fix(strix): remove the 300s LLM_TIMEOUT cap

[seonghobae]

Description

.github/workflows/strix.yml's Run Strix (quick) step sets four inference-related timeout env vars together (line 786-789). Three are already 0: STRIX_MEMORY_COMPRESSOR_TIMEOUT, STRIX_PROCESS_TIMEOUT_SECONDS, STRIX_TOTAL_TIMEOUT_SECONDS. LLM_TIMEOUT was left hardcoded at 300 (5 minutes) — contradicting both its three siblings and the job's own concurrency-block comment a few lines above:

Large, actively-growing repositories (e.g. contextual-orchestrator) can legitimately require well over two hours to scan -- this org's own standing operating directive accepts that central OpenCode/Strix/Noema scans may take more than two hours per model (docs/product-goal-directive.md). Inference has no wall-clock deadline; cancellation is reserved for an explicit operator action or a superseded head.

docs/product-goal-directive.md §8 is explicit on this: no uniform application/agent/gateway-wide LLM timeout ceiling; default unlimited, with admin-configurable per-model overrides only.

scripts/ci/test_strix_quick_gate.sh already asserts export LLM_TIMEOUT=0 (line 302, "strix disables the model client inference timeout"). Confirmed this assertion currently fails on main itself (fb02129), independent of any other PR's diff — a real, pre-existing base-branch contract violation, not something introduced by this change.

Fix

One line: export LLM_TIMEOUT=300 → export LLM_TIMEOUT=0, matching the three sibling env vars on the following lines.

Type of change

• Bug fix (non-breaking change which fixes an issue)

Checklist

• Confirmed the failing assertion against origin/main directly (git show origin/main:.github/workflows/strix.yml | grep LLM_TIMEOUT → export LLM_TIMEOUT=300) before making this change
• Diff is exactly one line (git diff origin/main → 1 file changed, 1 insertion(+), 1 deletion(-))
• No other LLM_TIMEOUT references in the file need changing (the assert_file_not_contains ... "LLM_TIMEOUT:" check in the same test guards against a different pattern — a YAML env: mapping key — which this change does not introduce)

🤖 Generated with Claude Code

https://claude.ai/code/session_019YgmaEKnPhrsbSyQBierFH

---

Generated by Claude Code

---

[coderabbitai]

Warning

Review limit reached

Next included review available in 57 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Team

Run ID: 090a0885-01fe-449e-8486-58cb7e96b42c

📥 Commits

Reviewing files that changed from the base of the PR and between fb02129 and 3196edd.

📒 Files selected for processing (1)

• .github/workflows/strix.yml

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[devin-ai-integration]

Devin Review found 0 potential issues.

[seonghobae]

QUEUE_SATURATION_CHICKEN_EGG merge decision for exact head 3196edde85ed7f4a909c3a627af75b47593c7f5e.

Fresh verification immediately before merge:

• GitHub reports the PR Ready and mechanically mergeable with a one-file / one-line effective delta: LLM_TIMEOUT=300 -> LLM_TIMEOUT=0 in the trusted Strix workflow.
• Protected main@fb021296afbe7c27e30363627971fc9d36d12979 still carries the defect, while the already-existing executable Strix contract asserts export LLM_TIMEOUT=0; the PR therefore repairs a pre-existing RED contract rather than weakening a test.
• Exact-head Devin Review reports 0 potential issues; the exact-head Devin and CodeRabbit commit statuses are both success; REST inline-review comments are empty.
• Exact-head repository/security workflows are queued rather than failing: OSV, SBOM, Security Scan, Semgrep, CodeQL, Python Security, Strix Changed Path Quality, Secret Scan and Scorecard are all waiting for runner admission.
• The repository currently has 1,252 queued Actions runs, so the remaining admission evidence cannot start promptly because of the central runner backlog. This one-line repair removes a hard 300-second inference ceiling from the required Strix control plane and is independently verified by the pre-existing failing contract.

No substantive failing test, security finding, meaningful CHANGES_REQUESTED, malformed provenance, merge conflict, or unrelated policy defect is being bypassed. Merge is bound to the exact head SHA; predecessor evidence is not transferred.