Jazz AI Conversation
What It Models
A Jazz AI conversation is a titled message history, optionally filed in a project or attached to a query or directory. A project groups conversations and provides context for general replies. This design describes the console prototype while leaving durable ownership open.
Identity and Attributes
The console server generates prefixed identifiers for conversations, projects and messages. Scoped creation searches for an existing query identity or draft key. Every record uses the same sample owner, so those identifiers do not establish a real person or workspace boundary. Titles, archive times and update times support browsing.
Relationships
Live query refinement consumes metered query generation. Directory assistance uses local interpretation and name lookup. General chat sends a direct model request outside Platform accounting. The query workspace and directory retain their own state and save behavior.
Invariants
Scoped creation normally reuses and restores an existing matching conversation, but scope edits are not uniqueness constrained. Scoped conversations reject general replies. Project deletion keeps conversations and removes their filing. Archiving retains messages; message append and metadata writes remain available in the store while archived.
How It Is Stored and Shared
Each server process keeps a shared memory store seeded with sample records. Requests reaching that process see the same sample-owner records. Restart loses authored changes, and separate instances have independent copies. Platform stores no conversation history at this revision.
Open Design Questions
Product must choose durable ownership, personal or shared projects, retention, deletion, evidence handling and access policy. A proposed customer-ready boundary uses isolated durable storage and metered general replies with permission and AI-switch checks.