Skip to main content

Vector search

Vector search enables users to examine how the bot constructs its responses. Key features include:

  1. Question Input: Users can input a question to analyze how the system generates an answer, including identifying which chunks of information are used to construct the response.

  2. Score Threshold and Result Configuration: Users have control over the response generation by adjusting the score threshold, which defines how relevant or confident a chunk must be to be included. Additionally, the number of results can be configured, determining how many chunks will be displayed for review.

  3. Chunk Display: Each response is constructed from chunks of information, which are displayed with the following details:

    • Document Source: Displays the name of the document the chunk was extracted from, helping users trace the origin of the information.
    • Content: Shows the specific text within the chunk that contributes to the response.
    • Relevance Score and Character Count: Found in the lower-left corner, this shows how relevant the chunk is to the query and the size of the chunk (number of characters).

  4. Edit Capability: Users can refine or correct the information by clicking the edit icon located in the upper-right corner of each chunk. This allows direct modification of the content, providing flexibility in tailoring the response.