@nodetool-ai/transformers-js-provider

Local-models BaseProvider backed by @huggingface/transformers and kokoro-js (chat, TTS, ASR, embeddings) for NodeTool.

Runs models locally in-process — no API keys, no network — via ONNX through @huggingface/transformers, with kokoro-js for text-to-speech. Registers a transformers_js provider on the NodeTool runtime registry, so chat, TTS, ASR, and embedding calls route to local inference like any other provider.

Install

npm install @nodetool-ai/transformers-js-provider

Exported symbols

| Symbol | Kind | Description | | --- | --- | --- | | TransformersJsProvider | class | Local-models BaseProvider (chat, TTS, ASR, embeddings) | | registerTransformersJsProvider | function | Register the provider as transformers_js on the runtime registry (idempotent) | | discoverLanguageModels | function | Discover available local chat models | | discoverTTSModels | function | Discover available local text-to-speech models | | discoverASRModels | function | Discover available local speech-recognition models | | discoverEmbeddingModels | function | Discover available local embedding models |

Usage

import { registerTransformersJsProvider } from "@nodetool-ai/transformers-js-provider";
import { getRegisteredProvider } from "@nodetool-ai/runtime";

// During server bootstrap
registerTransformersJsProvider();

// Later, resolve the provider class from the registry and use it like any other
const ProviderClass = getRegisteredProvider("transformers_js");

Links

• NodeTool
• GitHub