Pi Life Sciences Literature

pi-literature is a Pi package mirroring the focused Life Sciences Literature Codex plugin. It provides three on-demand skills:

• biorxiv-skill — compact bioRxiv and medRxiv metadata and publication linkage
• ncbi-entrez-skill — high-level, batched PubMed literature searches with abstracts, plus low-level E-utilities access
• ncbi-pmc-skill — PMC Article Dataset Open Access, license, retraction, and file metadata

The package uses Node's built-in fetch (Node 18.17 or newer), so installation does not require Python or third-party runtime packages. Each skill reads one JSON object from stdin and writes one JSON object to stdout. Raw upstream bytes are saved unchanged when save_raw is requested.

Install

From this checkout:

pi install .

From npm after publishing:

pi install npm:pi-literature

Pi discovers the skills from the package's pi.skills manifest. They can also be invoked explicitly as /skill:biorxiv-skill, /skill:ncbi-entrez-skill, or /skill:ncbi-pmc-skill.

Development

npm test
npm run validate
npm run benchmark -- --cold 2 --warm 20

The benchmark makes live requests to the public bioRxiv, PubMed, and PMC endpoints. It is intentionally separate from npm test: cold samples spawn a fresh Node process, warm samples reuse one process, and p50/p95 are calculated only from successful samples. Use --skill biorxiv, --skill entrez, or --skill pmc to measure one source. For machine-readable output, add --json and use npm run --silent benchmark -- ... (or invoke the Node script directly).

The source-presentation contract keeps provenance evidence separate from checked-only connectivity/metadata responses, emits deterministic canonical links for supported DOI/PMID/PMCID identifiers, redacts private query values, and never rewrites requested raw HTTP bytes.

Security

The skills call public biomedical APIs and execute local helper scripts. Review the package before installing it, as Pi packages and skills run with the permissions of the Pi process.