Warning: I will flag out when it could be fit for everybody to use with one-click manner. For time being, it's still a toy of IT professionals because it's risky to be used by non-IT people.
The diagram shows above is an ideal and high level workflow but the key is how to ensure LLM to maximize its capability by using context.
- version 1 which is a chatbot to consume PDF inspired by OpenAI.
- version 2 which is a web UI to put two LLM in and human player to manage two AI chat with each other.
- version 3 which is the baseline of 2nd brain.
All designs here is based on version 3. And thanks for Gemini as my copilot.
the application structure:
- Activity Capture Module: handles capturing user activity within the browser using browser extension APIs
- Communication Module: Manages communication with the background script, sending activity and receiving responses.
- LLM Module: to provide AI.
- Knowledge Base Module: the core and key place to mix context you're surfing internet with AI's understanding together with your insutrction (protect, support, recommend)
- watch on me when I'm surfing internet.
- provide reference, recommendation and protection at the side window along with my browser.
- No online LLM is used and no private data is sent out of my laptop/pc.
- persist my thinking, my preference locally and provide a UI to allow me to adjust it for better help me and protect me.
- Google Chrome (128.0.6613.84 (Official Build) (64-bit))
coming soon...