Search the documentation

Pick a page and open it with Enter.

Quickstart

From an empty project to your first answer in a few minutes.

From an empty project to your first streamed answer — three steps, no setup beyond an API key.

Prerequisites

You need a running backend and, if it is reachable over the network, an API key. Locally the backend trusts localhost, so a bare request works with no key at all. Node 18+ or Python 3.10+ is enough to run the examples below.

  1. 1

    Create a key

    Create a new key in the dashboard under Settings → API keys.

  2. 2

    Set an environment variable

    Store the key as SMEEWARE_API_KEY, never directly in source.

  3. 3

    Send your first call

    Copy the example below and run it — the answer comes back as a stream.

First call

Send a list of messages and a model id. The answer streams back as Server-Sent Events — one JSON frame per line.

bash
curl https://api.smeeware.dev/v1/chat \
  -H "Authorization: Bearer $SMEEWARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [
      { "role": "user", "content": "Hello!" }
    ]
  }'

Next steps

From here, learn how to read the stream frame by frame, how to recover from errors and rate limits, and what each endpoint accepts. The two cards below are the usual next stops.