Skip to main content

Documents

The Documents module manages clinical documents through upload, indexing, and search.

Capabilities​

  • Upload PDFs, images, and supported office documents.
  • Extract text and metadata through the ingestion pipeline.
  • Search with hybrid BM25 plus vector retrieval.
  • Associate documents with patients, groups, and tenant scope.

RAG Integration​

Indexed documents can contribute to the local corpus used by the Clinical Research Agent. External sources such as PubMed remain separate and are governed by retrieval policy.

The current ingestion path uses PaddleOCR 3.x for scanned PDFs and images, pdfplumber for structured extraction from born-digital PDFs, and pypdf as a plain-text fallback. Availability depends on environment configuration.

Privacy​

Documents are tenant-scoped. PHI scanning, retention, and log coverage must be confirmed for the enabled ingestion path.