Documentation

Platform Documentation

API reference, core concepts, and integration guidance for the VetIOS clinical intelligence platform.

Authentication

All API requests require a valid session token obtained via the VetIOS authentication surface. The platform uses Supabase-based JWT sessions. Include your token in the Authorization header:

Request Header
Authorization: Bearer <your-session-token>
Content-Type: application/json

Tokens are tenant-scoped. An operator token cannot access another tenant's data. Token expiry and rotation are managed automatically by the platform session layer.

API Reference

POST/api/inferenceCore

Submit a clinical case for AI-assisted differential diagnosis. Returns ranked hypotheses with confidence scores, CIRE signals, and a traceable inference event ID.

Request Body
{
  "model": { "name": "VetIOS Diagnostics", "version": "latest" },
  "input": {
    "input_signature": {
      "species": "canine",
      "breed": "mixed",
      "symptoms": ["vomiting", "lethargy"],
      "metadata": {
        "age_years": 3,
        "labs": { "wbc": 4.1, "pcv": 29 }
      }
    }
  }
}
Response
{
  "inference_event_id": "9f2c1b6a-...",
  "data": {
    "confidence_score": 0.82,
    "differentials": [
      { "label": "canine_parvovirus", "p": 0.82 },
      { "label": "hemorrhagic_gastroenteritis", "p": 0.41 }
    ]
  },
  "cire": { "phi_hat": 0.71, "cps": 0.12, "safety_state": "nominal" },
  "meta": { "tenant_id": "...", "request_id": "..." },
  "error": null
}
POST/api/outcomeLearning

Submit a confirmed outcome for a resolved case. Outcome events are used by the closed-loop learning layer to refine inference priors within your tenant.

Request Body
{
  "inference_event_id": "9f2c1b6a-...",
  "outcome": {
    "type": "confirmed_diagnosis",
    "payload": {
      "label": "canine_parvovirus",
      "confidence": 0.98
    },
    "timestamp": "2026-05-04T12:00:00.000Z"
  }
}
Response
{
  "outcome_event_id": "evt_2841...",
  "clinical_case_id": "case_4XK3...",
  "linked_inference_event_id": "9f2c1b6a-...",
  "request_id": "..."
}
POST/api/simulateSimulation

Run a synthetic simulation against a base case to pressure-test model behaviour before rollout. Returns a stability report and simulation trace.

Request Body
{
  "steps": 10,
  "mode": "adaptive",
  "base_case": {
    "species": "canine",
    "symptoms": ["vomiting", "lethargy"],
    "metadata": { "wbc": 4.1, "pcv": 29 }
  },
  "inference": {
    "model": "gpt-4o-mini",
    "model_version": "gpt-4o-mini"
  }
}
Response
{
  "simulation_event_id": "sim_901A...",
  "clinical_case_id": "...",
  "stability_report": {
    "passes": 9,
    "failures": 1,
    "mean_confidence": 0.79
  },
  "request_id": "..."
}

Core Concepts

Inference Engine

The core runtime that normalises clinical inputs, routes them through the model layer, and returns ranked differential hypotheses with confidence scores and CIRE signals.

input normalisationmodel routingranked output

CIRE Signals

Clinical Inference Reliability Estimation. Each inference response includes phi_hat (hypothesis confidence), cps (calibration pressure score), and safety_state — giving operators visibility into output reliability.

phi_hatcpssafety_state

Outcome Learning

Closed cases submitted via /api/outcome become supervision events. The learning layer uses confirmed diagnoses to refine inference priors within your tenant's isolated partition.

supervision eventsclosed-looptenant-isolated

Simulation Layer

Run synthetic case traffic against new model versions or policy changes before promoting them to live inference. Validates stability across edge cases and replayed historical patterns.

pre-rollout testingstability reportsynthetic traffic

Tenant Isolation

All clinical data, inference history, and outcome learning are partitioned per tenant. Data does not cross tenant boundaries. Operators within a tenant control access roles.

data isolationRBACmulti-tenant

Traceable Runtime

Every inference event produces a persistent inference_event_id. All requests, model calls, latency spans, and policy checks are logged to an immutable audit trail.

audit traillatency tracingevent IDs

Species Support

The inference engine accepts species-typed inputs. Correct species specification improves differential ranking accuracy as physiological priors are species-specific.

CanineFelineEquineBovineOvine / CaprinePorcineAvian (poultry)Exotic / Wildlife (beta)

Rate Limits

Standard
60 inference requests / min
Default for operator accounts
Partner
300 inference requests / min
Available under partner agreement
System / Admin
Configurable
Governed by tenant policy

Rate limit headers are returned on every response: X-RateLimit-Remaining and X-RateLimit-Reset.

Ready to integrate?

API access requires a VetIOS account. Request access or sign in to get your credentials.