AI Endpoints - API Catalog (EN)

Voir en Markdown

Découvrez comment utiliser l'API Catalog pour parcourir le catalogue OVHcloud AI Endpoints et consulter les métadonnées ainsi que les fonctionnalités de chaque modèle

Info

AI Endpoints is covered by the OVHcloud AI Endpoints Conditions and the OVHcloud Public Cloud Special Conditions.

Introduction

https://oai.endpoints.kepler.ai.cloud.ovh.net/v1 is the base endpoint used to actually call AI Endpoints models (see the AI Endpoints - Getting Started and AI Endpoints - Responses API guides). This article does not cover that endpoint; instead, it covers the separate Catalog API, a discovery service for browsing available models and their metadata.

The catalog URL https://catalog.endpoints.ai.ovh.net/rest/v1/models_v2 is the source for accessing information and metadata about AI Endpoints models. This API allows you to list models, supported features, pricing, and metadata specific to each model.

Warning

The catalog API is a discovery endpoint, not a text-generation endpoint. It returns metadata for models and capabilities rather than generated content.

This API may also change. Please note that a breaking change may be possible.

Objective

This documentation shows how to query the Catalog API, read model metadata, and filter the response client-side, with examples in Python, JavaScript, and cURL.

Requirements

The examples provided during this guide can be used with one of the following environments:

Python

JavaScript

cURL

A Python environment with the requests library.

pip install requests

Authentication

Unlike the v1/responses and v1/chat/completions routes, the catalog endpoint is public and does not require an API key. You can query it directly, without an Authorization header.

If you also plan to call the models it lists, follow the AI Endpoints - Getting Started guide to obtain an API key for those calls.

Listing all models

A single GET returns the whole catalog as a JSON array. The endpoint takes no query parameter.

Python

JavaScript

cURL

import requests

url = "https://catalog.endpoints.ai.ovh.net/rest/v1/models_v2"

response = requests.get(url)
response.raise_for_status()

catalog = response.json()
print(catalog)

import fetch from "node-fetch";

const url = "https://catalog.endpoints.ai.ovh.net/rest/v1/models_v2";

const response = await fetch(url);

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

Model metadata

Each entry in the array exposes the following fields:

id, name, description
available: whether the model can currently be called. Some entries are listed with available: false, so check this field before routing traffic to a model
category: the model family, one of Large Language Models (LLM), Visual LLM, Reasoning LLM, Code LLM, Embeddings, Image Generation, LLM Guard, Speech To Text, or Text To Speech
tags: short labels such as Reasoning, Multimodal, Code Assistant, or Moderation
metadata.model_specs.capabilities: input_modality and output_modality, arrays such as ["text", "image"], plus the reasoning, function_calling, and streaming booleans
metadata.publishing_information: publisher, series, licence_name, model_publication_date
metadata.usage_information.pricing: price and unit for prompt and completion usage
metadata.usage_information.rate_limit: request-rate limits, split into rate_limit_api (with an API key), rate_limit_api_anonymous (without one), and rate_limit_playground

Use the capability fields to pick a model before calling the v1/responses route: an input_modality containing image for vision use cases, function_calling for tool calling, and reasoning for reasoning modes.

Filtering the catalog

Every call returns the full catalog, so filtering, searching, and sorting happen on the returned array. The examples below keep only vision-capable models, then look up a single entry by id.

Python

JavaScript

cURL

import requests

url = "https://catalog.endpoints.ai.ovh.net/rest/v1/models_v2"

response = requests.get(url)
response.raise_for_status()
catalog = response.json()

vision_models = [
    model for model in catalog
    if "image" in model["metadata"]["model_specs"]["capabilities"]["input_modality"]
]
print(vision_models)

model = next((m for m in catalog if m["id"] == "gpt-oss-20b"), None)
print(model)

import fetch from "node-fetch";

const url = "https://catalog.endpoints.ai.ovh.net/rest/v1/models_v2";

const response = await fetch(url);
const catalog = await response.json();

const visionModels = catalog.filter((model) =>
  model.metadata.model_specs.capabilities.input_modality.includes("image")
);
console.log(visionModels);

const model = catalog.find((m) => m.id === "gpt-oss-20b");
console.log(model);

curl -s "https://catalog.endpoints.ai.ovh.net/rest/v1/models_v2" | \
  jq '[.[] | select(.metadata.model_specs.capabilities.input_modality | index("image"))]'

curl -s "https://catalog.endpoints.ai.ovh.net/rest/v1/models_v2" | \
  jq '.[] | select(.id == "gpt-oss-20b")'

Conclusion

Use the Catalog API to discover the available models and check their capabilities, features, pricing, and rate limits before calling the runtime API.

Go further

Use Catalog API is required to Deploy Langfuse on OVHcloud MKS for LLM observability and AI cost tracking. That article pulls the metadata.usage_information.pricing block from the catalog to register each model's price with Langfuse, which then computes the cost of your AI Endpoints calls automatically.

If you need training or technical assistance to implement our solutions, contact your sales representative or click on this link to get a quote and ask our Professional Services experts for a custom analysis of your project.

Feedback

Please send us your questions, feedback, and suggestions to improve the service:

On the OVHcloud Discord server.

Cette page vous a-t-elle aidé ?