clo-author-rag

Research scaffold for empirical papers, combining clo-author multi-agent pipeline with a local RAG engine for PDF knowledge bases.

Built on top of clo-author (Hugo Sant'Anna) and Proyecto-RAG.

What It Does

• Multi-agent research pipeline — librarian, strategist, coder, writer, and critic pairs guide you from idea to submission
• Local RAG — drop reference PDFs into papers/, they get indexed automatically; Claude queries them instead of re-reading every session
• 10 slash commands — /discover, /strategize, /analyze, /write, /review, /revise, /talk, /submit, /tools, /checkpoint
• Simulated peer review — /review --peer [journal] runs a full desk review + two independent referees

Clone and Set Up

If you already have the repo cloned but the submodule is missing:

First Steps After Cloning

• **Fill in CLAUDE.md** — replace [BRACKETED PLACEHOLDERS] with your project details
• **Fill in .claude/references/domain-profile.md** — your journals, data sources, notation, and seminal references. Use /discover interview to populate it interactively.
• Add your PDFs to papers/ and run python watcher.py --once
• Open Claude Code and run:

RAG: What Gets Indexed

PDFs placed in papers/ are processed into rag-data/ (gitignored — local only). The index is rebuilt with python watcher.py --once on any new machine after cloning.