DEVELOPER PLATFORM / TECHNICAL MANUAL

Developer Documentation.

Integration contract guide, request structures, model parameters, bearer authentication, and structured response handling.

01. Overview

Vessel Vision is developing a consistent HTTP JSON API for specialist medical imaging and diagnostic models. The examples below describe the planned integration contract.

Preview Architecture: This documentation details planned production contracts. AI inference is not connected. To register interest in future access, join the Beta Program.

02. Model Selection

Specify capabilities using family and task parameters:

  • model: "prism" - Radiology (e.g. pediatric-wrist-fracture, chestvision-screen)
  • model: "lumen" - Optical (e.g. sightdx-retinal-grade)
  • model: "aperture" - Audio biosignals (e.g. tb-cough-screening)

03. Authentication

All API requests require a Bearer token passed in the Authorization header:

Authorization: Bearer vv_live_YOUR_API_KEY

API keys are provisioned in the Developer Console. Ensure secret keys are kept strictly server-side and never exposed in client-side bundles.

04. Structured Outputs

The Vessel Vision response contract separates diagnostic findings, confidence concepts, review necessity flags, and usage telemetry:

ILLUSTRATIVE JSON RESPONSE
JSON
{
  "id": "vv_inf_98240172",
  "model": "prism",
  "task": "pediatric-wrist-fracture",
  "status": "completed",
  "findings": [
    "Incomplete transverse fracture line noted along the distal radial metaphysis.",
    "No displaced epiphyseal plate involvement observed."
  ],
  "concepts": {
    "fracture_presence": 0.981,
    "complexity_simple": 0.948,
    "joint_involvement": "none",
    "anatomic_location": "distal_radius"
  },
  "review_required": true,
  "clinical_disclaimer": "Model outputs require qualified professional review.",
  "usage": {
    "unit": "image_study",
    "quantity": 1,
    "inference_credits": 10
  }
}

05. Usage & Metering

Workspace results and prepaid API credits are separate balances. AI inference is not connected and demonstrations consume no credits. Usage rates and metering will be confirmed before AI access is enabled.

06. Error Handling

The API uses conventional HTTP status codes:

  • 400 Bad Request - Unsupported image format or missing parameters
  • 401 Unauthorized - Invalid or expired API key
  • 422 Unprocessable - Poor image resolution or out-of-distribution input
  • 429 Rate Limit - Exceeded organization concurrent limit

Help shape the next stage of medical AI.

Join Beta Program