Slack now answers messages whose first delivery was lost

The Agno Slack interface in AgentOS now remembers the event_id of each event it receives and uses it to sort Slack's retries. Slack retries any event it can't confirm within 3 seconds and marks the retry with X-Slack-Retry-Num. The Slack interface used to treat every retry as a duplicate and drop it. When the original delivery never reached AgentOS, the user's message got no reply and no error.

Now the Slack interface skips a retry only when it has already seen that ID. A retry for an event it hasn't seen runs like a first delivery.

The Slack interface now acknowledges a signed request whose body isn't valid JSON and logs it at warning level. It used to return 500, which made Slack retry the request.

The Slack interface keeps seen event IDs in memory for each process, for ten minutes. If you run several workers or replicas, a retry that lands on a different process can run a second time. Run one process per Slack app to avoid duplicate replies.

See the cookbook, and learn more about the Slack interface in the documentation.