Memory Chatbot
The Memory Chatbot template builds a conversational AI that remembers previous messages across chat sessions. Unlike a basic chatbot that treats every message in isolation, this flow stores chat history and injects it into each new prompt — so users can say "remember what I told you earlier" and the LLM can actually do it.
How the flow works
_10Chat Input → Message History → Prompt → Language Model → Chat Output
The Message History component retrieves past messages from the session store and passes them to the Prompt component. The prompt combines the history with the new user message and sends the full context to the Language Model, which generates a response. The response is saved back into history for the next turn.
Prerequisites
- An API key for a supported LLM provider (e.g., OpenAI)
Quickstart
-
In Orbis Studio, open the Templates library and select Memory Chatbot.
-
In the Language Model component, add your LLM API key and select a model.
-
Open the Playground and introduce yourself — for example: "My name is Deniz."
-
Start a new chat session in the Playground, then ask: "What is my name?" — the LLM retrieves your name from the stored session history.
Key components
| Component | Role |
|---|---|
| Chat Input | Receives the user's message from the Playground |
| Message History | Loads previous messages for the current session (configurable up to 100 messages) |
| Prompt | Combines history and the new user message into a structured LLM prompt |
| Language Model | Calls the selected LLM and returns a response |
| Chat Output | Displays the response in the Playground |
Configuration tips
- Session isolation: Each session ID in the Playground maintains its own history. Switch sessions to start a fresh conversation without deleting stored history.
- History depth: By default the Message History component retrieves up to 100 messages. Reduce this if you want shorter context windows.
- Prompt customization: Edit the Prompt component's template to change the system personality or format of injected history.
Next steps
- Add a Knowledge Retrieval component to ground responses in your own documents.
- Connect a Guardrails component to enforce content policies on every response.