Side Eye

Chrome extension · On-device & local models

Privately Chat, Research, and Command your local LLM in the Chrome Sidebar.

Side Eye lives in Chrome’s side panel — beside the tab you’re on. Chat with LiteRT-LM (Gemma 4 on-device via WebGPU), or stream from LM Studio or Ollama on your machine. Optional web search and a local bookmark index stay under your control. Your chats never go to our servers.

Add to Chrome See what it does

Features

Built for browsing with a local LLM

On-device in the browser by default — or point Side Eye at models you already run on your machine. Optional web search and a bookmark index when you want them.

On-device Gemma 4

Chat with Gemma 4 (E2B or E4B) via LiteRT-LM and WebGPU — no local server required. Download the model once; inference stays in Chrome.

Side panel chat

Open Side Eye beside any tab. Multiple chat tabs, streaming markdown replies, and a profile picker for LiteRT-LM, LM Studio, or Ollama.

Page context

Toggle the eye to include URL, selection, and readable page text — only when your message needs it. Ignore sensitive domains entirely.

Web search

Opt in to look up the web before chatting. DuckDuckGo, Exa, or Firecrawl — replies still come from your local model, with source links under the answer.

Bookmark index

Build a local index of summaries, tags, and categories. Browse or search saved pages, then ask about them — citations show as From your bookmarks.

LM Studio & Ollama

Prefer your own stack? Connect LM Studio, Ollama, or another OpenAI-compatible local server for vision models and larger contexts — still on your machine, not our cloud.

Vision & images

Attach up to eight images per message with capable models through LM Studio or Ollama. Image attach is disabled for text-only LiteRT-LM.

Quick actions

Summarize this page, or use Related and Reviews. With web search on, those two look up the live web instead of only the current tab.

On-device voice

Optional read-aloud with Supertonic TTS. Models download once; synthesis runs in the browser with a reactive orb control.

Search & bookmarks

Look up the web. Remember what you saved.

Both are optional. Search is off until you enable it. The bookmark index lives on this device. Your model still writes the reply.

Options → Web search

Web search

Side Eye can query the web before it talks to your model. Use it for facts, “search the web” questions, and Related / Reviews when search is on.

  • DuckDuckGo — no API key
  • Exa or Firecrawl — key stored locally, not synced
  • Sources — titles and URLs under the reply
  • Same provider is used to gather pages when you index bookmarks

Bookmarks manager tab

Bookmark indexer

Open Chrome’s bookmarks manager and the side panel becomes a bookmarks workspace: Browse by category, search titles and summaries, and see Indexed vs Pending.

  • Start sync for a date range (default last 7 days)
  • Each bookmark is gathered, then summarized and tagged on-device
  • Index stays in IndexedDB — Clear index does not delete Chrome bookmarks
  • Ask about saved pages; matching hits can be cited in chat

How it works

Four steps to a side glance

  1. 1

    Install the extension

    Add Side Eye from the Chrome Web Store and open the side panel from the toolbar. Chrome with WebGPU is recommended for on-device chat.

  2. 2

    Load a model

    LiteRT-LM is on by default — in Options, click Load model once to download Gemma 4. Or connect LM Studio / Ollama (enable CORS / OLLAMA_ORIGINS) and pick that profile in the footer.

  3. 3

    Chat with context

    Turn on page context when you want the current tab included. Ask naturally — “summarize this page”, “explain this selection”, or attach images with LM Studio / Ollama vision models.

  4. 4

    Search the web or index bookmarks

    In Options, enable Web search if you want live lookups and source citations. Open Chrome’s bookmarks manager in the linked tab, pick a date range, and Start sync to build the local index (web search must be on for sync).

Privacy

Your data stays yours

  • Chat history lives in chrome.storage.local on your device.
  • With LiteRT-LM, prompts and replies stay in the browser (WebGPU). With LM Studio or Ollama, page and chat content go only to the server URL you configure — usually localhost.
  • Ignored domains are never read — banking, mail, or anything you block in settings.
  • Web search is off by default. When you enable it, queries go to DuckDuckGo, Exa, or Firecrawl — not to us. Chat completions still go only to your local / on-device profile.
  • The bookmark index (summaries, tags, categories) stays in IndexedDB on this device. During sync, titles, URLs, and gathered text may be sent to your search provider and analysis profile.
  • Gemma 4 (LiteRT-LM) and optional Supertonic TTS models download from Hugging Face once; inference and speech run on-device.

Install

Get Side Eye in Chrome

Side Eye requires Chrome 114+ for the Side Panel API. Install from the Chrome Web Store or read the docs on GitHub.

Docs on GitHub

Feature overview, LiteRT-LM / Gemma 4, LM Studio and Ollama setup, web search, bookmark sync, voice options, and troubleshooting — all in the public README.

  1. Open the side-eye-docs repository on GitHub.
  2. Read setup notes for LiteRT-LM, LM Studio CORS, Ollama OLLAMA_ORIGINS, web search, and bookmarks.
  3. Follow the Chrome Web Store steps on the right to install the extension.
View docs on GitHub

Chrome Web Store

The fastest way to install — no build step, no developer mode. Free from the Chrome Web Store.

  1. Click Add to Chrome on the Chrome Web Store listing.
  2. Confirm the install when Chrome prompts you.
  3. Pin Side Eye from the extensions menu, then load LiteRT-LM (or connect LM Studio / Ollama) in Options.
Add to Chrome — Free