citesleuth 0.1.1

Verify citations and spot hallucinations.

CiteSleuth

Verify citations, spot hallucinations.

CiteSleuth ingests BibTeX, DOCX, PDF, or plain text references and verifies them against public scholarly data providers. It produces machine-readable JSON and human-readable Markdown reports with suggested fixes.

Install

pip install citesleuth
# or from source
python -m pip install -e .

Usage

citesleuth verify refs.bib
citesleuth verify paper.pdf --format md
citesleuth verify paper.docx --out report.json
citesleuth verify refs.bib --write-bib refs.verified.bib --apply-fixes

Common flags:

• --offline: use cache only
• --no-cache: disable cache writes
• --max-rps: global request cap
• --provider-delay semanticscholar=1.1,dblp=2.0,openlibrary=0.5

Tests

python -m pip install -e .[test]
pytest
cargo test -p citesleuth-core

Rust + Node

Core matching logic lives in crates/citesleuth-core. Node bindings are in crates/citesleuth-napi.

cargo test -p citesleuth-core
cd crates/citesleuth-napi
npm install
npm run build

Quick Node usage:

const citesleuth = require('citesleuth')
const result = citesleuth.normalizeTitle('Attention Is All You Need')
console.log(result)

Notes

• Text-based PDFs only (no OCR in v1).
• Semantic Scholar uses an API key for best reliability. Set SEMANTIC_SCHOLAR_API_KEY or pass --api-key-semanticscholar.
• Open Library requests a descriptive User-Agent string; use --user-agent to override.