docs: complete D1.4a second-book evidence and extraction contract - #33
Conversation
sunpole
changed the title
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 766ec19126
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".
| "warnings": [] | ||
| } | ||
| ], | ||
| "full_output_sha256": "deterministic aggregate hash", |
There was a problem hiding this comment.
Specify exactly which bytes feed full_output_sha256—including page order, separators, filename inclusion, and encoding. In the D1.4b pilot, concatenating page contents, concatenating page hashes, or hashing serialized manifest data would all satisfy the current wording while producing different values for the same extraction, so the raw-extraction gate cannot actually require a reproducible aggregate hash.
AGENTS.md reference: AGENTS.md:L145-L149
Useful? React with 👍 / 👎.
| ] | ||
| }, | ||
| "text_layer": { | ||
| "tool": "pdftotext -layout (poppler-utils)", |
There was a problem hiding this comment.
Replace this generic tool label with the exact pdftotext/Poppler version and execution platform used by the one-time workflow. Because that workflow installed whatever ubuntu-latest provided and the evidence asserts exact layout-derived character, word, and line counts, a future reproduction with a different Poppler release cannot distinguish tool-output drift from incorrect evidence.
Useful? React with 👍 / 👎.
| ## Documentation, data and extraction baseline | ||
|
|
||
| `v23.8.12` |
There was a problem hiding this comment.
Add the v23.8.12 milestone to CHANGELOG.md and map the new evidence, contract, and validator in docs/FILE_MAP.md. This line advances the maintained baseline to v23.8.12, but the changelog still ends at v23.8.10 and the file map omits every newly introduced D1.4a artifact, leaving the repository's maintained history and file inventory inconsistent with VERSION.md and ROADMAP.md.
AGENTS.md reference: AGENTS.md:L199-L204
Useful? React with 👍 / 👎.
1 participant
Цель
Завершить D1.4a: доказать техническую извлекаемость второй сохранённой книги и определить безопасный путь к отдельной базе без изменения текущего runtime, PWA и существующих данных.
Доказанный результат
_archive/source-files/Unlocking-Your-Dream-Student-Ma.pdf;740193bytes;edb4127915bf720ee56c96612f825ea4e0ef2f5fc15192a56fb96e5c7ec4745a;55;55;1.0;85103characters,12479words;direct-text-extraction-viable.Добавлено
docs/SECOND_BOOK_EVIDENCE.md;docs/second-book-evidence.json;docs/SECOND_BOOK_EXTRACTION_CONTRACT.md;scripts/validate-second-book-evidence.mjs;23.8.12вVERSION.mdиROADMAP.md;Безопасный контракт
ru-current-v1остаётся default и stable fallback.Границы
Не изменялись:
data/файлы и 4 086 записей;_archive/;versions/;Временные analysis/capture workflows удалены из итогового diff. Полный извлечённый текст книги не сохранён в репозитории.
Проверки
GitHub Actions run №145 прошёл полностью:
Следующий этап
D1.4b — private/non-public immutable raw-extraction pilot с page files, manifest и небольшой reviewed segmentation sample. Новый logical dataset в этом PR не создаётся.
Closes #32.