#agents.defaults.timeoutSeconds

1 messages · Page 1 of 1 (latest)

hollow surge
#

Is there a provider or general best practice regarding setting agents.defaults.timeoutSeconds beyond the default value?

#

@red terrace

red terraceBOT
#

Short answer: generally no, I would not increase agents.defaults.timeoutSeconds beyond the default.

A couple important bits from the docs/source:

  • agents.defaults.timeoutSeconds is the overall agent run timeout, and the default is already 172800s (48 hours), which is huge.
  • The thing that usually matters more for “slow provider” behavior is agents.defaults.llm.idleTimeoutSeconds, which aborts when the model stops sending chunks for too long.
  • If llm.idleTimeoutSeconds is unset, OpenClaw will fall back to agents.defaults.timeoutSeconds when you configured it, otherwise 120s.

So the practical best practice is:

  1. Leave agents.defaults.timeoutSeconds alone unless you truly want very long autonomous runs.
  2. If a provider/model is just slow or pauses a lot, tune agents.defaults.llm.idleTimeoutSeconds instead.
  3. Prefer per-use overrides for special cases, like heartbeat/cron/session-specific timeouts, instead of making the global default enormous.

That fallback behavior is the main gotcha: bumping agents.defaults.timeoutSeconds can also make stuck LLM calls hang much longer unless you set agents.defaults.llm.idleTimeoutSeconds explicitly.

Docs: