Semantics Search with Gemini

Semantics Search with Gemini

Comment

Inspiration

The inspiration for this project came from the frustration of navigating large amounts of text on websites. We wanted to create a tool that goes beyond traditional Ctrl + F searches by providing semantic search capabilities. The idea of leveraging the Gemini model to fetch contextually relevant information made this an exciting and challenging project.

What it does

  • Performs semantic keyword searches using the Gemini API, retrieving relevant content from websites.
  • Provides highlighting and auto-jump functionality to navigate results efficiently.
  • Allows users to toggle between semantic search and exact match search for more precise control.
  • Offers keyboard shortcuts (Cmd + Shift + Y for Mac, Ctrl + Shift + Y for Windows) to quickly activate the extension.

How we built it

  1. Core Features:
  2. Integrated the Gemini model to process user-provided keywords and extract related website content.
  3. Used the mark.js library to implement dynamic keyword highlighting and auto-jump functionality.
  4. User Interface:
  5. Built an intuitive popup UI using HTML and CSS for easy interaction.
  6. Added toggle options for switching between semantic and exact match modes.
  7. Optimization:
  8. Bundled the project with webpack for efficient performance and loading.
  9. Cached API results to minimize redundant calls.
  10. Testing:
  11. Tested the extension across various websites to ensure accuracy and performance.

Challenges we ran into

  • Semantic Search Integration: Understanding and properly implementing the Gemini API for contextual searches was initially complex.
  • Performance Optimization: Large websites caused performance bottlenecks when rendering highlights. We solved this by optimizing DOM manipulations and API calls.
  • Cross-Platform Shortcuts: Ensuring consistent behavior for keyboard shortcuts across Mac and Windows required careful debugging.
  • UI/UX Design: Designing a simple yet powerful interface that accommodates semantic search without overwhelming the user was a key challenge.

What we learned

  • How to effectively integrate and work with Google’s Gemini API for semantic processing.
  • Techniques for optimizing performance in browser extensions, including DOM manipulations and API call minimization.
  • Building and deploying Chrome Extensions with an emphasis on user-centric design.
  • The importance of cross-platform compatibility when implementing keyboard shortcuts.

What's next for Semantics Search with Gemini

  • Add PDF Support: Extend the extension’s functionality to support semantic searches in PDFs directly within the browser.
  • Fine-Tune the Gemini Model: Customize and fine-tune the Gemini model to improve semantic search accuracy and responsiveness.
  • Enhance UI: Introduce additional user settings, such as customizable highlight colors and improved toggle options.

Built With

Share this project:

Updates