Skip to main content
analyse applies a free-text clinical rule (an “équation type”) to one or more patient health records, decides which health records match, and attaches a knowledge-base-grounded recommendation with citations to each match. Endpoint: POST /v1/analyse · MCP tool: analyse It is synchronous (non-streamed) and powers punctual or scheduled reviews — a scheduler simply calls the endpoint on a batch.

​
How it works

rule ──(compile)──▶ criteria
                     │
   per health record ◀─────┘
     ├─ deterministic eval  (medication present, age, lab staleness…)
     └─ LLM judgment        (therapeutic class, clinical interpretation)
            │
   match? ──▶ recommendation via the knowledge base (cited)
The evaluation is hybrid: criteria that map cleanly to a health record field are checked deterministically; the rest are judged by the model. Health records are passed inline — Locus stores nothing.

​
Request

{
  "rule": "patient sous anticoagulant oral direct ET DFG non contrôlé depuis plus de 6 mois",
  "records": [
    {
      "id": "patient-4815",
      "age": 78,
      "medications": ["apixaban 5mg x2/j"],
      "recent_labs": [{ "name": "DFG", "value": "48 ml/min", "measured_at": "2024-11-30" }]
    }
  ]
}

​
Response

{
  "rule": "...",
  "compiled_criteria": [
    { "id": "c1", "description": "sous AOD", "kind": "judgment" },
    { "id": "c2", "description": "DFG ancien > 6 mois", "kind": "structured",
      "field": "lab", "op": "stale", "value": "DFG", "max_age_days": 180 }
  ],
  "findings": [
    {
      "record_id": "patient-4815",
      "matched": true,
      "criteria": [
        { "criterion_id": "c1", "met": true, "basis": "judgment", "rationale": "apixaban est un AOD" },
        { "criterion_id": "c2", "met": true, "basis": "deterministic", "rationale": "lab DFG daté de 208 j > 180 j" }
      ],
      "rationale": "...",
      "recommendation": "Réévaluer la fonction rénale et adapter la posologie...",
      "sources": [{ "reco_id": 412, "reco_title": "...", "source_name": "HAS", "deep_link": "..." }]
    }
  ],
  "metadata": { "records_scanned": 1, "matched_count": 1, "processing_ms": 1840 }
}
Latency scales with the number of health records (one judgment call and one recommendation per match, run concurrently). Requests are bounded per call; for large cohorts, batch across requests.
Run it live in the API Reference tab.