Skip to main content
Vector databases are specialized storage systems designed to handle vector embeddings that represent data in high-dimensional space. Vector databases are integral to modern artificial intelligence (AI) applications, offering the ability to efficiently manage and query high-dimensional data. By leveraging Domino’s platform capabilities, users can implement sophisticated AI solutions that require fast and accurate similarity searches, enhancing both the performance and scalability of their applications. Domino supports a wide array of vector databases, essential for powering advanced AI applications such as retrieval augmented generation (RAG), recommender systems, and media recognition tasks. This documentation outlines the process of efficiently indexing, loading, and retrieving data from vector databases using the robust capabilities of the Domino platform. Vector databases provide:
  • Efficient similarity searches: Quickly find the most similar items in large datasets, which is essential for applications like RAG and personalized recommendation systems.
  • Scalable infrastructure: Handle large volumes of data and complex query patterns without sacrificing performance.
  • Enhanced data handling: Store and manage high-dimensional data effectively, optimizing storage and retrieval operations.

​
Set up vector databases

Set up vector databases in Domino with the following steps:

​
Step 1: Choose a vector database

Select a vector database that best fits your application needs. Domino supports popular vector databases like Pinecone, QDrant, and others, offering a range of features tailored to different requirements.

​
Step 2: Index data

  • Prepare your data: Organize your data into a suitable format for vectorization. This might involve preprocessing steps like normalization or tokenization, depending on the nature of your data.
  • Generate embeddings: Use a machine learning model to convert your prepared data into vector embeddings. Domino can facilitate this process through its scalable compute resources and integration with machine learning frameworks.
  • Load data into the database: Upload the generated embeddings into your chosen vector database. Domino’s job scheduler can automate this process, ensuring data is indexed efficiently and regularly updated as needed.

​
Step 3: Integrate a vector database

  • Configure access: Set up connectivity between Domino and the vector database, ensuring secure and reliable data transfer.
  • Implement API calls: Use Domino’s APIs to query the vector database directly from your applications or during analytical workflows.

​
Step 4: Retrieve data

  • Execute queries: Perform queries to retrieve data based on similarity or other criteria, integrating these operations within your Domino Workspaces or Runs.
  • Utilize results: Use the retrieved data to enhance your AI applications, whether it’s generating responses in a RAG setup, recommending products, or recognizing images or videos.

​
Use cases

  • RAG: Enhance chatbots and other generative AI applications by providing contextually relevant data at runtime.
  • Recommender systems: Improve recommendation accuracy by matching user profiles with products or content based on similarity in vector space.
  • Media recognition: Quickly identify and classify media files by comparing embeddings generated from audio or video content.

​
Next steps

Find out how to connect to Pinecone and Qdrant.