For developers

Build on the YG3 API.

Four models behind one OpenAI-compatible endpoint. Elysia for text reasoning. Taurus for vision. Vani for image generation. Merlin for the technical jobs. Point your existing OpenAI client at our base URL and start sending.

Request an API key Read the docs

The whole API in one call

$ curl https://elysia-api.ngrok.io/api/public/v1/chat/completions \
  -H "Authorization: Bearer $ELYSIA_API_KEY" \
  -d '{"model":"elysia","messages":[{"role":"user","content":"hi"}]}'

Quickstart

The fastest call.

Drop these into a notebook and run them. The OpenAI Python SDK works unchanged against the YG3 base URL.

Authenticate

bash

Set ELYSIA_API_KEY in your environment. Every request takes the standard Authorization header.

curl https://elysia-api.ngrok.io/api/public/v1/models \
  -H "Authorization: Bearer $ELYSIA_API_KEY"

Chat with Elysia

python

Same request body as the OpenAI Chat Completions API.

from openai import OpenAI

client = OpenAI(
    base_url="https://elysia-api.ngrok.io/api/public/v1",
    api_key=ELYSIA_API_KEY,
)

response = client.chat.completions.create(
    model="elysia",
    messages=[
        {"role": "user", "content": "Summarize this article in three lines."},
    ],
    temperature=0.7,
    max_tokens=1000,
)
print(response.choices[0].message.content)

Vision with Taurus

python

Multimodal content array. Image as base64 or URL, text alongside it.

# Taurus handles vision + text in the same call.
response = client.chat.completions.create(
    model="taurus",
    messages=[{
        "role": "user",
        "content": [
            {"type": "image_url", "image_url": {
                "url": f"data:image/png;base64,{img_b64}",
            }},
            {"type": "text", "text": "What is in this image?"},
        ],
    }],
)

Image gen with Vani

python

Returns base64-encoded PNG. Omit response_format for a URL.

# Vani returns a base64-encoded image.
response = client.images.generate(
    model="vani",
    prompt="A storefront on a quiet street at dawn",
    response_format="b64_json",
)
img_b64 = response.data[0].b64_json

Models