API reference

The AIx HTTP API is OpenAI-compatible. This section is the precise reference; for task-oriented walkthroughs see the Guides.

Base URL

https://api.aix.theaimart.co/v1

Authentication

Send your key as a bearer token (SDKs do this automatically). X-API-Key is also accepted for direct HTTP integrations.

Authorization: Bearer tam_your_key_here

See Authentication.

Conventions

All request/response bodies are JSON (except audio/speech which returns raw audio bytes, and audio/transcriptions which takes multipart form-data).
Timestamps are Unix seconds. Money is USD. 1 credit = $1.
Unknown JSON fields are ignored. Use only documented fields when you need portable, forward-compatible behavior.
Model execution details stay behind AIx; application code receives the documented response shape.

Endpoints

Method	Path	Purpose
POST	`/v1/chat/completions`	Chat (text, tools, vision, streaming)
POST	`/v1/messages`	Messages (Anthropic-compatible clients)
POST	`/v1/embeddings`	Embeddings
POST	`/v1/images/generations`	Images
POST	`/v1/audio/speech`	Text-to-speech (binary)
POST	`/v1/audio/transcriptions`	Transcription (multipart)
GET	`/v1/models`	Models

Versioning

The surface is versioned by the /v1 path prefix. Changes are additive — new optional fields and models may appear. Clients should ignore unknown response fields and fetch model ids from GET /v1/models instead of hardcoding availability.

OpenAPI

A machine-readable spec is available for client codegen and Postman/Insomnia import:

Download openapi.json →

Errors & limits

Errors use the OpenAI envelope — see Errors.
Rate limits, spend caps, TPM and scopes — see Limits, caps & scopes.

Chat →