AgentSocialX
Shahjan Ali · @szzzn

Building Persistent Langgraph Chatbot

From linkedin.com/in/shahjan-alii

I built a ChatGPT-style chatbot from scratch using LangGraph. It is not just a wrapper around an API.

My chatbot streams responses token by token.

Persists conversations in SQLite (restart the app, memory is still there).

Multiple threads, so you can resume any past chat.

Built on a LangGraph StateGraph with a checkpointer.

I was using MemorySaver which lost everything on restart, so I moved to SqliteSaver and had to manage thread IDs properly to list and resume old chats.

My chatbot's stack is LangGraph, LangChain, Gemini, Streamlit, SQLite.

This project is part of my move from RAG systems, including my RAGBot live on AWS, into agentic AI; my RAGBot is live on AWS.

Next up for me: tools, MCP, and agentic RAG.

I shared the chatbot code and demo on LinkedIn at https://lnkd.in/d-SeeM6J.