VantageCloud Lake | Vector store in AWS - Creating a Vector Store in AWS - Teradata VantageCloud Lake

Lake - Analyze Your Data with ClearScape Analytics™
Deployment
VantageCloud
Edition
Lake
Product
Teradata VantageCloud Lake
Release Number
Published
February 2025
ft:locale
en-US
ft:lastEdition
2026-02-20
dita:mapPath
tcl1683670667798.ditamap
dita:ditavalPath
pny1626732985837.ditaval
dita:id
tcl1683670667798
PrerequisiteEnsure that the organization administrator has enabled the vector store for your environment. The administrator can enable it by selecting Enable from the Vector store option in the Settings menu of the environment.

To create a Vector Store in AWS environment:

  1. Sign in into the VantageCloud Lake console.
  2. In the Environments, select the environment where you need to create a vector store.
  3. Select Vector stores in the Manage data section.
  4. Select + to create a new vector store.
  5. Enter the Name and Description for the vector store.
  6. Select Content based for the vector store creation.
  7. Select either Prepared data or Raw data and select Next.

    The Prepared data is a pre-loaded data that exists in a database table and you can upload chunked text into a VARCHAR or CHAR column in a table that resides in the same database as the vector store.

    The Raw data is a file-based Vector store creation method that uses a PDF file on your local storage which is converted to vectors.

    After creating a vector store, you cannot change its content type.
  8. Select the required database and tables in Data sources.
  9. In the Settings, select Vector distance, K-means or HNSW as the Search algorithm. See Teradata® Enterprise Vector Store User Guide for more details.

    Vector distance

    Once you select, you can continue with the defaults or select required values by disabling Use defaults. You can modify:
    • Similarity matches: Select the number (between 1 to 1024) of similarity matches to generate.
    • Similarity threshold: To consider matching tables and views.
    • General settings: Select Euclidean, or Cosine, or Dotproduct. See Teradata® Enterprise Vector Store User Guide for more details.
    You can generate the vector store in batches by selecting Use batch processing for embedding generations.

    K-means

    Once you select, you can continue with the defaults or select required values by disabling Use defaults. You can modify:
    • Indexing Settings: Select the required values, such as, number of clusters to train, number of iterations to run, initial seed value, Centroid initialization algorithm (Random or Kmeans++), and so on.
    • Search settings: Select the training cluster to count, similarity matches, and threshold values.
    • General settings: Select Euclidean, or Cosine, or Dotproduct. See Teradata® Enterprise Vector Store User Guide for more details.
    You can generate the vector store in batches by selecting Use batch processing for embedding generations.

    HNSW

    Once you select, you can continue with the defaults or select required values by disabling Use defaults. You can modify:
    • Indexing settings: Select the number of layers for the graph, number of node connections, EF construction values, and so on.
    • Search settings: Select the similarity matches and EF search values.
    • General settings: Select Euclidean, or Cosine, or Dotproduct. See Teradata® Enterprise Vector Store User Guide for more details.
    You can generate the vector store in batches by selecting Use batch processing for embedding generations.
  10. [Optional: If required, modify the settings by selecting Previous.
  11. Select Create to create the vector store with the previously set settings.

    Once created, you can view the newly created vector store in the Vector stores page.