Session state
Each session gets its own thread in a LangGraph checkpointer that stores the message history.
Live · LLM · Conversational AI
Chatbot with memory
Most demo chatbots forget everything between messages. This one uses a LangGraph checkpointer to persist conversation state per session — so it genuinely remembers what you told it earlier. Tell it your name and what you do, then ask it to recall them a few turns later.
Start by telling the assistant your name and what you do — then ask it to recall that a few messages later.
Replies run on a local CPU model — expect 8–20 s each. Session memory persists until you reset.
Try this sequence
session:
How it works
Append turn
Your message is appended to the thread state via the graph reducer.
Generate
The model sees the recent history and replies with that context in mind.
Persist
The reply is written back to the thread, so the next turn remembers it.
Bounded memory
History is capped and sessions expire after 30 minutes to keep memory in check.
Production swap
Move the checkpointer to Postgres/Redis, add long-term memory and retrieval, swap in Claude — same graph.
Want a real assistant for your business?
We build production conversational AI with long-term memory, retrieval over your knowledge, tool actions, guardrails, and human handoff — grounded in your data, not generic chit-chat.
Ready to start
Turn one AI use case into measurable production value.
Book a 30-minute consultation. We will walk through the use case, sketch the value case, and tell you honestly whether we can help.