Introduction

What Karmat builds, which APIs are live today, and where to start.

Karmat Engineering builds AI products and the models behind them. Two of those products expose a public HTTP API, documented here in full. A third is still in development.

These docs also cover how we design and build AI solutions on Azure and AWS, the same process we run on client work.

What is live today

API	What it does	Status
Optical Diagnostics	Eye condition decision support from symptoms, fundus images, or both together.	Live
Seerah AI	Retrieval-grounded Islamic Q&A over Quran, Hadith, scholarly fatwa and Seerah history.	Live
Attender	Attendance verification for schools and teams.	In development

Optical Diagnostics APIText, image and multimodal inference, uploads, webhooks and usage.Seerah AI APIAsk a question, get an answer with its sources, and keep per-user threads.Access and API keysRequest access, receive a key, rotate and revoke it.AttenderWhat is being built, and what to expect when it ships.

How you call them

Both APIs are plain HTTPS and JSON. There is no SDK. You call them with whatever HTTP client your language already has. Authentication is a single header on every request:

bash

curl -X POST "https://api-seerahai.karmat-hq.com/api/v1/chat" \
  -H "X-API-Key: $SEERAH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "What does the Quran say about patience?"}'

Access is reviewed

Keys are not self-serve. You submit an access request describing your integration, and an administrator approves or rejects it. See Access and API keys for the full flow. It is the same on both APIs.

Interactive reference

Each API publishes a live OpenAPI document and a Swagger UI you can call directly from the browser once you have a key.

API	Swagger UI	OpenAPI
Optical Diagnostics	/api/docs	/api/openapi.json
Seerah AI	/docs	/openapi.json

Building AI on cloud

The second half of these docs is about method rather than reference: how to design an AI solution, and how to build it on the two platforms our clients actually run on.

Designing AI solutionsScope, model selection and evaluation before infrastructure.Azure AIAI Foundry, speech, vision, documents and knowledge mining.AWS AIBedrock, SageMaker, Textract, Transcribe and Kendra.

Conventions

Every request and response body is JSON, except file uploads, which are multipart/form-data.
Timestamps are ISO 8601 in UTC. Examples read the key from an environment variable, and never inline one in code you ship.
Both APIs return 401 for a bad key, 403 for insufficient role or scope, 422 for a validation error and 429 when rate limited.