API Documentation

Quick start

Three steps and you're sending requests.

1. 1

   Grab an API key

   Visit the homepage — a key is automatically generated for your IP. Keys are bound to that IP.

2. 2

   Point your OpenAI SDK at us

   Override base_url and you're done — everything else is identical to OpenAI.

3. 3

   Pick any model

   From gpt-5 to veo-3. See the full catalog.

# Returns OpenAI-shaped { "data": [{ "url": "https://…/video.mp4" }] }
curl "https://your-apiarium.example.com/v1/videos/generations" \
  -H "Authorization: Bearer api-xxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "veo-3",
    "prompt": "a red apple rotating on white, studio lighting",
    "duration_seconds": "8",
    "resolution": "1080p",
    "size": "16:9"
  }'

Authentication

Every request to /v1/* needs a Bearer token.

Authorization: Bearer api-xxxxxxxxxxxxxxxx

• 🔒 Each key is bound to one IP address. Requests from a different network are rejected with HTTP 401.
• ⏱️ Default rate limit: 4 RPM per key. The master key bypasses limits.
• 🪪 Lose your key? Visit the homepage again — your IP will get the same key back.

Base URL

All endpoints below are relative to:

↑ this page auto-detects the running host

Endpoints

Models catalog

Live snapshot from GET /v1/models. Click any tile to copy the model id.

Chat completions

Drop-in replacement for POST https://api.openai.com/v1/chat/completions. Streaming via "stream": true works out of the box.

{
  "model": "gpt-5",
  "messages": [
    {"role": "user", "content": "Hello"}
  ],
  "stream": true,
  "temperature": 0.7
}

{
  "id": "chatcmpl-…",
  "object": "chat.completion",
  "model": "gpt-5",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "Hi! How can I help?"
    },
    "finish_reason": "stop"
  }]
}

Veo 3 — video

Dedicated endpoint: POST /v1/videos/generations. Models: veo-3, veo-3-fast, veo-3.1-fast. Returns an OpenAI-style payload: {"data": [{"url": "https://…/video.mp4"}]}.

You can also send model: "veo-3" to /v1/chat/completions — the assistant message will then contain the bare MP4 URL — but the dedicated endpoint is more ergonomic.

Supported parameters

{
  "model": "veo-3",
  "prompt": "a red apple rotating on white, studio lighting",
  "duration_seconds": "8",
  "resolution": "1080p",
  "size": "16:9"
}

{
  "created": 1779570172,
  "data": [{
    "url": "https://…/video.mp4",
    "revised_prompt": "a red apple rotating on white…"
  }]
}

Gemini 3 — image

Dedicated endpoint: POST /v1/images/generations. Models: gemini-3-image, gemini-3-flash-image, gemini-3.1-flash-image. Append -pro to any of them for the higher-quality variant (~2 min).

Supported parameters

{
  "model": "gemini-3-image",
  "prompt": "a glass of orange juice on a sunny kitchen counter",
  "size": "1:1"
}

{
  "created": 1779570172,
  "data": [{
    "url": "https://…/image.png",
    "revised_prompt": "a glass of orange juice…"
  }]
}

Gemini 3 — text (RAG)

Models: gemini-3, gemini-3-flash, gemini-3-pro.

These route to Gemini Hub's RAG-grounded /query endpoint. For free-form chat without a knowledge base, prefer gemini-2.5-pro, gemini-3.1-pro or one of the GPT/Claude aliases.

gpt-image-1 / gpt-image-2

Same endpoint as Gemini 3 image: POST /v1/images/generations. Just swap the model id.

{
  "model": "gpt-image-1",
  "prompt": "a watercolor of Istanbul rooftops",
  "n": 1,
  "size": "1024x1024"
}

Errors & limits

SDK examples

from openai import OpenAI

client = OpenAI(
    base_url="https://your-apiarium.example.com/v1",
    api_key="api-xxxxxxxxxxxxxxxx",
)

# 1) chat
chat = client.chat.completions.create(
    model="gpt-5",
    messages=[{"role": "user", "content": "hi"}],
)
print(chat.choices[0].message.content)

# 2) image (Gemini 3 or gpt-image)
img = client.images.generate(
    model="gemini-3-image",
    prompt="a glass of orange juice on a sunny kitchen counter",
    size="1:1",
)
print(img.data[0].url)

# 3) video (Veo 3) — /v1/videos/generations
import httpx
vid = httpx.post(
    "https://your-apiarium.example.com/v1/videos/generations",
    headers={"Authorization": "Bearer api-xxxxxxxxxxxxxxxx"},
    json={"model": "veo-3",
          "prompt": "a red apple rotating on white",
          "duration_seconds": "8",
          "resolution": "1080p",
          "size": "16:9"},
    timeout=600,
).json()
print(vid["data"][0]["url"])

import OpenAI from "openai";

const BASE_URL = "https://your-apiarium.example.com";
const API_KEY  = "api-xxxxxxxxxxxxxxxx";

const client = new OpenAI({
  baseURL: `${BASE_URL}/v1`,
  apiKey:  API_KEY,
});

// 1) image — uses the OpenAI SDK's images.generate
const img = await client.images.generate({
  model: "gemini-3-image",
  prompt: "a glass of orange juice on a sunny kitchen counter",
  size: "1:1",
});
console.log(img.data[0].url);

// 2) video — bypass the SDK, hit /v1/videos/generations directly
const vid = await fetch(`${BASE_URL}/v1/videos/generations`, {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${API_KEY}`,
    "Content-Type":  "application/json",
  },
  body: JSON.stringify({
    model: "veo-3",
    prompt: "a serene mountain sunset timelapse",
    duration_seconds: "8",
    resolution: "1080p",
    size: "16:9",
  }),
}).then(r => r.json());

console.log(vid.data[0].url);