BioAgents Architecture

BioAgents is the AI layer of the platform. You hand it a research paper; it hands back structured metadata, recognized entities, and RDF triples ready for the knowledge graph. The point is not to summarize papers - it's to make a corpus of them queryable as data.

The framework underneath

BioAgents runs on the Eliza OS agent framework, which supplies the pieces you'd otherwise rebuild for every agent system: a runtime, a plugin mechanism for specialized capabilities, persistent memory across runs, and task workers so that a forty-page PDF doesn't block an HTTP connection while it's processed.

From PDF to knowledge graph

A paper moves through three stages, and each one narrows the representation.

Document processing comes first. GROBID pulls text and structure out of the PDF - which sounds mundane until you've tried it; scientific PDFs are hostile territory, and getting a clean "this is the methods section, this is a citation" segmentation is half the battle. The output is normalized, section-labelled text.

Knowledge extraction runs over that text: entity recognition for genes, proteins, diseases and compounds; relationship detection between them; claim extraction with its supporting evidence; and the bibliographic metadata - authors, citations, publication details.

Knowledge representation is the final narrowing. Extracted facts become RDF triples and JSON-LD documents, then get linked into the existing graph. From here on the paper isn't a document anymore - it's rows in a queryable structure.

How it connects to Bio-DID-Seq

API integration

DID-based metadata management

Data processing flow

API

Endpoints

Request/Response Examples

Process Paper Request

{
  "file_cid": "QmXg9Pp2ytZ14xgK35M6iTC2Vz6jR9zYgooNp2UHPTMnPN",
  "title": "CRISPR-Cas9 Gene Editing for Neurodegenerative Diseases",
  "authors": ["Jane Smith", "John Doe"],
  "doi": "10.1038/s41586-021-03819-2"
}

Process Paper Response

{
  "task_id": "b8e5c9a4-2c7a-4d64-b4b3-7c01e2f8b54e",
  "status": "processing"
}

Security and privacy

Nothing changes at the trust boundary just because AI is involved. BioAgents endpoints sit behind the same authentication and UCAN authorization as the rest of the API, data is encrypted in transit and at rest, and access is audit logged. A paper you haven't shared can't be processed by someone else's agent, the capability check happens before the pipeline ever sees the file.