MSDS_Module-3B_AI_Agent

📘 Description

This project showcases the creation of a stateful, memory-enabled AI agent using the LangGraph framework from LangChain. Built as part of MSDS 442 at Northwestern, this tutorial walks through how to wire up a language model, connect external tools (like search), stream responses, and persist memory across conversation threads — all in one cohesive agent pipeline.

⸻

🔧 Features

  • LLM Integration: GPT-4o-mini via LangChain’s OpenAI wrapper
  • Tool Calling: Integrated with Tavily’s search API for live web results
  • ReAct Agent Framework: Uses create_react_agent() to reason + act in steps
  • Memory Management: Threaded memory via MemorySaver for multi-turn conversation flow
  • Streaming UX: Live stream of responses and tokens for enhanced interactivity
  • Conversation Control: Scoped sessions using thread_id to simulate new or continued chats

⸻

💡 Key Insight

Even simple chatbots gain serious capability when you give them memory and tools. LangGraph’s modular structure makes it easy to go from stateless Q&A to dynamic agents that reason, act, recall, and respond — all while maintaining a clear UX.