📚 renderscholar

renderscholar lets you search Google Scholar from the terminal, scrape real papers, and render them into a static HTML page.
No more hallucinated citations from LLMs — everything comes straight from Scholar.

The HTML includes two views:

• 👤 Human View — nicely formatted papers (title, authors, year, citations, link, snippet).
• 🤖 LLM View — plain-text block you can copy-paste into ChatGPT/Claude/etc.

Inspired by Andrej Karpathy's rendergit.

---

🔧 Installation

You'll need Python 3.10+. Then install directly from GitHub:

pip install git+https://github.com/peterdunson/renderscholar.git

First-time setup

Playwright requires one extra step (to install the headless Chromium browser it uses):

playwright install chromium

Optional: Semantic mode

If you want to enable the semantic ranking mode (using transformer embeddings for smarter similarity), install with extras:

pip install "git+https://github.com/peterdunson/renderscholar.git#egg=renderscholar[semantic]"

---

🚀 Usage

Search Google Scholar from the terminal:

renderscholar "Bayesian nonparametric survival analysis"

Options:

• --pool-size N → number of raw results to scrape (default: 100)
• --filter-top-k N → number of top papers to keep after ranking (default: 20)
• --mode balanced|recent|famous|influential|hot|semantic → ranking style (default: balanced)
• --no-open → don't auto-open in browser
• -o out.html → write to a specific file instead of a temp file

📊 Modes

Each mode changes how papers are ranked (weights on query similarity, citation count, and recency):

• balanced → even weighting (default)
• recent → favors newest work
• famous → favors most cited classics
• influential → balances impact + relevance
• hot → boosts both citations and recency (trending work)
• semantic → uses transformer embeddings for semantic similarity (requires optional install)
• single → scrapes only the top 10 results and returns the single most relevant paper (best for quick lookups)

Example:

renderscholar "Bayesian factor models in genomics" --pool-size 80 --filter-top-k 15 --mode hot

This will:

1. Open Scholar in a Chromium browser (so you can solve captcha if needed).
2. Scrape ~80 results.
3. Rank and filter to the top 15 (favoring "hot" papers).
4. Render an HTML file and open it in your browser.

---

📂 Output

• 👤 Human View
  Nicely formatted papers with titles, authors, years, citations, snippets, and links.

• 🤖 LLM View
  Copiable plain text, ready for ChatGPT/Claude/etc. — no risk of hallucinated papers.

---

📝 Notes

• If you see a captcha, just solve it in the opened browser — scraping will resume automatically.
• The LLM view is plain Unicode text (titles, authors, citations, links, snippets).
• Designed for research workflows: quickly scan papers, then copy results into your favorite AI assistant.

---

📄 License

MIT © 2025 Peter Dunson