Save vectors
Generates and saves vector records to a SQL Server / Azure SQL database from text input.
Example
This flow reads files from OneDrive, converts their content to text, splits it into chunks, generates embedding vectors, and stores them in a SQL server database for semantic search.
This action allows saving a text as a list of records in a SQL server vector collection, in a specific format (Profitbase). This is done using the following steps:
- Split - The input text is split into chunks using the Text splitter (link).
- Embed - Each of the chunks resulting from the previous step is embedded using the Text embedder (link)
- Build - A new collection record is built for each of the embedded chunk.
- Upsert - The newly created record is inserted in the database collection.
Properties
Name | Type | Description |
---|---|---|
Title | Optional | The title or name of the action. |
Connection | Required | Select or define your SQL server connection. |
Enable dynamic connection | Optional | A 'Dynamic Connection' will override the 'Connection' on flow execution. |
Input text | Required | Text to vectorize and save. Can be a variable or static string. |
Collection name | Required | The name of the table or collection in the database where vectors are saved. |
Record definition | Optional | The definition of columns used in the vector search. Note: The vector column is not returned. |
Description | Optional | Explanation of what this action does. |
Collection Schema
The collection has the following columns:
key (String):
Contains the record's key. This is automatically generated by the node.content (String):
Contains the plain-text of a chunk.vector (Vector):
Contains the vector representation of thecontent
column.context (String):
Contains the context of the text inputted in the node. This can be a document name, a URL, or a generic ID representing the resource to which the text pertains.additional columns:
For each custom column defined in the node, a new collection column will be created (including the specified data type).