Skip to content

Latest commit

 

History

History
57 lines (35 loc) · 3.04 KB

README.md

File metadata and controls

57 lines (35 loc) · 3.04 KB
Raw

Webgati AI

Chrome Web Store

NOTE: This AI-powered extension is under active development, please verify the results as necessary and provide feedback to help us improve.

Webgati AI is a Chrome & MS Edge browser extension for chatting with any webpage. Whether you're researching, studying, or just exploring diverse content, Webgati AI is designed to efficiently guide you through webpages by providing quick summaries and answers directly from the content.

Features

  • Interactive Sidebar: Open a chat by clicking the extension icon, ask questions or seek summaries from any webpage effortlessly.

  • Content Selection Dialog: Highlight text, and a dialog box will pop up to allow direct interaction with the content.

  • Vision chat: Easily ask questions about the visual information on the page (images, graphs, etc.) by using the extension to take a screenshot or select an area of the page.

  • Privacy-Focused:

    • Use private local models via Ollama for complete privacy
    • OR use your own API key to access cloud-based model providers, your chat is directly sent to their API, not to any other server, thus ensuring your data remains private and secure. (Currently OpenAI and Anthropic are supported, more model providers will be available soon, reach out for any specific model request)

  • Open Source: Community-driven and transparent, allowing for continuous improvement and collective auditing.

Demo

Demo Video

Roadmap

  • Save chat messages to local storage
  • Custom prompts/instructions for each website
  • Index and chat with all pages in a website
  • Add more model providers including multi-lingual and domain-specific models
  • Web agents for automated tasks

Ollama Setup

  • Install Ollama
  • Run OLLAMA_ORIGINS=* OLLAMA_HOST=127.0.0.1:11434 ollama serve

Local Setup

  • npm install
  • npm run build:dev (Hot reloading with npm run dev will not work due to WASM dependencies needed for in-memory models via transformers.js)
  • Load the extension from dist folder using Load unpacked option in Chrome extensions (instructions)

Tech/Frameworks

Contributing

This project is open to and encourages contributions to benefit the wider community! Feel free to reach out to get started.

Feedback

Please provide feedback by creating an issue, sending an email to [email protected] or contacting sqzsh269