REST API — VeldrixAI Docs

Base URL & auth

Base URL: https://api.veldrixai.ca. Authenticate with a bearer token:

HTTP

Authorization: Bearer vx-live-...
Content-Type: application/json
x-veldrix-sla-tier: STANDARD     # optional: REALTIME | STANDARD | BACKGROUND

POST /trust/evaluate

Field	Type	Required	Description
`prompt`	string	Yes	The input/prompt that produced the response.
`response`	string	Yes	The model output to evaluate.
`model`	string	No	Model identifier, for reporting (e.g. `gpt-4o`).
`context`	string	No	Ground-truth/source for hallucination checks.
`metadata`	object	No	Arbitrary tags stored with the audit record.

cURL

curl https://api.veldrixai.ca/trust/evaluate \
  -H "Authorization: Bearer vx-live-..." \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Summarize the patient record.",
    "response": "John Smith, DOB 1985-03-12, has hypertension.",
    "model": "gpt-4o"
  }'

Response shape

JSON

{
  "success": true,
  "data": {
    "request_id": "req_a1b2c3d4",
    "final_score": { "value": 34.0, "confidence": 0.92, "risk_level": "high_risk" },
    "pillar_results": {
      "compliance": { "score": { "value": 12.0 }, "flags": ["pii_detected"] },
      "safety":     { "score": { "value": 95.0 }, "flags": [] }
    },
    "execution_time_ms": 312
  }
}

REST scores are 0–100 (final_score.value); the Python SDK normalizes to 0–1 (overall). risk_level corresponds to the verdict band.

Audit & reports endpoints

Endpoint	Purpose
`GET /api/audit-trails`	List/filter audit records.
`GET /api/audit-trails/{request_id}/detail`	Full record for one evaluation.
`POST /api/audit-trails/{request_id}/intelligence`	Generate the AI Risk Thesis.
`POST /api/reports/generate-pdf`	Produce a PDF compliance report.

Errors & rate limits

Status	Meaning
`401`	Missing/invalid API key.
`422`	Malformed request body.
`429`	Quota exceeded or rate limited — back off and retry.
`5xx`	Transient server error — retry with exponential backoff.

Was this page helpful?