ARI Models Gateway

The public client endpoint is https://models.ari-motors.com. The gateway forwards requests to a private Ollama instance on the Mac Mini through Tailscale. Clients never talk to Ollama directly.

Base URL

Authentication

Public API calls require a Bearer token in the Authorization header.

Authorization: Bearer <prod-token>

The admin-only endpoint /installed-models requires the admin token.

Authorization: Bearer <admin-token>

The browser playground does not expose tokens. It uses Auth0 session login on the server and internal proxy routes under /playground-api/*.

Endpoints

What the API can return

Request formats

{
  "model": "qwen3:4b",
  "messages": [
    { "role": "user", "content": "Write one short product summary." }
  ],
  "stream": false,
  "keep_alive": "5m",
  "options": {},
  "format": "json"
}

{
  "model": "qwen2.5vl:3b",
  "prompt": "Describe the image in English.",
  "images": ["<base64-image>"],
  "stream": false,
  "keep_alive": 0,
  "options": {}
}

Vision images are always preprocessed on the gateway: max long side 768 px, converted to JPEG, quality 82 by default.

JavaScript examples

1. Check health

const response = await fetch('https://models.ari-motors.com/health', {
  headers: {
    Authorization: 'Bearer ' + PROD_TOKEN
  }
});

const data = await response.json();
console.log(data);

2. Load public models

const response = await fetch('https://models.ari-motors.com/models', {
  headers: {
    Authorization: 'Bearer ' + PROD_TOKEN
  }
});

const data = await response.json();
console.log(data.models);

3. Send a chat request

const response = await fetch('https://models.ari-motors.com/chat', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    Authorization: 'Bearer ' + PROD_TOKEN
  },
  body: JSON.stringify({
    model: 'qwen3:4b',
    messages: [
      { role: 'user', content: 'Write a short English greeting.' }
    ],
    stream: false,
    keep_alive: '5m'
  })
});

const data = await response.json();
console.log(data.message?.content);
console.log(data.message?.thinking);

4. Consume streaming chat output

const response = await fetch('https://models.ari-motors.com/chat', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    Authorization: 'Bearer ' + PROD_TOKEN
  },
  body: JSON.stringify({
    model: 'qwen3:4b',
    messages: [
      { role: 'user', content: 'Explain HTTP streaming in one paragraph.' }
    ],
    stream: true
  })
});

const reader = response.body.getReader();
const decoder = new TextDecoder();
let buffer = '';

while (true) {
  const { value, done } = await reader.read();
  if (done) break;

  buffer += decoder.decode(value, { stream: true });
  const lines = buffer.split('\n');
  buffer = lines.pop() || '';

  for (const line of lines) {
    if (!line.trim()) continue;
    const chunk = JSON.parse(line);

    if (chunk.message?.thinking) {
      console.log('thinking:', chunk.message.thinking);
    }

    if (chunk.message?.content) {
      console.log('content:', chunk.message.content);
    }

    if (chunk.done) {
      console.log('finished:', chunk.done_reason);
    }
  }
}

5. Send a vision request from JavaScript

async function fileToBase64(file) {
  return new Promise((resolve, reject) => {
    const reader = new FileReader();
    reader.onload = () => {
      const result = String(reader.result || '');
      resolve(result.includes(',') ? result.split(',')[1] : result);
    };
    reader.onerror = reject;
    reader.readAsDataURL(file);
  });
}

const imageBase64 = await fileToBase64(fileInput.files[0]);

const response = await fetch('https://models.ari-motors.com/vision', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    Authorization: 'Bearer ' + PROD_TOKEN
  },
  body: JSON.stringify({
    model: 'qwen2.5vl:3b',
    prompt: 'Describe everything visible in this image.',
    images: [imageBase64],
    stream: false
  })
});

const data = await response.json();
console.log(data.response);

Response fields

{
  "status": "ok",
  "gateway": "ok",
  "macMiniReachable": true,
  "ollamaResponding": true,
  "responseTimeMs": 217
}

{
  "models": [
    {
      "name": "qwen3:4b",
      "model": "qwen3:4b",
      "type": "chat",
      "family": "qwen3",
      "size": 1234567890
    }
  ]
}

{
  "models": [
    {
      "name": "qwen3:4b",
      "model": "qwen3:4b",
      "size": 1234567890,
      "modified_at": "2026-04-16T10:00:00Z",
      "details": {
        "family": "qwen3",
        "parameter_size": "4B",
        "quantization_level": "Q4_K_M"
      }
    }
  ]
}

{
  "model": "qwen3:4b",
  "created_at": "2026-04-16T10:00:00Z",
  "message": {
    "role": "assistant",
    "content": "Hello.",
    "thinking": "Optional reasoning text"
  },
  "done": true,
  "done_reason": "stop"
}

{
  "model": "qwen2.5vl:3b",
  "created_at": "2026-04-16T10:00:00Z",
  "response": "Image description",
  "done": true,
  "done_reason": "stop"
}

Errors and limits