[aw-failures] [aw-fix] P1: Daily Cache Strategy Analyzer — Codex requests decommissioned model `gpt-5-codex-alpha-2025-11-07` (404, not retrie [Content truncated due to length]

[github-actions]

Pin the Codex request model to a supported name — the engine is resolving to the decommissioned alpha snapshot gpt-5-codex-alpha-2025-11-07, which 404s at the proxy and hard-fails the run before any agent work.

Parent: #46170

Problem statement

The Daily Cache Strategy Analyzer Codex run dies in Execute Codex CLI with a fatal 404 Not Found: Model not found gpt-5-codex-alpha-2025-11-07 from (apiproxy/redacted) The codex-harness classifies it isInvalidModelError=true and explicitly does **not** retry (invalid/unsupported model configuration — not retrying`), so the job exits 1 with no agent work. This is a deterministic config failure that will recur on every scheduled run until the model is fixed.

Affected workflows & run IDs

• Daily Cache Strategy Analyzer (.github/workflows/daily-cache-strategy-analyzer.lock.yml) — representative run §29604657433 (2026-07-17T18:39Z).
• No comparator success for this workflow in the 6h window; nearest prior failure is the same signature.

Probable root cause

The workflow's experiment spec (GH_AW_EXPERIMENT_SPEC) offers model_size variants gpt-5.4 / gpt-5.4-mini, wired to GH_AW_INFO_MODEL telemetry. But the actual Codex CLI request (engine codex v0.144.4) resolves to the hardcoded/default gpt-5-codex-alpha-2025-11-07 snapshot, which the api-proxy no longer serves. The experiment model_size value is not being propagated into the Codex request model — it only reaches telemetry, so the request falls back to the decommissioned alpha default.

Proposed remediation

• Propagate the resolved model_size (gpt-5.4 / gpt-5.4-mini) into the Codex CLI request model (--model / codex config), not just GH_AW_INFO_MODEL, so no request ever falls back to gpt-5-codex-alpha-*.
• Add a compile-time / startup validation that rejects unknown or alpha-snapshot Codex model names and fails fast with a clear message pointing at frontmatter.
• Recompile affected Codex workflows against the current supported model set and remove any references to gpt-5-codex-alpha-2025-11-07.

Success criteria / verification

• Execute Codex CLI for Daily Cache Strategy Analyzer no longer emits Model not found gpt-5-codex-alpha-2025-11-07.
• Next scheduled run reaches the agent with a supported model and non-zero token usage.
• No isInvalidModelError=true in codex-harness across 5 consecutive runs.

Related to #46170
Related to #46170

Generated by 🔍 [aw] Failure Investigator (6h) · 93 AIC · ⌖ 41.5 AIC · ⊞ 5.2K · ◷

• expires on Jul 24, 2026, 11:20 AM UTC-08:00

[github-actions]

🍪 Issue Monster selected this for Copilot

I identified this as a good candidate for automated resolution and requested assignment to the Copilot coding agent.

If assignment succeeds, the Copilot coding agent will analyze the issue and create a pull request with the fix.

Om nom nom! 🍪

🍪 Om nom nom by Issue Monster · 3.71 AIC · ⌖ 6.38 AIC · ⊞ 7.8K · ◷

[github-actions]

This issue was automatically closed because it expired on 2026-07-24T19:20:22.556Z.

Closed by Workflow