API Documentation (v2.1)

1. Architectural Overview

Nom AI exposes its core inference engine via a highly available API gateway routed through Amazon API Gateway and balanced by AWS Network Load Balancers. The API is designed around RESTful principles and strictly consumes/produces application/json payloads.

2. Authentication (Zero-Trust mTLS)

All requests require Bearer token authentication. Tokens must be provisioned via the Nom AI Dashboard. For enterprise VPC peered connections, mutual TLS (mTLS) is enforced via AWS Certificate Manager (ACM).

Authorization: Bearer nom_live_XXXXXXXXXXXXXXXXXXXX X-Nom-Organization-Id: org_XXXXXXXXXXXXX

3. Extraction Endpoint (POST /v2/extract)

Synchronous document extraction for payloads under 10MB. For larger batches (up to 500MB), use the asynchronous /v2/jobs endpoint backed by Amazon SQS.

POST https://api.me-south-1.nom-ai.com/v2/extract{"document_url": "s3://bucket/document.pdf", "language_mode": "ar-SA", "confidence_threshold": 0.98}

4. Rate Limits & Quotas

Standard enterprise limits are enforced via AWS WAF. Burst capacity allows up to 10,000 requests per second (RPS) per isolated tenant VPC. 429 Too Many Requests will be returned with a Retry-After header if limits are exceeded.