docs: define MinerU practices roadmap

[konard]

Summary

• audits MinerU 3.4.4 at pinned commit 79d6d8d against the current Rust and JavaScript web-capture implementations
• turns the issue's broad "all best practices" request into 20 finite, testable practices with current-state mappings and acceptance gates
• proposes a clean-room, Unlicense-compatible CaptureDocument v1 architecture and a seven-phase delivery roadmap
• preserves the issue, PR, release, repository, CI, and related-work metadata used by the research
• keeps Cargo.lock synchronized with Rust package version 0.3.33 after merging the latest main

Root cause and scope

MinerU parses local documents with OCR/layout/model pipelines, while web-capture acquires rendered URLs. The foundational gap is therefore not a missing parser dependency; it is the absence of a shared typed document model between acquisition and serialization.

This PR is intentionally contract-first, matching the issue's documentation and enhancement labels. It does not copy MinerU code or artifacts, add an ML runtime, or expose an untested parser API. Follow-up implementation starts with shared Rust/JavaScript schema fixtures and golden tests before adding optional document or OCR backends.

Evidence

• Reproduction: issue Use all best practices from https://github.com/opendatalab/mineru but in Rust/JavaScript and using Unlicense (Public Domain) #143 has a title but no body, comments, screenshots, or acceptance test.
• Automated gates defined: each BP-01 through BP-20 entry includes a verification gate; the acceptance matrix makes the research scope finite.
• Provenance: online sources are pinned and compact API snapshots are archived under docs/case-studies/issue-143/data/.
• Licensing: the plan treats MinerU's custom license, models, datasets, and fixtures as separate review boundaries and keeps original project code Unlicense-compatible.
• CI diagnosis: Rust workflow run 30238377295 failed at rust/src/main.rs:211; the failure was reproduced locally with Rust 1.97.1. The equivalent compatibility fix landed on main in fix(js): remove deprecated glob dependency #145 and is included here by merge.

Validation

• Prettier 3.7.4 check for the Markdown documents
• local Markdown relative-link check
• parse all archived JSON with Node.js
• node scripts/check-js-rust-parity.mjs
• npm run lint (0 errors; 30 pre-existing warnings)
• npm run format:check
• npm run check:duplication
• npm test -- --runInBand --testPathIgnorePatterns=tests/e2e/docker.test.js (55 suites and 493 tests passed; 4 suites/42 Docker tests skipped)
• Node 22 targeted API integration suite after merging main (18 passed)
• cargo +1.97.1 fmt --all -- --check
• cargo +1.97.1 clippy --all-targets --all-features -- -D warnings
• cargo +1.97.1 test --all-features (267 executable tests and 2 doc tests passed)

Fixes #143

[konard]

Working session summary

Implemented and finalized PR #146.

• Added a pinned MinerU 3.4.4 audit with 20 testable practices, licensing boundaries, CaptureDocument v1, and a seven-phase Rust/JavaScript roadmap.
• Preserved research sources and API snapshots.
• Merged the latest main and synchronized Rust 0.3.33 lock metadata.
• Updated the PR title/description and marked it ready for review.
• Local validation: 493 JavaScript tests and 269 Rust tests passed.
• Final SHA fa24ce4: JavaScript, Rust, and parity CI all pass.
• PR is cleanly mergeable; working tree is clean.

Main document: README.md

---

This summary was automatically extracted from the AI working session output.

[konard]

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

• Model: GPT-5.6 Sol
• Provider: OpenAI
• Public pricing estimate: $28.419403

📊 Context and tokens usage:

• 499.4K / 200K (250%) input tokens, 56.6K / 128K (44%) output tokens

Total: (499.4K + 20.9M cached) input tokens, 56.6K output tokens, $28.419403 cost

🤖 Models used:

• Tool: OpenAI Codex
• Requested: gpt-5.6-sol
• Thinking level: medium (~15999 tokens)
• Model: GPT 5.6 Sol (gpt-5.6-sol)

📎 Log file uploaded as Gist (7868KB)

• View complete solution draft log

---

Now working session is ended, feel free to review and add any feedback on the solution draft.

[konard]

🎉 Auto-merged

This pull request has been automatically merged by hive-mind.

• All CI checks have passed

---

Auto-merged by hive-mind with --auto-merge flag