docs: define MinerU practices roadmap

docs/case-studies/issue-143/data/initial-ci-runs.json

@@ -0,0 +1 @@
+[{"conclusion":"success","createdAt":"2026-07-27T04:35:53Z","databaseId":30237547903,"headSha":"432a8ecee2d651b334d7eaa9f93f53a82f6af363","workflowName":"Rust Checks and Release"},{"conclusion":"success","createdAt":"2026-07-27T04:35:53Z","databaseId":30237547904,"headSha":"432a8ecee2d651b334d7eaa9f93f53a82f6af363","workflowName":"JavaScript Checks and Release"},{"conclusion":"success","createdAt":"2026-07-27T04:35:53Z","databaseId":30237547898,"headSha":"432a8ecee2d651b334d7eaa9f93f53a82f6af363","workflowName":"JS/Rust Parity"}]

docs/case-studies/issue-143/data/issue-143-comments.json

@@ -0,0 +1 @@
+[]

docs/case-studies/issue-143/data/issue-143.json

@@ -0,0 +1 @@
+{"author":{"login":"konard"},"body":"","createdAt":"2026-07-14T16:46:43Z","labels":["documentation","enhancement"],"number":143,"state":"OPEN","title":"Use all best practices from https://github.com/opendatalab/mineru but in Rust/JavaScript and using Unlicense (Public Domain)","updatedAt":"2026-07-14T16:46:43Z","url":"https://github.com/link-assistant/web-capture/issues/143"}

docs/case-studies/issue-143/data/mineru-latest-release.json

@@ -0,0 +1 @@
+{"body":"fix: enhance PDF font analysis by adding Latin and CJK font usage detection; fix #5238: add offset duplicate character detection for improved text rendering","html_url":"https://github.com/opendatalab/MinerU/releases/tag/mineru-3.4.4-released","name":"mineru-3.4.4-released","published_at":"2026-07-10T11:54:35Z","tag_name":"mineru-3.4.4-released","target_commitish":"master"}

docs/case-studies/issue-143/data/mineru-repository.json

@@ -0,0 +1 @@
+{"created_at":"2024-02-29T08:52:34Z","default_branch":"master","description":"Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.","forks_count":6363,"full_name":"opendatalab/MinerU","html_url":"https://github.com/opendatalab/MinerU","license":{"key":"other","name":"Other","spdx_id":"NOASSERTION"},"open_issues_count":65,"pushed_at":"2026-07-25T03:52:10Z","snapshot_commit":"79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7","snapshot_date":"2026-07-27","stargazers_count":75808,"topics":["ai4science","document-analysis","docx","extract-data","layout-analysis","ocr","parser","pdf","pdf-converter","pdf-extractor-llm","pdf-extractor-pretrain","pdf-extractor-rag","pdf-parser","pptx","python","xlsx"],"updated_at":"2026-07-27T04:20:03Z"}

docs/case-studies/issue-143/data/omnidocbench-repository.json

@@ -0,0 +1 @@
+{"default_branch":"main","description":"[CVPR 2025] A Comprehensive Benchmark for Document Parsing and Evaluation","full_name":"opendatalab/OmniDocBench","html_url":"https://github.com/opendatalab/OmniDocBench","license":{"key":"apache-2.0","name":"Apache License 2.0","spdx_id":"Apache-2.0"},"pushed_at":"2026-07-24T07:14:06Z","snapshot_date":"2026-07-27","stargazers_count":1915,"updated_at":"2026-07-25T21:34:26Z"}

docs/case-studies/issue-143/data/pr-146-comments.json

@@ -0,0 +1 @@
+[]

docs/case-studies/issue-143/data/pr-146-review-comments.json

@@ -0,0 +1 @@
+[]

docs/case-studies/issue-143/data/pr-146-reviews.json

@@ -0,0 +1 @@
+[]

docs/case-studies/issue-143/data/pr-146.json

@@ -0,0 +1 @@
+{"baseRefName":"main","headRefName":"issue-143-614fca5161bb","isDraft":true,"number":146,"state":"OPEN","title":"[WIP] Use all best practices from https://github.com/opendatalab/mineru but in Rust/JavaScript and using Unlicense (Public Domain)","url":"https://github.com/link-assistant/web-capture/pull/146"}

docs/case-studies/issue-143/data/recent-merged-prs.json

@@ -0,0 +1 @@
+[{"mergedAt":"2026-06-25T23:06:49Z","number":142,"title":"feat: add shared dialog capture","url":"https://github.com/link-assistant/web-capture/pull/142"},{"mergedAt":"2026-06-14T21:13:55Z","number":140,"title":"fix(ci): repair merge detection and Rust Docker build","url":"https://github.com/link-assistant/web-capture/pull/140"},{"mergedAt":"2026-06-14T16:12:36Z","number":138,"title":"fix(rust): repair fresh dependency resolution (#137)","url":"https://github.com/link-assistant/web-capture/pull/138"},{"mergedAt":"2026-06-12T16:43:00Z","number":136,"title":"docs: add FormalAI capture contract","url":"https://github.com/link-assistant/web-capture/pull/136"},{"mergedAt":"2026-05-30T10:03:09Z","number":134,"title":"Add Browserbase comparison case study","url":"https://github.com/link-assistant/web-capture/pull/134"}]

docs/case-studies/issue-143/data/web-capture-repository.json

@@ -0,0 +1 @@
+{"created_at":"2025-05-06T21:59:30Z","default_branch":"main","description":"A library, CLI and microservice to capture the web in required format","forks_count":0,"full_name":"link-assistant/web-capture","html_url":"https://github.com/link-assistant/web-capture","js_version":"1.10.10","license":{"key":"unlicense","name":"The Unlicense","spdx_id":"Unlicense"},"open_issues_count":4,"rust_version":"0.3.32","snapshot_commit":"4c5443e1ab83a4a250457933079eaf2fbd61ed21","snapshot_date":"2026-07-27","stargazers_count":0}

docs/case-studies/issue-143/online-sources.md

@@ -0,0 +1,72 @@
+# Online sources
+
+Research snapshot: 2026-07-27.
+
+The analysis in this case study is pinned to MinerU commit
+[`79d6d8d`](https://github.com/opendatalab/MinerU/commit/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7)
+and web-capture commit
+[`4c5443e`](https://github.com/link-assistant/web-capture/commit/4c5443e1ab83a4a250457933079eaf2fbd61ed21).
+That pin matters because both projects evolve quickly.
+
+## Primary MinerU sources
+
+| Source                     | URL                                                                                                                            | Used for                                                                                       |
+| -------------------------- | ------------------------------------------------------------------------------------------------------------------------------ | ---------------------------------------------------------------------------------------------- |
+| Repository                 | <https://github.com/opendatalab/MinerU>                                                                                        | Product scope, supported inputs, backends, platforms, CLI, API, deployment, and project layout |
+| README at snapshot         | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/README.md>                                | Feature inventory and release history                                                          |
+| License at snapshot        | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/LICENSE.md>                               | Custom Apache-2.0-derived terms, thresholds, and attribution obligation                        |
+| Package manifest           | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/pyproject.toml>                           | Optional dependency groups, supported Python versions, entry points, and test configuration    |
+| Quick usage                | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/docs/en/usage/quick_usage.md>             | Local/remote execution, async tasks, health, router, and model configuration                   |
+| CLI tools                  | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/docs/en/usage/cli_tools.md>               | Input, page-range, backend, output, OCR, formula, and table controls                           |
+| Advanced CLI               | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/docs/en/usage/advanced_cli_parameters.md> | Accelerator selection and inference-engine pass-through                                        |
+| Output files               | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/docs/en/reference/output_files.md>        | Intermediate representation, content list, debug overlays, coordinates, and artifacts          |
+| Model source configuration | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/docs/en/usage/model_source.md>            | Cache reuse, automatic source selection, and offline model paths                               |
+| Backend directory          | <https://github.com/opendatalab/MinerU/tree/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/mineru/backend>                           | Pipeline, VLM, hybrid, Office, and shared output stages                                        |
+| API implementation         | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/mineru/cli/fast_api.py>                   | Task lifecycle, limits, result packaging, and service boundary                                 |
+| Router implementation      | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/mineru/cli/router.py>                     | Multi-worker routing and local/remote worker management                                        |
+| Public HTTP client policy  | <https://github.com/opendatalab/MinerU/blob/79d6d8d79fb8f3ddba5cc34c07a16f0ec36f56c7/mineru/cli/public_http_client_policy.py>  | Public-bind guard for caller-selected inference endpoints and SSRF considerations              |
+| Latest release             | <https://github.com/opendatalab/MinerU/releases/tag/mineru-3.4.4-released>                                                     | Version and recent font/duplicate-glyph hardening                                              |
+
+## Benchmarks and technical reports
+
+| Source                  | URL                                           | Used for                                                           |
+| ----------------------- | --------------------------------------------- | ------------------------------------------------------------------ |
+| MinerU technical report | <https://arxiv.org/abs/2409.18839>            | Original pipeline motivation and evaluation context                |
+| MinerU 2.5 report       | <https://arxiv.org/abs/2509.22186>            | VLM architecture evolution                                         |
+| OmniDocBench repository | <https://github.com/opendatalab/OmniDocBench> | Reproducible evaluation dimensions and current dataset information |
+| OmniDocBench paper      | <https://arxiv.org/abs/2412.07626>            | Benchmark design and annotations                                   |
+| OHR-Bench repository    | <https://github.com/opendatalab/OHR-Bench>    | Downstream RAG impact of OCR errors                                |
+
+Benchmark scores quoted by MinerU are upstream claims, not results reproduced by
+this case study. The roadmap requires web-capture to publish its own pinned
+results before making comparative accuracy claims.
+
+## web-capture sources
+
+| Source                      | URL or path                                                                     | Used for                                                             |
+| --------------------------- | ------------------------------------------------------------------------------- | -------------------------------------------------------------------- |
+| Repository                  | <https://github.com/link-assistant/web-capture>                                 | Project purpose and current code                                     |
+| Root README                 | [`README.md`](../../../README.md)                                               | Current formats, endpoints, image policies, and parity promise       |
+| Architecture                | [`ARCHITECTURE.md`](../../../ARCHITECTURE.md)                                   | Current browser pipeline, deployment, security notes, and known gaps |
+| JavaScript package          | [`js/package.json`](../../../js/package.json)                                   | Runtime, dependencies, checks, version, and Unlicense declaration    |
+| Rust package                | [`rust/Cargo.toml`](../../../rust/Cargo.toml)                                   | Runtime, dependencies, lints, version, and Unlicense declaration     |
+| JS README                   | [`js/README.md`](../../../js/README.md)                                         | JS CLI/API contract and development commands                         |
+| Rust README                 | [`rust/README.md`](../../../rust/README.md)                                     | Rust CLI/API contract and development commands                       |
+| Parity script               | [`scripts/check-js-rust-parity.mjs`](../../../scripts/check-js-rust-parity.mjs) | Existing source/test parity guard                                    |
+| Browserbase precedent       | [`docs/case-studies/issue-127/README.md`](../issue-127/README.md)               | Comparison-study scope and evidence style                            |
+| FormalAI contract precedent | <https://github.com/link-assistant/web-capture/pull/136>                        | Contract-first documentation precedent                               |
+
+## Archived metadata
+
+The `data/` directory preserves compact, machine-readable snapshots for:
+
+- issue 143 and its comments;
+- PR 146 conversation comments, review comments, and reviews;
+- MinerU repository and latest release metadata;
+- OmniDocBench and web-capture repository metadata;
+- recent merged web-capture PRs; and
+- the initial CI runs on PR 146.
+
+No screenshots or issue attachments existed to download. No authenticated MinerU
+cloud or model-backed run was attempted: the issue asks for transferable
+practices, and a model download would not validate a Rust/JavaScript design.