TabGist

Summarize any web page in one click — powered by LLM, streamed in a new tab.

한국어

TabGist is a Chrome extension (Manifest V3) that extracts the current tab's content and generates a concise summary using an LLM of your choice. Click the toolbar icon, and a new tab opens with the summary streamed in real time.

Features

• One-click summarization — Click the toolbar icon on any web page
• Real-time streaming — Summary appears token by token as the LLM generates it
• Dual provider support
  • Anthropic — Claude models via the Messages API
  • OpenAI-compatible — OpenAI, LM Studio, Ollama, llama.cpp, OpenRouter, Groq, vLLM, and any /v1/chat/completions endpoint
• 7 built-in presets — One-click configuration for popular providers
• 8 output languages — Auto (follows page language), Korean, English, Japanese, Chinese, Spanish, French, German
• Dark mode — Follows your system preference
• Connection test — Verify your API settings before use
• Thinking timer — Real-time elapsed timer shown while waiting for the first token
• Stats display — Thinking time, total time, and tokens/second shown after each summary
• Truncation warning — Alert when output is cut off due to max_tokens limit
• Custom system prompt — Override the default prompt to customize output format
• max_completion_tokens option — Support for models that require max_completion_tokens instead of max_tokens
• Action buttons — Regenerate, copy, open original, close

Installation

1. Clone this repo
2. Open chrome://extensions in Chrome
3. Enable Developer mode (top right)
4. Click Load unpacked and select the TabGist folder
5. Click the TabGist icon in the toolbar to get started

Setup

1. Right-click the TabGist icon → Options (or go to chrome://extensions → TabGist → Details → Extension options)
2. Select a Provider (Anthropic or OpenAI-compatible)
3. Enter your API Key (leave empty for local servers like LM Studio / Ollama)
4. Adjust Base URL, Model, Max Tokens, and System Prompt as needed — or use the preset dropdown
5. (Optional) Enable Use max_completion_tokens for models that require it (e.g. o1, o3, gpt-5.4-nono)
6. Click Test Connection to verify
7. Click Save

Usage

1. Navigate to any web page (article, blog, wiki, etc.)
2. Click the TabGist toolbar icon
3. A new tab opens showing:
   • Analyzing page... while content is extracted
   • Generating summary... while waiting for the first token
   • Streaming summary with a typing cursor
   • Real-time timer while waiting for the first token
   • Stats at the bottom when complete
   • Warning if output was truncated
4. Use the action buttons to regenerate, copy, open the original, or close

Supported Presets

Preset	Base URL	Default Model
Anthropic	https://api.anthropic.com	claude-sonnet-4-6
OpenAI	https://api.openai.com/v1	gpt-4o-mini
Z.AI	https://api.z.ai/api/paas/v4	glm-5.1
LM Studio	http://localhost:1234/v1	—
llama.cpp server	http://localhost:8080/v1	—
Ollama	http://localhost:11434/v1	—
OpenRouter	https://openrouter.ai/api/v1	—

Language Settings

Setting	Behavior
Auto	Summarize in the same language as the source page
Specific language	Translate and summarize in the chosen language regardless of source

Project Structure

TabGist/
├── manifest.json           # Manifest V3 config
├── background.js           # Service worker (orchestration)
├── content.js              # Readability-based content extraction
├── providers/
│   ├── index.js            # Common interface + language-aware prompts
│   ├── anthropic.js        # Anthropic Messages API + SSE
│   └── openai.js           # OpenAI-compatible /chat/completions + SSE
├── lib/
│   └── Readability.js      # Mozilla Readability (vendored)
├── summary/
│   ├── summary.html        # Summary result page
│   ├── summary.js          # Streaming renderer + stats
│   └── summary.css         # Styles with dark mode
├── options/
│   ├── options.html        # Settings page
│   ├── options.js          # Settings logic + presets + test
│   └── options.css         # Settings styles
└── icons/                  # 16/48/128 PNG icons

Tech Stack

• Vanilla JavaScript (no build step, no framework)
• Mozilla Readability.js for content extraction
• CSS with custom properties (dark mode via prefers-color-scheme)
• Chrome Extension Manifest V3

Notes

• chrome://, file://, and other restricted pages cannot be summarized
• Pages with very little text content (< 50 characters) will show an error
• API keys are stored in chrome.storage.local; other settings in chrome.storage.sync
• Max Tokens default is 2048 — increase in settings if summaries are truncated
• A truncation warning will appear automatically when the output reaches the token limit

License

MIT