DocumentEmbeddingsServiceImpl
A service that is used by the DocumentEmbeddings processor to generate embeddings for the text in your documents.
This service uses an LLM from Hugging Face, which NiFi downloads and caches when the service first runs.
Properties
Name | Default Value | Description |
---|---|---|
Idol License Service | An IdolLicenseServiceImpl that provides a way to communicate with an IDOL License Server. | |
Model Name |
The name of the model to use to generate embeddings. This model name must match exactly the name of a model on Hugging Face. You can choose any sentence-transforme r model (OpenText recommends the sentence-transformers/sentence-t5-large model: https://huggingface.co/sentence-transformers/sentence-t5-large). The first time that you use this model, NiFi Ingest downloads it from Hugging Face. It then caches it for future use. |
|
Cache Directory | ~/.cache/huggingface/hub | The location where NiFi Ingest caches your model files. |
Device | CPU | The type of processor to use process your embeddings model, cpu or cuda. If you set this option to cuda, you must have a CUDA-compatible GPU. |
Embedding Precision | The number of decimal places to use in the embedding values. You can use a value between 1 and 10 (inclusive). | |
Model Max Sequence Length |
The maximum chunk size permitted by the model that you use to generate embeddings. |
|
Model Minimum Final Sequence Length | The minimum length of the final chunk of text used to generate embeddings, when multiple embeddings are required for a piece of text. | |
Access Token | An access token for Hugging Face to use to access a private model. This option allows you to use private models that you have access to. | |
Model Revision | main | The version of the model to use from Hugging Face. This value can be a branch name, tag name, or commit ID for the model. |
Model Sequence Overlap |
The length of overlap required for text used to generate successive embeddings, when multiple embeddings are required for a piece of text. |
|
Offline Mode | false | Set to true to use only a locally cached version of the configured model (and never download an updated version). In this case, you must have the required models cached in the appropriate place. |
Data Limit | -1 (unlimited) | The maximum start offset (in bytes) of text used to generate any embedding input. You can use this option to limit the amount of processing if the processor receives a large file. Set this value to a positive integer, or -1 for no limit. |