Models now have categories

Models now have categories. Owners can assign up to three categories to a model, and anyone can filter models by category on the search page or through the API.

What changed

Finding a model used to mean searching by name or browsing the full list. Categories describe what a model does, so you can narrow the list to the models that fit your task. Building transcription? Filter by Speech to Text. Adding retrieval to a RAG pipeline? Filter by Text Embedding or Reranking. Generating images? Text to Image.

A model's categories show up on its model card, on its model detail page, and in search results. Model owners choose them from the dashboard, either when creating a model or by editing an existing one.

Using categories with the API

List the available categories with GET /model_categories, then pass their slugs as category_slugs to GET /models. A model matches if it has any of the slugs.

The official clients already support categories, so you don't need to build the requests yourself. This example returns models that do either speech to text or text to speech:

1import { WriftAI } from 'wriftai';
2
3const wriftai = new WriftAI();
4
5const models = await wriftai.models.list({
6  category_slugs: ['speech-to-text', 'text-to-speech'],
7});
8
9for (const model of models.items) {
10  console.log(model.name);
11}

Owners can set categories with category_slugs when creating or updating a model. See the API reference for details.