Revenue-Critical Fix

Revenue Action Timeout Resolver Sprint

Stop 120-second TimeoutErrors from destroying your MiniMax revenue actions. Deterministic diagnosis, concrete artefacts, 5 business days.

What you get

• 1. Timeout Diagnostic Report (PDF, 18–22 pages) Structured analysis of your MiniMax API timeout pattern. Identifies whether the hang originates in slow-read loops on tool calls, bash execution, or sessionTarget=isolated contexts. References your specific version regressions (v2026.3.2, v2026.3.23-2). Includes latency breakdown charts and root-cause hypothesis ranked by confidence.
• 2. Session Replay Fixture (Python, deterministic) Standalone test case that reproduces the exact 120s timeout condition in a controlled environment. Isolates whether the failure triggers via direct api.minimax.io, openrouter proxy, or both. You can run this locally to validate any fix against a known failure signature.
• 3. Timeout Configuration Patch (YAML + JSON) Production-ready configuration values replacing your current timeoutMs thresholds. Includes dynamic threshold logic that adapts timeout based on action type (tool call vs. bash vs. chat completion). Compatible with OpenClaw session configs and direct API deployments.
• 4. Routing Strategy Document (Markdown) Decision tree for routing revenue-critical actions through openrouter vs. direct MiniMax endpoint. Covers fallback triggers, latency budgets per action type, and automated failover thresholds. Designed to prevent future revenue actions from hitting the direct endpoint during regression windows.
• 5. Regression Test Suite (pytest + fixture bundle) Automated test suite you can integrate into CI/CD to catch timeout regressions before deployment. Includes test cases for each action type that previously triggered 120s hangs, with pass/fail assertions tied to your new timeout configuration. Validates both direct and proxy endpoint behavior.

How it works

FAQ

My timeouts happen intermittently — will this still help?

Yes. The session replay fixture creates a deterministic reproduction of your intermittent failure, eliminating the need to wait for the next occurrence. The fixture tests multiple conditions (action type, endpoint, session mode) so the failure doesn't need to happen live for us to diagnose it.

We're on OpenClaw — does that change anything?

No. The diagnostic explicitly references OpenClaw-specific issues including GitHub #34487 and #54004 which document the v2026.3.2 and v2026.3.23-2 regressions in OpenClaw contexts. The configuration patches target OpenClaw's sessionTarget and timeoutMs parameters directly.

What if the fix requires more than timeout tuning?

The timeout configuration patch is the first actionable solution. If the diagnostic reveals the problem requires architectural changes (e.g., mandatory openrouter routing, sessionTarget adjustments), those recommendations are documented in the Routing Strategy Document. The sprint is scoped to deliver artefacts, not ongoing consulting — if deeper remediation is needed, that becomes a separate engagement.

Is there a warranty period?

You receive 14 days of artefact support after delivery. If the regression test suite fails against your environment due to a configuration issue in the delivered artefacts, I provide corrected versions at no additional charge. If your codebase changes (new action types, different MiniMax versions), that's outside scope but can be scoped separately.