mirror of
https://github.com/openclaw/openclaw.git
synced 2026-08-25 03:45:46 -06:00
17abdfc78c
* fix(system-agent): increase setup inference probe max tokens from 32 to 128 The 32-token cap starves reasoning models that spend tokens on thinking before emitting visible text. The probe gets a reasoning-only turn and reports healthy local models as broken inference. Increase to 128 tokens to accommodate reasoning model thinking overhead. Fixes #124345 * fix(system-agent): leave reasoning room in setup probes Co-authored-by: synthclaw <synthalorian@gmail.com> --------- Co-authored-by: Peter Steinberger <steipete@gmail.com>