[Feature Request] Make AUTO_RECALL_TIMEOUT_MS configurable via config

[jlin53882]

Problem

When using the OpenClaw plugin's auto-recall feature, if the embedding model (e.g., qwen3-embedding:0.6b) runs on local CPU, it takes ~8-10 seconds per embedding call. This easily triggers the hardcoded 15s timeout in auto-recall.js, causing failures at ~6-7 seconds.

Suggestion

Add a configurable timeout option (e.g., autoRecallTimeoutMs) so users can adjust based on their hardware. Example config:

{
  "openviking": {
    "autoRecallTimeoutMs": 30000
  }
}

Environment

• OpenViking: 0.3.21 (local install)
• Deployment: Local Python server (non-Docker)
• Embedding: qwen3-embedding:0.6b on CPU
• OpenClaw: 2026.5.28

Related

• Issue [Bug]: memory-openviking plugin before_agent_start auto-recall hangs indefinitely, causing total silent agent failure #673 (closed): auto-recall hangs indefinitely — similar category, no external config
• This issue is about hardcoded timeout not adapting to different hardware

---

Reported by: OpenClaw user (2026-06-03)

[qin-ctx]

Thanks for reporting this!

Yes, this is a reasonable request. The OpenClaw plugin currently has an outer auto-recall timeout to prevent the agent from hanging if recall/search gets stuck, but that timeout is currently fixed at 5 seconds. That can be too short for local CPU embedding models.

We would support making this configurable, for example via autoRecallTimeoutMs, while keeping the default at 5000 ms so existing behavior remains unchanged.

A good PR would update:

• examples/openclaw-plugin/config.ts to add and validate autoRecallTimeoutMs
• examples/openclaw-plugin/auto-recall.ts to use cfg.autoRecallTimeoutMs instead of the hardcoded timeout
• examples/openclaw-plugin/openclaw.plugin.json so the option is exposed in the plugin schema/UI hints
• focused config/unit tests to cover the default and configured value

If you're interested in contributing this, we'd be happy to review the PR.