Skip to content

Code-first video & image, rendered natively on WebGPU.

1.6 KB
Raw Blame

@framefields/ai

AI models (images, video, music, sound effects, voice, upscaling, background removal) through the framefields AI gateway, billed to your framefields account. No dependencies. Use it from the main package:

import { ai } from "framefields";

// FRAMEFIELDS_API_KEY from the environment, or:
ai.config({ apiKey: process.env.FRAMEFIELDS_API_KEY });

const { data } = await ai.subscribe<{ images: { url: string }[] }>("flux/schnell", {
  input: { prompt: "Neon street at night" },
});
console.log(data.images[0].url);
Method Use
ai.run(model, { input }) One request; best for fast models.
ai.subscribe(model, { input, onQueueUpdate, logs }) Queues the job, polls, returns the output. Use for video and music.
ai.stream(model, { input }) Iterate partial output; await stream.done() for the last.
ai.queue.submit / status / result / cancel Run many jobs and collect them later by requestId.
ai.storage.upload(blob) Upload a local file and pass the URL as input. Blobs inside input upload automatically.

Failed calls throw ApiError (status, type, body, requestId), or ValidationError with fieldErrors when the input doesn't match the model's schema. type is the gateway's code, e.g. insufficient_credits (top up at https://framefields.dev/dashboard/billing) or rate_limited.

Model ids and their input schemas: https://framefields.dev/models. Keys: https://framefields.dev/dashboard/keys.

FRAMEFIELDS_AI_PROXY_URL overrides the gateway URL, for running the gateway locally.