[CEPBEP]

Summary

This PR hardens the GLAM computational pipeline and promotes the qFoldIT adapter from an execution-status scaffold to a provenance-bound run manifest.

Changes

• Deterministic random/scaffold splits with local seeded RNGs; split-cache keys bind to source dataset SHA-256.
• Explicit split propagation through dataset constructors and worker config.
• Perturbation datasets can defer parent split initialization.
• Seeded architecture search and stable SHA-256 configuration fingerprints.
• High-fidelity execution uses argv and shell=False; worker failures propagate.
• Training config is parsed with ast.literal_eval, not executable eval.
• Per-job output isolation and required high-fidelity/model-selection artifacts.
• qfoldit.glam.run-result/1.0 manifest with mission/action provenance, dataset digest, runtime versions and artifact hashes.
• Request/response/result schemas and CI checks.

Authority boundary

GLAM remains a computational engine. scientific_validation_status=NOT_PERFORMED and evidence_status=NOT_CREATED.

CI checks adapter tests, compilation, JSON schemas, split/worker invariants and artifact-manifest requirements. End-to-end training on a prepared production dataset and locked holdout remains a separate release gate.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	⚠️ Failed 2026-11-02T11:30:54.908294Z We couldn’t complete this review. Comment @codex review to try again.	003d473	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.