Skip to content

v0.15.2 — Resilient model calls (transient retry)

Choose a tag to compare

View all tags
@z-fab z-fab released this 06 Jun 00:50
· 149 commits to master since this release
v0.15.2
ed01c8c

Fixed

  • Transient transport drops no longer kill a run — model calls now retry on transport-level errors (e.g. RemoteProtocolError: Server disconnected without sending a response, a flaky preview endpoint dropping the connection) up to 3 times with exponential backoff, instead of failing the whole execution. This was most visible during HILT: a dropped connection on resume would error the execution and the conversation only "continued" via history-seeding on a manual re-run. Genuine API errors (bad request, auth) are not retried. Applies to both run and resume.
  • Resume errors now log a traceback — AgentRunner.resume() logs the full exception (parity with run()), making backend-side failures diagnosable in backend.log.

Full changelog: https://github.com/z-fab/agentmd/blob/master/CHANGELOG.md

Assets 2
Loading