Shared AI Memory

BunkerAI maintains a unified conversation context across all connected channels. Whether you chat in the Web UI, send a message on Slack, or use the Telegram bot, the AI remembers your previous exchanges and your broker's context.

What Shared AI Memory Is

All three channels - Web Chat, Slack, and Telegram - connect to the same BunkerAI backend for your BunkerM instance. There is one unified conversation store for your instance that all channels read from and write to. Shared memory means:

• A conversation started in Slack can be continued in Web Chat without repeating yourself.
• Anomaly context discussed via Telegram is remembered when you open Web Chat.
• Setup and topology you have explained in one session carry forward to future sessions.
• The AI does not need to relearn your broker every time you open a new channel.

How It Works

When you send a message, BunkerAI:

1. Retrieves your conversation history from the shared store.
2. Includes that history as context in the AI request.
3. Generates a response that is aware of everything previously discussed.
4. Stores the new exchange back in the shared store.

What Is Remembered

• Previous questions and answers - the full text of your previous exchanges, retained and used as context.
• Broker topology you have discussed - which clients are sensors, which groups exist, what your topic structure means.
• Actions the AI has taken - if the AI created a client or set up a watcher in a previous session, it remembers that.
• Topic annotations - annotations from Settings > Annotations are always included as a persistent knowledge base, independently of conversation history.

Benefits

• No repetition - explain your setup once. Whether you return the next day or a week later, the AI still has context.
• Consistent behavior across channels - switching channels does not change the AI's understanding of your broker.
• Faster, more accurate responses - as context accumulates, the AI answers more precisely.
• Cross-channel workflows - get an anomaly alert on Telegram, investigate via Web Chat, resolve via Slack - all as one continuous conversation.

Privacy and Scoping

Conversation memory is strictly scoped to your BunkerAI tenant. Your conversations are not visible to other BunkerM users or instances. Each BunkerM installation connected to BunkerAI has its own isolated conversation store.

Clearing Conversation History

1. Open the Web Chat in the BunkerM dashboard.
2. Click Clear History or New Conversation.
3. Confirm the action.

This clears the shared history for your entire instance. The next interaction on any channel starts without previous context. Topic annotations are not cleared - they persist separately.