A Chrome extension that turns browsing history into a conversational, AI-powered memory layer. Ask natural-language questions about pages you have already visited, retrace research trails from fuzzy memories rather than exact keywords, and return to prior work through a privacy-aware stack that combines on-device Rust/WebAssembly processing with a serverless Go backend.
Chromium History Chat is a Manifest V3 Chrome extension for asking natural-language questions about pages you have already visited. It processes history client-side with Rust compiled to WebAssembly, then pairs it with a Go Google Cloud Function for OpenAI requests, rate limiting, and credential management—so users can query their browsing history without managing backend infrastructure.
Chrome Extension (Manifest V3), Rust to WebAssembly, Dart compiled to JavaScript for UI and service worker, Go (Google Cloud Functions Gen 2), Google Secret Manager, OpenAI GPT-4o-mini, HTTPS-only transport.
Published on the Chrome Web Store with a public GitHub repository, a dedicated privacy policy route, and a portfolio post that documents both the product story and the architecture behind it.
The product, client architecture, and backend + AI layers work together to make history useful again.
Chromium History Chat lets you have a real conversation with your browsing history. Rather than remembering the right keyword, digging through old tabs, or scrolling the default History page, ask a plain-English question and get an answer backed by pages you actually visited. No exact URL, title, or search query required—describe what you were doing or remember, and the extension can help you find it again.
The default History view treats prior browsing as a timestamped list. Chromium History Chat turns it into a searchable, conversational memory: instead of "I know I saw this somewhere..." you can ask, "Past me already researched this—what did I find?"
This extension works with your browsing history, so it needs permission to read history entries. Without that access, there is nothing to chat with. The permission lets the product understand pages you visited, organize them into a reusable index, and answer questions grounded in your browsing rather than generic web search.
This project delivers a full vertical slice: browser-native UX, semantic recall over real user history, on-device WASM performance, and a production-shaped serverless backend with rate limits and secret management. It answers "I know I saw this somewhere" without asking users to change how they browse.
Shipping a Chrome Web Store extension with Rust/WASM, a Go Cloud Function, and a clear privacy story shows end-to-end ownership across client performance, cloud operations, and AI product design. It is a product-shaped implementation with an actual distribution channel, an explicit permissions narrative, and a real user workflow.
