Models

The text API follows the OpenAI Chat Completions shape. Point an OpenAI-compatible client to the Omev base URL and keep the key on your server.

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://llmapi.omev.be/v1",
    api_key=os.environ["OMEV_API_KEY"],
)

response = client.chat.completions.create(
    model="omev-lite",
    messages=[{"role": "user", "content": "Summarise this page in five bullets: ..."}],
    max_tokens=500,
)

print(response.choices[0].message.content)

Public models

Model	Recommended use	Availability
omev-lite	High-volume, clearly specified text work such as extraction, data organisation, short rewriting and structured output	Standard text access
omev-pro	Complex instructions, planning, long-form drafting and work that needs deeper reasoning	Standard text access
omev-image-lite	One or more inline 1024x1024 images	Separate image agreement
omev-image-main	One inline 1K-class image, up to 1536x1024 or 1024x1536	Separate image agreement
omev-image-pro	One or more inline 1K-class images	Separate image agreement
omev-image-jobs	Asynchronous 1K image generation for automated pipelines	Separate image agreement

GET /v1/models returns the documented products enabled for the key. Do not build an integration around an unfamiliar or internal name.

Choosing between omev-lite and omev-pro

Start with omev-lite when the instruction and output are clear. Use omev-pro when the difficulty is reasoning: several dependent constraints, planning, long context or a response that must keep a complex argument consistent.

Moving to omev-pro does not replace a clear prompt. State the goal, required facts, audience, language, format and constraints in the request.

Text endpoint

• POST /v1/chat/completions
• streaming with "stream": true
• structured JSON with response_format
• standard system, user and assistant messages

Validate JSON against your own schema before storing or publishing it. When you set max_tokens explicitly, it caps the output tokens billed to your account. Omev's internal generation floors and hidden reasoning budgets remain unchanged, but they cannot raise billed output above max_tokens. Also set a client timeout that matches your use case.

Image endpoints

Image instructions are available to everyone, but image access is enabled only under a separate agreement with support@omev.be. See Image generation for the supported request shapes.

Stable public surface

Build against the public omev-* name, documented route and documented fields. The underlying technical model, provider, internal attempts and Omev's own cost are not part of the client response and may change without requiring an integration change.