Lu_Wang_ENB_DBX
Databricks Employee
Databricks Employee

Genie agents are documented to support up to 10,000 conversations per agent and 10,000 messages per conversation, and the API is designed to retain prior messages as conversation context rather than silently omitting them.

More likely causes are:

  • You’re only listing the caller’s own conversations. GET /conversations returns only the requesting user’s conversations by default; to get all users’ conversations you must set include_all=true, and the caller needs CAN MANAGE on the agent.
  • include_all may not be honored in some clients if passed wrong. For example, Python boolean True could end up as "True" instead of "true", causing the API to return only a partial set.
  • If your dashboard pipeline uses audit logs as the source of truth, that can miss conversations depending on which actions you filter for. Missing Genie conversations could actually be present under different event names, and the fix was to include API actions such as genieGetConversationMessage in addition to UI events.

What to check:

  1. Verify you are pulling from GET /api/2.0/genie/spaces/{space_id}/conversations, not only from audit logs.
  2. If you need all users, send include_all=true and confirm the caller has CAN MANAGE.
  3. If calling from Python/SDK, inspect the raw request and make sure include_all is serialized exactly as expected.
  4. If you do rely on audit logs, broaden the event filters to include both UI and public API Genie actions.

View solution in original post