SDKs & tools

The Syntic API speaks the widely used chat completions format over plain HTTPS. That means you can call it from almost any language today, either with a client library you may already use or with a few lines of HTTP code.

Choose a route

What every route needs

Whichever you pick, three settings do the work:

Base URL   https://app.syntic.ai/v1
API key    sk_live_...           (Console → Settings → API Keys)
Model      Amara S1.2            (or another name from the models list)

Keep the key in an environment variable. The examples in these docs read it from SYNTIC_API_KEY.

Official Syntic SDKs

Syntic does not publish its own client libraries yet. The OpenAI-compatible route is the supported way to call the API from code until it does, and the Direct HTTP page shows complete, dependency-light examples.

Official libraries are being built. Their pages describe the planned design and are marked as previews:

Behaviour to plan for

These hold no matter which client you use:

• Streaming is on by default. Set stream: false when you want one JSON object.
• Errors are flat. The body is {"error": "<code>", ...} with a string code, which some libraries do not unpack. See Error codes.
• Sampling settings are ignored. temperature, top_p, n and the penalties have no effect.
• One endpoint. Only chat completions exists. Calls a library makes to other endpoints (model lists, embeddings, files) do not work.
• Unknown model names fall back. Today a misspelt model name is answered by the default model, not rejected. Not live yet — on develop: it is rejected with 404 model_not_found.

OpenAPI description

A machine-readable OpenAPI 3.0 file is served at https://app.syntic.ai/api/openapi. See OpenAPI description for how to use it and where it differs from the live API.