Skip to content
Nappai Documentation

NVIDIA Embeddings

The NVIDIA Embeddings component lets you turn text into numerical vectors using NVIDIA’s powerful AI models. These vectors can be used for search, clustering, or any task that needs a numeric representation of text.

How it Works

When you add this component to a workflow, it sends your text to NVIDIA’s cloud API.

  • Model – Choose which NVIDIA embedding model to use (e.g., nvidia/nv-embed-v1 or snowflake/arctic-embed-I).
  • Base URL – The endpoint where the API is hosted (default is https://integrate.api.nvidia.com/v1).
  • Temperature – Controls how deterministic the embeddings are; lower values give more consistent results.

The component then receives a list of vectors (embeddings) that represent the input text. These vectors can be passed to other components such as similarity search, clustering, or storage.

Inputs

The component requires a credential of type NVIDIA API.

  1. Configure the NVIDIA API credential in the Credentials section of Nappai.
  2. Select that credential in the component’s Credential field.

Input Fields

  • NVIDIA Base URL: The API endpoint to connect to.
  • Model: The specific NVIDIA embedding model you want to use.
  • Model Temperature: A numeric value that influences the randomness of the embeddings (recommended to keep it low for consistent results).

Outputs

  • Embeddings: A list of numerical vectors that represent the input text. These can be used for similarity comparison, clustering, or stored for later retrieval.

Usage Example

  1. Add the NVIDIA Embeddings component to your workflow.
  2. Set the Model to nvidia/nv-embed-v1.
  3. Leave the Base URL at its default value.
  4. Set Temperature to 0.1 for stable embeddings.
  5. Connect the output of a Text Input component to the NVIDIA Embeddings component.
  6. Use the resulting Embeddings as input to a similarity search component or store them in a database for later use.

This simple flow turns raw text into vectors that can be compared or indexed, enabling powerful AI-driven search and recommendation features.

  • OpenAI Embeddings – Uses OpenAI’s models for generating embeddings.
  • Cohere Embeddings – Generates embeddings with Cohere’s language models.
  • SentenceTransformers Embeddings – Uses Hugging Face models for embeddings.

These components share a similar interface and can be swapped depending on the model provider you prefer.

Tips and Best Practices

  • Choose the right model: nvidia/nv-embed-v1 is great for general-purpose embeddings, while snowflake/arctic-embed-I may offer better performance for certain domains.
  • Keep temperature low (e.g., 0.1) to get consistent embeddings for the same text.
  • Batch your requests: If you have many texts, batch them to reduce API calls and improve speed.
  • Validate the output: Check the shape and dimensionality of the embeddings to ensure they match downstream components.

Security Considerations

  • The NVIDIA API key is stored securely in Nappai’s credential store and never exposed in the workflow.
  • Always use the Credential field to reference the stored key instead of hard‑coding it.
  • Monitor API usage in NVIDIA’s dashboard to detect any unexpected activity.