Offline Research Paper Agent

The need

Research papers, reference records and evidence excerpts become difficult to search and reconcile across drafts.

Why it was developed

Keep a local indexed paper library, preserved BibTeX records, grounded drafting and phase-aware systematic-review audit trails.

How to use it

1. Create and activate a Python 3.10+ virtual environment. Run python -m pip install .
2. Run research-agent --home ./library init, then research-agent --home ./library add ./sources.
3. Run research-agent --home ./library search "your research question". Import references using bib-import and link them to paper IDs.
4. Install Ollama and an appropriate local model for ask/draft commands. Verify generated claims and citation labels against the source documents. ORIGINAL-README.md contains the full command workflow.

How it can save time or money

Indexes papers once, reuses extracted passages and reference records, and detects duplicates instead of repeatedly searching PDFs and reconciling bibliography files by hand.

These are potential benefits, not measured savings. Include setup, maintenance, review and local operating costs when comparing your own workflow.

Current scope & limits

Search/library work needs no model; generative drafting needs a configured local model. Database subscription access, lawful source acquisition and scientific verification are outside the tool. Personal paper libraries are excluded.