Choose theme
AI Resources
Gemma 4 Browser Extension
Gemma 4 Browser Extension is an independent experiment by nico-martin for asking about the page you are reading, finding previously indexed pages and operating browser tabs from a Chrome side panel.
It runs a Gemma 4 ONNX model through Transformers.js and WebGPU in the browser. Local model inference and what a browser tool does are separate parts of the workflow. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
A model connected to browser tools
The side panel combines chat with page retrieval, highlighting, history search and tab actions. This is a source-build experiment, not an official Google or Gemma browser extension.
Why it stands out
Retrieve browser context before answering
The extension uses embeddings to find relevant page passages and indexed history entries. Those retrieval tools supply selected context rather than assuming the model has read every page you have visited.
Availability
Build and load it in Chrome
The documented path is to clone the repository, build with pnpm and load the dist directory as an unpacked extension. Model files are downloaded and cached during setup.
Why it matters
What makes it useful
For a question about the current webpage, ask_website retrieves relevant extracted passages rather than putting the entire page into an answer. Its results include element IDs that the highlighting tool can use to scroll to a passage. This connects an answer’s context back to something you can read on the page; it does not establish that the answer is correct.
What to know
Where it fits
Use find_history when you remember the subject of an indexed page but not its title. It searches stored titles, descriptions and URLs by similarity, and can narrow the results with a date range. This is a search over the extension’s own metadata index, not a complete text archive of every visited page.
Notable points
What stands out
The tab tools can open, switch and close tabs. Unlike retrieving a passage or history entry, these actions change the browser workspace. Keep that difference in mind when asking the agent to organise tabs: a plausible chat reply is not evidence that the intended tab was changed.
Before using
What to review
The documented requirements include Chrome 113 or later, WebGPU support and a capable GPU. Availability of WebGPU alone does not establish that the chosen model will run on your hardware.
The manifest requests sidePanel, storage, scripting and tabs permissions, with host access and content scripts across HTTP and HTTPS pages. Decide whether that page access fits the browser profile you intend to use.
History metadata and its vectors are kept in IndexedDB. Treat that stored browsing context as part of the extension’s footprint, not just the model download.
Reader fit
Who may find it relevant
Developers exploring how local inference, page retrieval and browser tools can share an extension.
People prepared to build and load an unpacked Chrome extension and check its actions against their tabs.
Someone seeking a ready-to-install Chrome Web Store assistant needs a different installation path from the one documented here.
Editorial note
Why LifeHubber lists it
Choose the search tool according to where you want the query to go. The google_search tool opens Google in a new tab with the query in its URL, even though model inference is local. If you do not want that query submitted to Google, avoid that tool; for a question about the current page or an already indexed page, the page or history retrieval tools address those narrower tasks. This distinction does not promise that the whole extension makes no network requests.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
More in AI Agents
Keep browsing this category
Explore more AI agent projects.
Paperclip
paperclipai/paperclip
A self-hosted server and dashboard for coordinating agent teams through companies, goals, roles, issues, heartbeats, budgets, approvals, and persistent activity records.
OpenSEO
every-app/open-seo
An SEO workspace with keyword, ranking, backlink, and site research tools, an MCP connection for AI agents, reusable research skills, and hosted or self-hosted access.
Magnitude
magnitudedev/magnitude
An early open-source local inference server that profiles a computer, recommends model configurations that fit, downloads the selected model, and connects it to existing agent harnesses through local OpenAI- and Anthropic-compatible APIs.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
