tldw Assistant

Browser extension frontend for tldw_server — a unified AI assistant with chat, RAG, media processing, and more.

Overview

tldw Assistant is an open‑source browser extension that provides a side panel and full‑page web UI for your own tldw_server instance. It connects to tldw_server (an API aggregator for multiple LLM providers) so you can:

• Chat with any model configured on your server
• Search and cite with RAG (retrieval‑augmented generation)
• Ingest and process media (web pages, videos, audio, documents)
• Transcribe speech (STT) and synthesize speech (TTS)
• Chat with the current page, use internet search, OCR snippets, and more

This repo refactors the original Page Assist extension into a dedicated, whitelabeled frontend for tldw_server.

Requirements

• Bun (or Node) for building: https://bun.sh/
• A running tldw_server instance (local or remote)
  • Single‑user: API key
  • Multi‑user: username/password (Bearer tokens)

Quick Start (Development)

bun install

# Chrome/Edge dev
bun dev             # Chrome
bun run dev:edge    # Edge

# Firefox dev
bun run dev:firefox

Then load the extension from the WXT dev server prompt, or open your browser’s extensions page and load the unpacked output from the prompted build directory.

Build & Package

# Build all targets (Chrome, Firefox, Edge)
bun run build

# Or build individually
bun run build:chrome
bun run build:firefox
bun run build:edge

# Create zipped artifacts for release
bun run zip          # Chrome by default
bun run zip:firefox  # Firefox

By default the build output is placed in build/. Load that directory as an “unpacked”/temporary extension in your browser.

Configuration (First Run)

Open Options → tldw Server and configure:

• Server URL: e.g., http://localhost:8000
• Authentication Mode:
  • Single‑user (API key)
  • Multi‑user (login via username/password)
• Timeouts: global and per‑API (chat, RAG, media, uploads)

The extension requests optional host permission (Chromium) for your configured origin so background requests can include auth headers and avoid CORS issues.

Features

• Sidebar: Chat from any page; quick RAG/search; page‑aware chat
• Web UI: Full chat experience with history, editing, and regeneration
• RAG: Simple/search modes; insert citations into context
• Media: Add URLs, ingest web content; progress via notifications
• STT/TTS: Transcribe uploads and play synthesized speech (where available)
• Knowledge Base: Load files/notes and chat with your data
• Internet Search: Integrations for web search providers
• OCR: Basic OCR for selections/screenshots
• Multi‑language UI: Locales under src/assets/locale/* and _locales/*

Want something else? Please open an issue.

Usage

Open the UI

• Side Panel: Ctrl+Shift+Y
• Web UI (new tab): Ctrl+Shift+L

Shortcuts are configurable from your browser’s extension settings and inside the app for in‑app actions.

In‑App Shortcuts (defaults)

• New Chat: Ctrl+Shift+O
• Toggle Sidebar: Ctrl+B
• Focus Textarea: Shift+Esc
• Toggle Chat Mode (page/chat): Ctrl+E

Browser Support

Browser	Sidebar	Chat With Webpage	Web UI
Chrome	✅	✅	✅
Brave	✅	✅	✅
Firefox	✅	✅	✅
Vivaldi	✅	✅	✅
Edge	✅	✅	✅
LibreWolf	✅	✅	✅
Zen Browser	✅	✅	✅
Opera	❌	❌	✅
Arc	❌	❌	✅

Model & Provider Support

Models are surfaced from your tldw_server configuration (OpenAI‑compatible providers, local runtimes, etc.). Model fetching uses /api/v1/llm/models and related endpoints exposed by your server.

Roadmap (Active Work)

• ✅ Foundation: branding, settings, auth (API key + login)
• ✅ Models: fetch and select models from tldw_server
• 🚧 Chat: streaming completions via /api/v1/chat/completions
• ⏳ RAG search and citations
• ⏳ Media ingestion (URL/page) and processing
• ⏳ STT/TTS integration

Privacy

• The extension does not collect analytics or telemetry.
• Credentials are stored using browser storage; tokens are handled by the background where possible.
• Data you process flows to the tldw_server you configure (local or remote). Review your server’s privacy/security settings.
• See PRIVACY.md for more details.

Development Notes

• Source lives in src/ with WXT entries under entries/ and entries-firefox/.
• TailwindCSS for UI (src/assets/tailwind.css, tailwind.config.js).
• Prettier + import sorting: bunx prettier --write .
• Type‑check before PRs: bun run compile
  • OpenAPI path enforcement: the web UI’s API calls are compile‑time checked against the bundled openapi.json. If you add or change an endpoint path/method, update openapi.json accordingly or your typecheck will fail.
  • CI: GitHub Actions runs the typecheck on each push/PR (.github/workflows/typecheck.yml).
  • Use the typed helpers bgRequest, bgStream, and bgUpload for all server calls. Direct browser.runtime.sendMessage({ type: 'tldw:request' ... }) calls should pass a path typed as AllowedPath to participate in checks.

Contributing

Contributions are welcome! Please open an issue or PR. Follow conventional commits (feat:, fix:, docs:, chore:, etc.) and include steps to test and screenshots for UI changes.

License

MIT

Acknowledgements

This project builds on the excellent work of the original Page Assist extension and community.