agents/tools/langchain-agent.md

model

model
sonnet

LangChain/LangGraph Agent Scaffold

Create a production-ready LangChain/LangGraph agent for: $ARGUMENTS

Implement a complete agent system including:

1. Agent Architecture:

   • LangGraph state machine
   • Tool selection logic
   • Memory management
   • Context window optimization
   • Multi-agent coordination

2. Tool Implementation:

   • Custom tool creation
   • Tool validation
   • Error handling in tools
   • Tool composition
   • Async tool execution

3. Memory Systems:

   • Short-term memory
   • Long-term storage (vector DB)
   • Conversation summarization
   • Entity tracking
   • Memory retrieval strategies

4. Prompt Engineering:

   • System prompts
   • Few-shot examples
   • Chain-of-thought reasoning
   • Output formatting
   • Prompt templates

5. RAG Integration:

   • Document loading pipeline
   • Chunking strategies
   • Embedding generation
   • Vector store setup
   • Retrieval optimization

6. Production Features:

   • Streaming responses
   • Token counting
   • Cost tracking
   • Rate limiting
   • Fallback strategies

7. Observability:

   • LangSmith integration
   • Custom callbacks
   • Performance metrics
   • Decision tracking
   • Debug mode

Include error handling, testing strategies, and deployment considerations. Use the latest LangChain/LangGraph best practices.