DeepPaperNote transforms dense research papers into structured Obsidian notes, automating the tedious work of evidence gathering and note production. It targets researchers and students building long-term knowledge bases, enabling them to focus on understanding complex material rather than mechanical tasks, resulting in higher-quality, reusable research assets.

How It Works

The project uses a "model-led understanding" and "evidence first" approach, prioritizing mechanism breakdown, experimental details, and limitations over simple summarization. It systematically gathers evidence from PDFs and metadata, plans figure integration, and leverages AI models for note generation, followed by quality linting. This ensures technically detailed, paper-type-aware notes grounded in source material, serving as valuable research assets.

Quick Start & Requirements

• Primary install / run command: Recommended: npx skills add 917Dhj/DeepPaperNote. Manual installation via release zip or git clone into agent skill directories (~/.codex/skills/, ~/.claude/skills/) is also supported.
• Non-default prerequisites and dependencies: Requires a compatible agent environment (e.g., Claude Code, Codex), Python 3.10+, and PyMuPDF (python3 -m pip install PyMuPDF). Optional: Obsidian vault, Zotero integration, OCR tools (tesseract, pytesseract, Pillow).
• Links:
  • Homepage: https://917dhj.github.io/DeepPaperNote/
  • Releases: https://github.com/917Dhj/DeepPaperNote/releases/tag/v2.0.0

Highlighted Details

• Evidence-First Workflow: Gathers comprehensive evidence from PDFs, metadata, and Zotero before note generation.
• Paper-Type-Aware Processing: Adapts strategies based on paper type (method, benchmark, survey).
• Native Knowledge-Base Output: Generates structured notes for Obsidian with YAML frontmatter and local image directories.
• Image-First Figure Handling: Integrates usable figures directly and provides placeholders for problematic ones.
• Zotero Integration: Supports leveraging local Zotero libraries for faster paper identification.
• OCR Fallback: Includes OCR for scanned PDFs when direct extraction is insufficient.

Maintenance & Community

Actively maintained with a clear release history (latest v2.0.0). Presented as a skill for agent platforms; specific community channels are not detailed.

Licensing & Compatibility

Licensed under the MIT License, permitting broad use, including commercial purposes. Compatible with multiple AI agent frameworks.

Limitations & Caveats

Notes are primarily generated in Chinese; English support is pending. Requires compatible agent environments. OCR is a fallback mechanism for specific PDF types. Optimal integration with Obsidian/Zotero requires user configuration.