paper-parser-skill（论文解析技巧）

Paper Parser Skill

CLI tool for automated academic paper processing.

🛡️ Data Privacy & Security

[!IMPORTANT] External Data Processing: This skill transmits PDF files and paper metadata to MinerU (opendatalab) for layout analysis and Markdown conversion. Please ensure you trust the service and understand their data handling policies before providing an API token in the configuration file.

Security & Provenance:

• Open Source: The full source code is available on GitHub.
• Verified Package: This tool is published on PyPI as a standard Python package.
• Local Control: All search results and downloaded PDFs are stored locally in your specified workspace.

Before Installing:

1. Review the source code at GitHub for unexpected behavior.
2. Understand that installing from PyPI executes third-party code on your system — use a virtual environment or container to limit blast radius if desired.
3. The MINERU_API_TOKEN grants MinerU access to receive and process uploaded PDFs — use a dedicated, revocable token with minimal scope.
4. Avoid uploading sensitive, unpublished, or confidential documents to MinerU — review their privacy and data retention policies.
5. For sensitive documents requiring offline parsing, consider local alternatives that do not transmit PDFs externally.

🚀 Setup

[!WARNING] Installing from PyPI executes third-party code. Use a virtual environment if you want to limit blast radius.

pip install paper-parser-skill==v0.1.3

⚙️ Configuration

Default path: ~/.paper-parser/config.yaml

[!IMPORTANT] MINERU_API_TOKEN is required for parsing functionality. Get a token at mineru.net.

PAPER_WORKSPACE: "~/paper-parser-workspace"
MINERU_API_TOKEN: "your_token_here"  # Required for parsing
MINERU_API_BASE_URL: "https://mineru.net/api/v4"
MINERU_API_TIMEOUT: 600

📖 CLI Usage

Alias: pp

Basic Commands

Parsing Commands

[!TIP] Recommended for agent/automation use: pp submit + pp check (non-blocking async workflow). pp parse and pp all block the process until cloud processing completes, which can take several minutes and may time out. Prefer the async approach when calling from an agent or pipeline.

Recommended Async Workflow (for agents)

# Step 1: Search for papers by keyword → pick an arXiv ID from results
pp search "retrieval augmented generation"
# → 1. Id: 2312.10997  Title: Retrieval-Augmented Generation for ...
# → 2. Id: 2401.00123  Title: ...

# Step 2: Submit for parsing and return immediately
#         (PDF is downloaded automatically if not already cached)
pp submit 2312.10997
# → ⬇️  Downloading PDF...
# → ✅ Submitted!  batch_id: xxxxxxxx

# (minutes later, the agent or user calls again)

# Step 3: Check status — downloads & extracts results automatically when done
pp check 2312.10997
# → "⏳ Still processing (state: running, 45s since submission)."
# → or "✅ Parsing complete!  📂 Results in: ~/paper-parser-workspace/2312.10997"

📂 Workspace Structure

PAPER_WORKSPACE/
└── <arxiv_id>/
    ├── paper.pdf
    ├── title.md
    ├── summary.md
    ├── .parse_task.json       ← async task state (batch_id, status, timestamps)
    └── markdowns/
        ├── 01_Introduction.md
        └── images/

🛠️ Requirements

• Python >= 3.8
• requests, click, PyYAML, arxiv, rapidfuzz
• MinerU API Token: Required for the parsing stage. Add it to your config.yaml file. Get one at mineru.net.