Basic Langgraph chatbot with memory

The Problem

We want a chatbot that can retain conversation context across multiple turns without relying on a persistent database, yet also provide an optional durable storage fallback. Implementing memory in LangGraph requires integrating a checkpointing mechanism that captures and restores the graph’s state.

Services and Libraries

• langgraph.graph – builds the conversational graph and manages state transitions.

• langgraph.checkpoint.memory.InMemorySaver – lightweight in‑memory state persistence.

• langgraph.checkpoint.sqlite.SqliteSaver – durable SQLite‑based checkpointing.

• langchain.chat_models.init_chat_model – initializes the language model backend.

• dotenv.load_dotenv – loads environment variables for secure credentials.

• sqlite3 – Python’s SQLite driver for on‑disk storage.

Implementation Details of the Topic

Choose a checkpointing strategy

• In‑memory: ideal for short‑lived sessions or testing.

Example RAM Memory: https://github.com/bernatsampera/langgraph-playground/blob/main/examples/introduction/memory_chatbot/main.py

• SQLite: persists state to disk, enabling recovery after restarts.

Example Sqlite: https://github.com/bernatsampera/langgraph-playground/blob/main/examples/introduction/sqlite_memory_chatbot/main.py

1. Instantiate the checkpoint saver

   # In‑memory
   from langgraph.checkpoint.memory import InMemorySaver
   checkpointer = InMemorySaver()
   
   # SQLite
   import sqlite3
   from langgraph.checkpoint.sqlite import SqliteSaver
   conn = sqlite3.connect("checkpoints.sqlite", check_same_thread=False)
   checkpointer = SqliteSaver(conn)
   

2. Compile the graph with the chosen saver

   graph = graph_builder.compile(checkpointer=checkpointer)
   

3. Configure thread‑specific context

   from langchain_core.runnables import RunnableConfig
   config = RunnableConfig({"configurable": {"thread_id": "1"}})  # or a custom ID
   

4. Run and stream updates

   def stream_graph_updates(user_input: str):
       events = graph.stream(
           {"messages": [{"role": "user", "content": user_input}]},
           config,
           stream_mode="values",
       )
       for event in events:
           event["messages"][-1].pretty_print()
   

5. Graceful shutdown (SQLite only)

   conn.close()
   

By swapping the checkpointer implementation, the same LangGraph workflow can switch seamlessly between transient in‑memory state and durable SQLite storage, providing flexible memory management for chatbot sessions.