feat(artifacts): add strict BM25 publication

[fishmingyu]

Summary

Restacks the isolated strict BM25 publication producer on current main. It plans and replays exact canonical BM25 generations through retained source and strict workspace authority without changing legacy SourceTrust or native-vector production routes.

Changes

• Add PlannedBm25View plus strict plan, replay, and high-level normalization APIs.
• Bind source and output records, repository fingerprint, and caller configuration into one exact workspace subject.
• Capture one detached authenticated repository identity for the complete high-level operation so mutable public projections cannot redirect policy, paths, or fingerprints.
• Enforce bounded UTF-8 JSON, duplicate and nonfinite rejection, exact inventories, repository separation, and staged and published verification.
• Export the isolated API and update the storage roadmap; no production provider or compiler route is enabled.
• Add adversarial coverage for overlap aliases, repository drift rollback, provider mismatches, forged plans, source drift, identity ABA, and BM25 query compatibility.

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Refactoring
• Performance improvement
• Tests

Testing

• Tests pass locally
• Added new tests for the changes
• Focused publication, source, provider, and artifact suites: 800 passed, 15 skipped.
• Unit tier under umask 022: 5,490 passed, 71 skipped, 191 deselected; the sole additional deselection is the locally installed 0.2.0 distribution metadata mismatch against source version 0.2.1, already reproduced on clean main and covered by clean CI.
• Exact-head Full CI light: 5,492 unit tests passed and 36 integration tests passed.
• Exact-head formal PR CI: 5,490 unit tests passed.
• Exact-head Docs and CodeNib Publish Smoke passed. Every release artifact build, Python 3.10-3.14 install, upgrade, CodeGraph, Wiki/MCP, and Ask verification job passed; the workflow-dispatch-only MCP Registry environment gate rejected this non-tag branch before running steps and is not a PR check.
• Pre-commit, py_compile, git diff --check, and python -m mkdocs build --strict.

Checklist

• My code follows the project style guidelines
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• My changes generate no new warnings
• Any dependent changes have been merged and published

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b3fdb8bd5d

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Raise the aggregate budget for streamed BM25 documents

When a generation contains 20,000 otherwise valid documents, this whole-file preflight rejects it with exceeds its 100000-node limit even if the JSON is only about 1.2 MB. The call inherits DEFAULT_MAX_NODES_PER_ELEMENT and DEFAULT_MAX_LEXICAL_TOKENS, but applies those limits to the entire top-level array before iter_bounded_json_array performs the intended per-element checks, so sufficiently large repositories cannot use strict BM25 publication despite the documented 256 MB file budget and one-million-item iterator limit. Use aggregate limits consistent with the allowed array size, or restrict this preflight to the whole-file lexical properties it actually needs to check.

Useful? React with 👍 / 👎.