A system that turns research PDFs into a Wikipedia-style static website.
Wiki-Gen processes research papers from the ingest/ directory, extracts metadata and figures, generates a Wikipedia-style summary (using a language model), and builds a static HTML site.
This repository is configured to be friendly with Claude, Gemini, and Qwen.
Custom skills are located in the skills/ directory and are symlinked to:
.claude/skills/.gemini/skills/.qwen/skills/
These skills provide the model with domain-specific knowledge and instructions for:
- ingest: Process one or all PDFs in
ingest/. - build: Regenerate the static site from the database.
- reingest: Reprocess a paper from scratch.
- fix-title: Correct a paper's title in the database.
- wiki-gen: General help and information about the system.
Slash commands are available in both Gemini and Qwen:
/ingest [filename]/build/reingest <filename or paper_id>/fix-title <paper_id> <new title>/wiki-gen
- Ingest: Place PDFs in the
ingest/directory. - Process: Run
/ingest(or the equivalent skill) to process the papers. - View: Open
wiki/site/index.htmlto view the generated site.
ingest/: Input PDFs.processed/: Successfully processed PDFs.scripts/: Python scripts for processing and site generation.skills/: Cross-tool skill definitions.wiki/data/: SQLite database.wiki/site/: Generated static website.