Troubleshooting Guide

Common issues and solutions for AI development tools.

Claude Code

"API Key Invalid"

# Set your API key
export ANTHROPIC_API_KEY=your-key-here

Context Too Long

• Use /compact to summarize conversation
• Start fresh with /clear
• Break large tasks into smaller chunks

Permission Denied

• Claude Code requires write access to project
• Check file/folder permissions
• Run from project root directory

FlowiseAI

Port Already in Use

# Use different port
npx flowise start -p 3001

Vector Store Connection Failed

1. Check database credentials
2. Verify database is running
3. Check network/firewall settings

"Model Not Found"

• Verify API key is set
• Check model name spelling
• Ensure model is available in your region

Memory Issues with Large Documents

• Enable chunking in document loader
• Reduce chunk size
• Use streaming for responses

n8n

Webhook Not Working

# Run with tunnel for external access
n8n start --tunnel

"Execution Timed Out"

• Increase timeout in workflow settings
• Split large data processing
• Use batch processing

Credentials Not Saving

• Check file permissions on ~/.n8n
• Verify encryption key is set
• Restart n8n service

Ollama

Model Download Stuck

# Cancel and retry
Ctrl+C
ollama pull modelname

"CUDA Out of Memory"

• Use smaller model (7B instead of 13B)
• Close other GPU applications
• Use CPU mode with --num-gpu 0

Slow Performance

• Check available RAM
• Use quantized models (Q4_0)
• Reduce context length

OpenAI API

Rate Limit Exceeded

• Implement exponential backoff
• Use batch API for large jobs
• Upgrade API tier

"Context Length Exceeded"

• Reduce input size
• Use summarization
• Switch to model with larger context

High Costs

• Use gpt-3.5-turbo for simple tasks
• Cache responses
• Optimize prompts for shorter outputs

General Tips

API Keys Best Practices

1. Never commit keys to git
2. Use environment variables
3. Rotate keys periodically
4. Use separate keys for dev/prod

Performance Optimization

1. Use streaming for long responses
2. Implement caching
3. Batch similar requests
4. Choose appropriate model size

Debugging AI Responses

1. Log prompts and responses
2. Use lower temperature for consistency
3. Add system prompts for behavior
4. Test with simple inputs first

Version Mismatches

Many issues come from version differences. Always check:

• Tool version matches tutorial
• API version is current
• Dependencies are compatible