Every model, one key.

Every model, one key.

An OpenAI-compatible API for text, embeddings, images, and video. Point the client you already use at Bayle and keep going.

Get an API keyMake your first request

cURL

curl https://api.bayle.ai/chat/completions \
  -H "Authorization: Bearer $BAYLE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MODEL_ID",
    "messages": [{"role": "user", "content": "In one sentence, what is an embedding?"}]
  }'

Python

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["BAYLE_API_KEY"],
    base_url="https://api.bayle.ai",
)

completion = client.chat.completions.create(
    model="MODEL_ID",
    messages=[{"role": "user", "content": "In one sentence, what is an embedding?"}],
)
print(completion.choices[0].message.content)

TypeScript

import OpenAI from 'openai'

const client = new OpenAI({
  apiKey: process.env.BAYLE_API_KEY,
  baseURL: 'https://api.bayle.ai',
})

const completion = await client.chat.completions.create({
  model: 'MODEL_ID',
  messages: [{ role: 'user', content: 'In one sentence, what is an embedding?' }],
})
console.log(completion.choices[0].message.content)

Endpoints

API reference

POST/chat/completionsChat with streaming, function calling and visionPOST/embeddingsVectors for search and retrievalPOST/images/generationsOne image per requestPOST/video/queueQuote, queue, poll, then download the clipGET/modelsEvery model with its type and capabilities, no key needed

Where to go next

Bayle serves an HTTP API shaped like OpenAI’s, at https://api.bayle.ai. If you already have code talking to an OpenAI-compatible endpoint, changing the base URL and the API key is usually the whole integration.

Start calling it

Quickstart gets a first response back. Authentication covers the header and which endpoints need it.

Already using an OpenAI SDK?

OpenAI compatibility shows the base URL change per language, and the three places behaviour differs from OpenAI’s own API.

Generate something

Text and streaming, embeddings, images, and video.

Work out what went wrong

Errors lists every code the API returns, with the message and the OpenAI error type it maps to.

Browse every model

See what’s available, with capabilities and current status, at the Models page. The /models guide covers listing and filtering the same catalogue from the API itself.

Got a consent error?

If a video request came back with 409 and "code": "needs_consent", that model needs a consent attestation before it will generate media containing a face. Nothing was charged.

Resend the same request with a consents object added. The full walkthrough, including what to show the person whose likeness appears in the media, is in the video guide.