This is a hands-on project to learn to create chatbot agent using LangChain & LangGraph. Use hatch package manager
Based on gpt-4o-mini LLM model, with added features:
- Tool calling: multiply tool, web search tool
- Memory: short term memory from conversation thread
- Human-in-the-loop: verify tool call before executing, in case conversational topic is sensitive
- State: add custom state "sensitivity", indicating if conversation topic is sensitive or not
Agent architecture: use 2 chat models for separation of concern
- Helper model to classify if conversation is sensitive or not
- Main model equiped with tools
I also made FastAPI backend and NextJS frontend to interact with chatbot Deploy chatbot as standalone container via LangGraph Platform
Deploy to AWS ECS as separate services Current deployment's URL: http://chatbot-alb-468353364.eu-central-1.elb.amazonaws.com/ (I don't keep this URL if I take it down, because I don't pay money for a fixed one))
PyPI package: https://pypi.org/project/chatbot_nam685/
pip install chatbot_nam685chatbot_nam685 is distributed under the terms of the MIT license.