Jina - Generate embeddings action
The Generate embeddings action creates vector embeddings for text in Jina.
Input
| Input field | Description |
|---|---|
| Model | Select an embedding model, or enter a custom model ID. |
| Input | Enter one or more strings to embed. |
| Task | Optional. Select the intended use for the embedding, such as Retrieval query, Retrieval passage, or Classification. |
| Dimensions | Optional. Enter the number of dimensions for the output embedding. |
| Embedding type | Optional. Select the embedding encoding. Options are float, base64, binary, or ubinary. |
| Normalized | Optional. Select Yes or No to normalize the embedding vector. |
| Late chunking | Optional. Select Yes or No to enable late chunking. |
| Truncate | Optional. Select Yes or No to truncate input that exceeds the model's token limit. |
| Input type | Optional. Select the input type for ColBERT-style models. Options are Query or Document. |
| Return multivector | Optional. Select Yes or No to return multi-vector embeddings. |
| Return tokenized input | Optional. Select Yes or No to return the tokenized input. |
Output
The following are a representative set of output fields. Refer to the Jina Embeddings documentation for more information on available output fields.
| Output field | Description |
|---|---|
| Model | The model used. |
| Object | The response type, for example list. |
| Usage | Token usage. |
| Total tokens (Usage) | Token usage. |
| Data | An array of embedding results. |
| Object (Data) | The result type. Typically embedding or embeddings when Return multivector is Yes. |
| Index (Data) | The input's position in the Input array, starting at 0. |
| Embedding (Data) | The embedding vector. |
Last updated: