LangGraph + SQLite | Chatbot with Database Integration | CampusX

Summary

Nitish Singh demonstrates how to add persistent conversation storage to a LangGraph chatbot using SQLite, replacing the in-memory checkpointer used in earlier tutorials.

Why Persistence Is Needed

The chatbot already supports a Streamlit interface, token streaming, and multiple conversation threads. However, its in-memory saver stores messages in RAM, so conversations disappear when the app closes or reloads. SQLite lets users reopen and continue old chats later.

Backend Changes

• Install the external langgraph-checkpoint-sqlite package.
• Create or connect to a SQLite database file, such as chatbot.db, using Python’s SQLite library. The tutorial sets check_same_thread=False so the connection can be used across threads.
• Configure LangGraph’s SQLite checkpointer with that connection instead of the in-memory saver.
• Test persistence by sending a message, ending the run, and starting another run with the same thread ID. The chatbot can still recall earlier messages.
• Use different thread IDs to keep conversations separate. For example, the tutorial shows one thread remembering Nitish and another remembering Rahul.

Inspecting and Retrieving Threads

• The instructor opens the database in VS Code using a SQLite viewer extension to inspect saved checkpoints and messages.
• LangGraph’s checkpointer list method returns checkpoints. Since each thread may have multiple checkpoints, the tutorial extracts thread IDs and uses a set to identify unique threads.
• This logic is wrapped in a retrieve_all_threads function for the frontend to call.

Frontend Changes

• Update the Streamlit app’s initialization so it fetches existing thread IDs from the backend instead of starting with an empty thread list.
• The app can then display saved conversations and their message history on launch.
• Newly created conversations are added to the thread list as usual.

Result: Chats remain available after closing or refreshing the app, and users can resume them by selecting their thread.

Guide Topics Covered

1. Installing and configuring the SQLite checkpointer.
2. Testing message persistence across separate program runs.
3. Isolating conversation histories with thread IDs.
4. Viewing SQLite checkpoints in VS Code.
5. Retrieving existing threads and loading them into a Streamlit UI.

Main speaker/source: Nitish Singh, presenting on the CampusX channel.