Skip to content
Nappai Documentation

Hugging Face Embeddings

This component helps Nappai understand the meaning of text. It turns text into numbers that computers can easily work with. This allows Nappai to perform tasks like comparing how similar different pieces of text are, or finding specific information within large amounts of text.

Relationship with Hugging Face API

This component uses the Hugging Face API, a powerful service that provides access to many different language models. These models are trained to understand the nuances of human language. Our component uses these models to create the numerical representations (embeddings) of your text.

Inputs

  • API Key: This is a secret code that allows Nappai to access the Hugging Face API. You’ll need to obtain this from Hugging Face. This setting is considered advanced and is hidden by default.
  • API URL: This is the web address of the Hugging Face API. You usually don’t need to change this; the default value is already set. This setting is considered advanced and is hidden by default.
  • Model Name: This specifies which language model Hugging Face should use to create the embeddings. A default model is provided, but you can change it if you need a different model for your specific needs.
  • Credential: This is a reference to your Hugging Face API credentials, which are managed within Nappai. Follow the instructions provided to set this up correctly.

Outputs

The component produces Embeddings. These are numerical representations of your input text. These numbers are then used by other Nappai components to perform tasks such as:

  • Finding similar text: Comparing the embeddings of different texts to see how similar they are.
  • Semantic search: Searching for information based on meaning, rather than just keywords.
  • Text classification: Categorizing text based on its content.

Usage Example

Imagine you want to find documents similar to a specific customer query. You would feed the customer query into this component. The component would then generate embeddings representing the meaning of the query. These embeddings are then passed to another component (like a vector database) to find similar documents based on their embeddings.

Templates

This component is used in many Nappai workflows. You can find it in various templates designed for tasks involving text analysis and semantic search. Check the template documentation for specific examples.

  • Semantic Text Splitter: This component breaks down large texts into smaller, more manageable chunks before sending them to the Hugging Face Embeddings component.
  • Couchbase, Upstash, Chroma DB, Weaviate, Vectara, Redis, PGVector, FAISS, Astra DB, Qdrant, Pinecone, MongoDB Atlas, Milvus, Supabase, Cassandra, Text Embedder: These components are vector databases or other embedding-related tools that work with the embeddings generated by this component to store and search for information.

Tips and Best Practices

  • Use the default settings unless you have a specific reason to change them.
  • Ensure your Hugging Face API key is correctly configured.
  • Consider using the semantic text splitter for very long texts to improve performance and accuracy.

Security Considerations

Protect your Hugging Face API key. Do not share it with unauthorized individuals. Nappai’s secure handling of secrets helps protect this sensitive information.