Add chat history and viewer memory to an AI bot

Supply explicitly configured context files and keep dialogue bounded.

Before you start

Use a test project connected to Core. Configure the relevant integrations in the intended local service profile. Review Apply and Save before testing a changed definition.

Steps

1. Complete the basic AI bot workflow and configure a writable context root.
2. Add a chat-message capture automation with Append JSONL for Chat/RecentChat.jsonl, followed by Trim JSONL to bound retained messages.
3. Append the current viewer's request and the AI reply to a separate viewer dialogue file using a stable viewer identifier.
4. Select only the intended chat, dialogue and profile files as context for Call and Store.
5. Optionally add a maintenance branch that summarizes sufficiently long dialogue, rewrites a compact profile and trims the old dialogue.
6. Save the automation, Test multiple viewers separately and inspect the created files through Data Explorer.

Verify

Replies receive the intended bounded context while separate viewers keep separate dialogue/profile files. Inspect the corresponding run, then Save the applied project when the result is correct.

If it does not work

Review missing files, viewer identity and concurrent file writes. Sending context to a hosted provider can disclose the selected viewer messages; trim and review the content deliberately.