Academic paper deep analysis pipeline. Given a paper URL or local PDF file (supports arXiv, OpenReview, PubMed, direct PDF links, and local files), orchestrate a full pipeline: (1) download PDF and fetch related articles, (2) write an accessible Chinese explainer article with generated diagrams, (3) produce a modern glassmorphism single-page HTML website. Use when the user asks to analyze, explain, or visualize an academic paper, or when they provide a paper URL/file and want a comprehensive breakdown. Trigger phrases: '解读论文', '分析这篇论文', 'analyze this paper', 'paper URL', 'paper file'.
End-to-end pipeline: paper URL → PDF download → article writing → HTML website.
User provides paper URL
↓
Step 0: Extract metadata (title, authors, abstract, arXiv ID)
Step 1: Create isolated workspace directory
Step 2: Download PDF + fetch related articles → references/resource-downloader.md
Step 3: Write explainer article + generate diagrams → references/content-writer.md
Step 4: Generate glassmorphism HTML website → references/web-designer.md
Step 5: Deliver results
Detect input type:
https://arxiv.org/abs/XXXX.XXXXX.pdf or containing /pdf/Extract metadata:
webfetch or WebReader toolGenerate standardized directory name:
For all papers:
{sanitized_title}
Base path: Papers are stored in ~/papers/ (user's home directory).
Full workspace path example:
~/papers/attention_is_all_you_need/
Title sanitization: lowercase → remove special chars (keep [a-z0-9_-]) → spaces to underscores.
Create directory structure:
mkdir -p "{workspace}/pdfs" "{workspace}/articles" "{workspace}/images"
Run scripts/download_paper.sh for any paper type:
# arXiv paper (backward compatible)
bash scripts/download_paper.sh "https://arxiv.org/abs/1706.03762" "{workspace}"
# Direct PDF URL
bash scripts/download_paper.sh "https://arxiv.org/pdf/1706.03762.pdf" "{workspace}"
# Local PDF file
bash scripts/download_paper.sh "/path/to/paper.pdf" "{workspace}"
The script automatically detects input type and downloads/copies the PDF to {workspace}/pdfs/.
Then fetch 2-3 related articles (blog posts, explanations) into {workspace}/articles/.
Read the full guide: references/content-writer.md
Key points:
{workspace}/article.md with 6 mandatory sectionsscripts/generate_diagrams_template.py as template{workspace}/images/Read the full guide: references/web-designer.md
Key points:
article.md → single self-contained index.htmlassets/html_template.html as skeletonimages/xxx.pngReport workspace path, HTML file path, and file statistics.
output/.images/xxx.png, never absolute paths.mkdir -p, curl, ls — never Windows commands.| Resource | Purpose |
|---|---|
scripts/download_paper.sh | Universal PDF downloader: supports arXiv URLs, direct PDF URLs, and local files |
scripts/generate_diagrams_template.py | Matplotlib diagram generation template |
references/content-writer.md | Article writing guide and structure |
references/resource-downloader.md | PDF + article download strategies |
references/web-designer.md | HTML generation guide and design specs |
assets/html_template.html | Glassmorphism HTML skeleton with particles, theme toggle, TOC |