Skip to main content
search queries the clinical knowledge base and returns structured document units — never a free-form blob. It is the retrieval foundation the other capabilities build on. Endpoint: POST /v1/search · MCP tool: search

​
When to use it

  • You want the raw sources for a query, to display or cite yourself.
  • You are building your own synthesis and just need retrieval.
For a written answer instead of sources, use answer.

​
Request

FieldTypeNotes
querystringFree-text clinical query (French).
source_idsint[]Optional. Restrict to specific sources.
limitintMax document units (default 10).
max_chunks_per_unitintBest chunks kept per unit (default 3).

​
Response

Each unit is one source document with its best-matching chunks: 
{
  "query": "antibioprophylaxie chirurgie colorectale",
  "units": [
    {
      "document_id": 14853,
      "document_title": "Antibioprophylaxie en chirurgie",
      "reco_id": 900,
      "reco_title": "Chirurgie colorectale",
      "source_id": 3,
      "source_name": "SFAR",
      "year": "2023",
      "deep_link": "https://.../14853.pdf",
      "score": 0.91,
      "chunks": [
        { "chunk_index": 0, "content": "Céfazoline en dose unique...", "score": 0.91 }
      ]
    }
  ],
  "retrieved_pool": 100
}
limit caps the number of document units, not chunks. retrieved_pool is the diagnostic count of raw chunks swept before grouping — it is not bounded by limit.
Run it live in the API Reference tab.