Inspiration
I was tired of scrolling endlessly through my Chrome history to find that one article — the one about Amelia Earhart or that AI paper I read last week. Echo was born from that frustration. Instead of searching by title or date, why not simply ask your browser what you’re looking for and let it recall it for you?
What it does
Echo lets you converse with your browsing memory. It can index pages locally in an on-device database or seamlessly tap into Chrome’s built-in history. Using fuzzy matching and on-device AI summarization, Echo can recall past pages, articles, and videos — giving you a summary and a reason why each result matches your query. All processing happens entirely on your device, preserving privacy while feeling almost magical.
How we built it
Echo is built with JavaScript, HTML, and CSS, powered by the Chrome Extensions API and the new Built-in AI APIs — specifically:
Summarizer API (Gemini Nano) to generate short, key-point summaries of visited pages.
LanguageModel Prompt API (Gemini Nano) to semantically re-rank search results and interpret natural-language queries. Data is stored locally using IndexedDB and the Chrome Storage API, while the Side Panel API provides an immersive “memory view” inside the browser.
Challenges we ran into
Chrome’s new sidePanel API had strict gesture and version constraints — opening it from within the popup required surgical timing and creative fallbacks. Integrating Gemini Nano’s Summarizer and Prompt APIs also demanded careful handling of availability, throttling, and offline behavior.
Accomplishments that we're proud of
Running fully on-device AI summarization and reasoning with zero external calls.
Seamlessly blending Chrome history + local indexed data into a single searchable memory.
Designing a UI that feels native — light, fast, and privacy-first
What we learned
We learned how to bridge local browser data with embedded AI models to deliver powerful context-aware search — all without a server. We also discovered how Chrome’s new APIs (Summarizer, Prompt, and SidePanel) can transform extensions from passive tools into intelligent assistant
What's next for Echo
Add multimodal memory using Gemini’s image and audio inputs.
Introduce temporal filtering (“Show me articles I read before my trip”).
Enable topic tagging and conversation continuity for long-term recall.
Explore hybrid AI with Gemini Developer API + Nano for deeper contextual search.
Log in or sign up for Devpost to join the conversation.