UniversalEmbedder performs multimodal embedding extraction directly using LiteRT LM. **Signature:** export declare class UniversalEmbedder ## Methods | Method | Modifiers | Description | |---|---|---| | [createFromOptions(wasmFileset, options)](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembeddercreatefromoptions) | `static` | Initializes the Wasm runtime and creates a new UniversalEmbedder instance. | | [createWebGpuDevice()](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembeddercreatewebgpudevice) | `static` | Creates a WebGPU device for GPU-accelerated universal multimodal embedding. | | [createFromModelBuffer(wasmFileset, modelAssetBuffer)](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembeddercreatefrommodelbuffer) | `static` | Initializes the Wasm runtime and creates a UniversalEmbedder from a model buffer. | | [createFromModelPath(wasmFileset, modelAssetPath)](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembeddercreatefrommodelpath) | `static` | Initializes the Wasm runtime and creates a UniversalEmbedder from a model path. | | [embedText(text)](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembedderembedtext) | | Performs embedding extraction on the input text. | | [embedImage(imageBytes)](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembedderembedimage) | | Performs embedding extraction on raw image bytes. | | [embedAudio(audioData)](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembedderembedaudio) | | Performs embedding extraction on audio float samples. | | [embedContent(content)](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembedderembedcontent) | | Performs embedding extraction on multimodal content parts. | | [cosineSimilarity(u, v)](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembeddercosinesimilarity) | `static` | Computes cosine similarity between two embeddings. | | [getProvider()](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembeddergetprovider) | | Returns an EmbeddingProvider for this embedder. | | [close()](https://developers.google.com/edge/api/mediapipe/js/tasks-retrieval.universalembedder#universalembedderclose) | | Shuts down the UniversalEmbedder and releases resources. | ## UniversalEmbedder.createFromOptions() {#universalembeddercreatefromoptions} Initializes the Wasm runtime and creates a new UniversalEmbedder instance. **Signature:** static createFromOptions(wasmFileset: WasmFileset, options: UniversalEmbedderOptions): Promise; ### Parameters | Parameter | Type | Description | |---|---|---| | wasmFileset | WasmFileset | | | options | UniversalEmbedderOptions | | **Returns:** Promise\ ## UniversalEmbedder.createWebGpuDevice() {#universalembeddercreatewebgpudevice} Creates a WebGPU device for GPU-accelerated universal multimodal embedding. **Signature:** static createWebGpuDevice(): Promise; **Returns:** Promise\ ## UniversalEmbedder.createFromModelBuffer() {#universalembeddercreatefrommodelbuffer} Initializes the Wasm runtime and creates a UniversalEmbedder from a model buffer. **Signature:** static createFromModelBuffer(wasmFileset: WasmFileset, modelAssetBuffer: Uint8Array | ReadableStreamDefaultReader): Promise; ### Parameters | Parameter | Type | Description | |---|---|---| | wasmFileset | WasmFileset | | | modelAssetBuffer | Uint8Array \| ReadableStreamDefaultReader | | **Returns:** Promise\ ## UniversalEmbedder.createFromModelPath() {#universalembeddercreatefrommodelpath} Initializes the Wasm runtime and creates a UniversalEmbedder from a model path. **Signature:** static createFromModelPath(wasmFileset: WasmFileset, modelAssetPath: string | string): Promise; ### Parameters | Parameter | Type | Description | |---|---|---| | wasmFileset | WasmFileset | | | modelAssetPath | string \| string | | **Returns:** Promise\ ## UniversalEmbedder.embedText() {#universalembedderembedtext} Performs embedding extraction on the input text. **Signature:** embedText(text: string): Promise; ### Parameters | Parameter | Type | Description | |---|---|---| | text | string | | **Returns:** Promise\ ## UniversalEmbedder.embedImage() {#universalembedderembedimage} Performs embedding extraction on raw image bytes. **Signature:** embedImage(imageBytes: Uint8Array): Promise; ### Parameters | Parameter | Type | Description | |---|---|---| | imageBytes | Uint8Array | | **Returns:** Promise\ ## UniversalEmbedder.embedAudio() {#universalembedderembedaudio} Performs embedding extraction on audio float samples. **Signature:** embedAudio(audioData: Float32Array): Promise; ### Parameters | Parameter | Type | Description | |---|---|---| | audioData | Float32Array | | **Returns:** Promise\ ## UniversalEmbedder.embedContent() {#universalembedderembedcontent} Performs embedding extraction on multimodal content parts. **Signature:** embedContent(content: readonly ContentPart[]): Promise; ### Parameters | Parameter | Type | Description | |---|---|---| | content | readonly ContentPart\[\] | | **Returns:** Promise\ ## UniversalEmbedder.cosineSimilarity() {#universalembeddercosinesimilarity} Computes cosine similarity between two embeddings. **Signature:** static cosineSimilarity(u: Embedding, v: Embedding): number; ### Parameters | Parameter | Type | Description | |---|---|---| | u | Embedding | | | v | Embedding | | **Returns:** number ## UniversalEmbedder.getProvider() {#universalembeddergetprovider} Returns an EmbeddingProvider for this embedder. **Signature:** getProvider(): EmbeddingProvider; **Returns:** EmbeddingProvider ## UniversalEmbedder.close() {#universalembedderclose} Shuts down the UniversalEmbedder and releases resources. **Signature:** close(): void; **Returns:** void