Show HN: Digital Twin Proxy – Turn web browsing into memory for AI agents

3 points

10 months ago

I built Digital Twin Proxy as an open-source tool that logs your web browsing and uses a local LLM to generate an analysis of your browsing patterns. The goal is to create a "digital twin" of your online activity that can be used to provide real-time context to other AI agents.

GitHub repo: https://github.com/kstonekuan/digital-twin-proxy

How it works: It uses a Squid proxy to capture HTTP/S traffic, and then a Rust application sends the logs to a local LLM (via Ollama) to analyze and create a digital twin based on web browsing patterns.

Why I built this: I wanted a way to give my AI agents more personalized and real-time context. By analyzing my web browsing, agents can tailor their responses, anticipate my needs, and even get better at using tools through "context engineering."

Key features:

- HTTP/S Traffic Logging: Captures all web requests.

- AI-Powered Analysis: Uses a local LLM (Ollama) to analyze traffic.

- Flexible Operation Modes: Can run in the background, log continuously, or perform one-off analysis.

- Customizable: You can easily change the AI model, analysis interval, and other settings.

Planned features:

- Exposing the context from your digital twin as an MCP server for AI agents.

- Injecting real-time context directly into your browser for any agentic AI app (like ChatGPT or Perplexity) to access.

Feedback and contributions are welcome!