ContextBridge – Local RAG Companion

What's New in Version 1.1.2 🚀
🆕 Highlight & Annotate Pages

You can now select any text on a web page, right-click, and choose "Highlight & Note" to save clippings and add personal notes.
A new Highlights tab in the side panel lets you easily browse, search, and manage your saved highlights.
🧠 Multi-Document Chat (Local RAG)

You can now switch the Chat context from "Current Page" to "All Indexed Pages" via the dropdown inside the Chat tab.
This uses advanced semantic retrieval to pull the most relevant snippets from across all your saved documents to answer your question.
🔍 Upgraded BM25 Search Engine

Upgraded the search algorithm to Okapi BM25 scoring. Matches are now ranked by relevance based on term frequency and document length, bringing the most relevant pages to the top of your search results.
⚙️ Custom Ollama Model Selection

Run custom local models with Ollama! You can now specify any model tag (e.g., llama3, mistral, phi3) directly inside the Settings tab.
💎 ContextBridge Pro Beta

Added a Pro Beta waitlist under the Stats tab. Sign up with your email to claim free lifetime access to upcoming features like multi-device sync, Obsidian Auto-Sync, and team libraries.
🔧 Bug Fixes & Refinements

Fixed a bug where switching chat providers (e.g., to Ollama) did not instantly refresh the Chat panel UI state.
Improved YouTube transcript extraction resilience to handle low-signal network conditions gracefully.
Cleaned up and optimized background workers to use less CPU and memory.

ContextBridge extracts clean, structured content from any web page and stores it locally in your browser using IndexedDB. No backend required — works offline out of the box.

Key Features:
• AI Chat — Ask questions about any indexed page using Claude, OpenAI, or Gemini (bring your own API key)
• Full-Text Search — Instantly find any page you've saved by keyword
• Markdown Export — Export pages as .md files or ZIP bundles for Obsidian/notes
• Smart Extraction — Site-adaptive parsers for GitHub, Stack Overflow, MDN, arXiv, and more
• Code Block Preservation — Language detection and fenced syntax maintained
• Auto-Tagging — Detects technologies (Python, Rust, React, SQL, etc.)
• Storage Mode — Choose local-only, endpoint-only, or both simultaneously
• Optional RAG Endpoint — Send content to your own vector store for advanced retrieval

Privacy:
• All data stored on-device only
• No analytics, no telemetry, no tracking
• AI chat calls go directly to your chosen provider — no intermediary
• API keys stored locally, never synced

How It Works:
1. Open any technical page (docs, GitHub, blog, tutorial)
2. Press Ctrl+Shift+E or click the sidebar button
3. Content is extracted, chunked, and stored locally
4. Search it, chat with it, or export it anytime

Requires: Your own API key for AI chat (Claude, OpenAI, or Gemini). The extension itself is completely free.