SF-agent
My search agent that understands user queries and finds information
What it does
SF-agent: Intelligent Video Content Search
SF-agent is an advanced search tool for finding specific video content. It uses a repository of embedded video subtitles and a database of subtitle timestamps and video identifiers.
How it works:
User submits a search query
Query is embedded and converted to a vector
System compares this vector with stored subtitle vectors
Most similar subtitles are identified
Corresponding video timestamps and numbers are retrieved
The SF-agent then provides users with the exact location of relevant content, eliminating manual searching.
Key Features:
Efficient: Quickly locates specific video segments
Accurate: Uses advanced embedding for precise matching
Versatile: Handles various query types
Gemini API Integration:
We integrated the Gemini API to refine user queries, addressing diverse search behaviors. Users may input keywords, use complete sentences, or have limited video content knowledge. Our custom prompt:
Takes the initial user query
References it against stored subtitle files
Generates a more specific version
This refined query is used for searching, yielding more accurate results.
Benefits:
Enhanced search experience
Efficient location of precise video segments
Accommodates various user familiarity levels and query styles
SF-agent transforms video content search, making it faster, more accurate, and user-friendly.
Built with
- None
Team
By
Aigenda
From
South Korea