index dimension
  • Recreate index with correct dimension:
    • text-embedding-3-large: 3072 dimensions
    • text-embedding-3-small: 1536 dimensions
    • text-embedding-ada-002: 1536 dimensions

​
Database Issues

​
Issue: Connection Failed

Error: Failed to connect to Supabase Solutions:
  • Verify URL includes https://
  • Check internet connection
  • Verify project is active
  • Check API key validity

​
Issue: Table Not Found

Error: Table 'xxx' not found Solutions:
  • Tables are auto-created on first run
  • Verify table exists in Supabase dashboard
  • Check table name spelling
  • Ensure proper permissions

​
Issue: Permission Denied

Error: Permission denied Solutions:
  • Check RLS policies
  • Use service_role key for server-side operations
  • Verify API key permissions
  • Review Row Level Security settings

​
Reranker Issues

​
Issue: Model Download Fails

Error: Failed to download model Solutions:
  • Check internet connection
  • Verify model name: "ms-marco-MiniLM-L-12-v2"
  • Check disk space
  • Ensure write permissions on cache directory

​
Issue: Import Errors

Error: Could not import FlashrankRerank Solutions: 
pip install langchain langchain-community
pip install flashrank

​
Issue: Slow Performance

Solutions:
  • Use smaller model: reranker_model="ms-marco-MiniLM-L-6-v2"
  • Reduce retrieval count before reranking
  • Use GPU if available

​
Performance Issues

​
Issue: Slow Response Times

Solutions:
  • Reduce retrieval_k: retrieval_k=5
  • Use smaller embedding model
  • Reduce reranker_top_n: reranker_top_n=3
  • Optimize Pinecone index
  • Use faster model: openai_model="gpt-4o-mini"

​
Issue: High Memory Usage

Solutions:
  • Reduce max_chat_history: max_chat_history=10
  • Use smaller reranker model
  • Clear session history periodically
  • Use streaming responses (if supported)

​
General Debugging

​
Enable Logging

import logging

logging.basicConfig(level=logging.DEBUG)
logger = logging.getLogger('langchat')
logger.setLevel(logging.DEBUG)

​
Check Configuration

from langchat import LangChatConfig

config = LangChatConfig.from_env()
print(f"OpenAI Model: {config.openai_model}")
print(f"Pinecone Index: {config.pinecone_index_name}")
print(f"Supabase URL: {config.supabase_url}")

​
Test Individual Components

from langchat import LangChat, LangChatConfig

config = LangChatConfig.from_env()
langchat = LangChat(config=config)

# Test OpenAI
engine = langchat.engine
test_response = engine.llm.invoke([{"role": "user", "content": "Hello"}])
print(f"OpenAI works: {test_response.content}")

# Test Pinecone
retriever = engine.vector_adapter.get_retriever(k=1)
docs = retriever.get_relevant_documents("test")
print(f"Pinecone works: {len(docs)} documents")

# Test Supabase
client = engine.supabase_adapter.client
result = client.table("chat_history").select("id").limit(1).execute()
print(f"Supabase works: {len(result.data)} records")

​
Getting Help