[CI failure] Paid OpenRouter on rage/96-versioned-syntax-facts #294

New issue

Closed

opened 2026-09-02 17:25:36 +00:00 by nsaspy · 1 comment

nsaspy commented

2026-09-02 17:25:36 +00:00

Owner

Automated prolog-rlm workflow failure report.

Workflow: Paid OpenRouter
Run: https://github.com/lost-rob0t/prolog-rlm/actions/runs/33546609366
Branch: rage/96-versioned-syntax-facts
Commit: bf13637af833011fab2e29da41df37b1b09c214c
Event: pull_request
Attempt: 1
Conclusion: failure
Pull request: #342

This issue is deduplicated by workflow + branch. Close it after the failure is resolved; a later regression will open a new issue.

Generated by .github/workflows/report-workflow-failures.yml.

Automated prolog-rlm workflow failure report. - Workflow: `Paid OpenRouter` - Run: https://github.com/lost-rob0t/prolog-rlm/actions/runs/33546609366 - Branch: `rage/96-versioned-syntax-facts` - Commit: `bf13637af833011fab2e29da41df37b1b09c214c` - Event: `pull_request` - Attempt: `1` - Conclusion: `failure` - Pull request: #342 This issue is deduplicated by workflow + branch. Close it after the failure is resolved; a later regression will open a new issue. _Generated by `.github/workflows/report-workflow-failures.yml`._

nsaspy closed this issue

2026-09-04 14:36:18 +00:00

nsaspy referenced this issue

2026-09-04 14:36:34 +00:00

[P1] Expose Tree-sitter query/capture APIs to Prolog #97

nsaspy commented

2026-09-04 14:37:07 +00:00

Author

Owner

Resolved by re-run + local replication at the exact PR head bf13637af833011fab2e29da41df37b1b09c214c (PR #342).

Root cause: stochastic live-model quality, not a diff regression. deep_constraint_openrouter passed 5/6; the single failure (core_minimal_depth_1) recorded root_decision: direct (no recursion attempted) and verification_status: rejected with 9 constraint violations — the verifier correctly rejected a wrong glm-5.3-flash answer. The PR diff (tree-sitter syntax projection) touches no planner/benchmark path.

Model pinning verified: every request in the failing run and the local replication used requested_model/selected_model = z-ai/glm-5.3-flash (repo variable OPENROUTER_TEST_MODEL; paid-openrouter.yml refuses unpinned/free fallback). No gpt-oss anywhere.

Evidence:

CI re-run on same head: Pinned paid OpenRouter integration → pass (https://github.com/lost-rob0t/prolog-rlm/actions/runs/33546609366)
Local replication of the exact failing step benchmark/run.pl -- deep-integration at bf13637: 6/6 passed, quality 1.00, all six cases passed verification at z-ai/glm-5.3-flash
Local exact-head gate: PlUnit 96 suites 1179/1179; deterministic 16/16; deep-experiment 15/15; demo --json pass; git diff --check clean; research-approval 19/19; nix flake check all checks passed (native 21/21 project syntax + FFI)

Resolved by re-run + local replication at the exact PR head `bf13637af833011fab2e29da41df37b1b09c214c` (PR #342). **Root cause**: stochastic live-model quality, not a diff regression. `deep_constraint_openrouter` passed 5/6; the single failure (`core_minimal_depth_1`) recorded `root_decision: direct` (no recursion attempted) and `verification_status: rejected` with 9 constraint violations — the verifier correctly rejected a wrong glm-5.3-flash answer. The PR diff (tree-sitter syntax projection) touches no planner/benchmark path. **Model pinning verified**: every request in the failing run and the local replication used `requested_model`/`selected_model` = `z-ai/glm-5.3-flash` (repo variable `OPENROUTER_TEST_MODEL`; `paid-openrouter.yml` refuses unpinned/free fallback). No `gpt-oss` anywhere. **Evidence**: - CI re-run on same head: Pinned paid OpenRouter integration → pass (https://github.com/lost-rob0t/prolog-rlm/actions/runs/33546609366) - Local replication of the exact failing step `benchmark/run.pl -- deep-integration` at `bf13637`: 6/6 passed, quality 1.00, all six cases passed verification at `z-ai/glm-5.3-flash` - Local exact-head gate: PlUnit 96 suites 1179/1179; deterministic 16/16; deep-experiment 15/15; demo --json pass; git diff --check clean; research-approval 19/19; `nix flake check` all checks passed (native 21/21 project syntax + FFI)

nsaspy referenced this issue from a commit

2026-09-05 03:17:41 +00:00

Merge pull request 'feat: implement #97 tree-sitter query pack and capture APIs' (#294) from rage/97-query-capture-apis into main