Oracle Database Vector Store icon

Oracle Database Vector Store

Oracle Database Vector Search
Join our community

Overview

This node integrates with an Oracle Database to perform vector similarity searches using stored embeddings. It is designed to take embedding data as input and query a specified Oracle database table to find vectors similar to the input, based on a chosen distance strategy.

Common scenarios where this node is beneficial include:

  • Semantic search applications where you want to find documents or items similar to a given query embedding.
  • Recommendation systems that rely on vector similarity.
  • Any AI workflow requiring fast nearest neighbor search within an Oracle database.

For example, you might use this node to find the most semantically similar customer reviews stored in an Oracle DB when given a new review embedding.

Properties

Name Meaning
Table Name The name of the Oracle database table containing the vector embeddings to search against.
Distance Strategy The method used to measure similarity between vectors. Options: Cosine, Euclidean, Dot Product

Output

The node outputs a JSON object under the response field which contains the results of the vector similarity search performed on the Oracle database. This includes the matched vectors and their similarity scores according to the selected distance strategy.

No binary data output is indicated by the source code.

Dependencies

  • Requires the oracledb Node.js module to connect to Oracle Database.
  • Uses a custom OracleDbVectorStore class for managing vector storage and querying.
  • Requires credentials for connecting to an Oracle Database instance (host, port, user, password, service name).
  • The Oracle connection string is constructed dynamically depending on whether the database is autonomous or not.

Troubleshooting

  • Missing OracleDB Module: If the oracledb package is not installed, the node will throw an error instructing to install it via npm install oracledb.
  • Connection Errors: Incorrect credentials or network issues may cause connection failures. Verify Oracle DB credentials and network accessibility.
  • Table Not Found: Ensure the specified table name exists in the Oracle database and contains the expected vector data.
  • Invalid Distance Strategy: Only "cosine", "euclidean", and "dot" are supported. Using other values may cause errors.
  • Embedding Input Issues: The node expects valid embedding data from its input connection; missing or malformed embeddings can cause failures.

Links and References

Discussion