LLM Chats Notes

An on-device Chrome extension that turns AI chat conversations into structured, multilingual notes , which is fast, private, and cloud-free.

Comment

Inspiration

I believe AI is one of the most powerful tools for learning in the modern era.
When I use AI to study and explore ideas, the conversation often grows longer and more complex as I ask new questions and the AI responds with additional insights, suggestions, and examples.
While this dynamic exchange deepens my understanding, I often find it difficult to revisit and organize all the information afterward.
That challenge inspired me to create LLM to Notes, a tool that helps learners extract, summarize, and translate their AI conversations into structured notes for easier review and knowledge retention.
By turning long, unstructured discussions into clear study materials, users can focus on what they learned rather than searching through hundreds of messages.

Each person’s chat with AI represents a unique learning process. It captures how someone thinks, explores ideas, and builds understanding over time. Instead of relying solely on AI-generated PDFs or summaries, I wanted to give users a way to capture their own learning journey and form a personal knowledge base that reflects their thinking style.

I also realized that this personal archive of conversations could grow into a connected learning ecosystem. With NotebookLM, users can bring multiple conversations or summaries together as resources within a single notebook.
This allows them to generate audio overviews, ask new questions across all their previous discussions, and gradually build an interlinked knowledge system powered by both AI and their own insights.

As a bilingual learner, I also wanted to bridge the gap between my native language and English.
I sometimes study or ask questions in my native language but want to know the exact English terms and expressions that are used in technical contexts.
That motivated me to include translation controls, allowing users to pre-translate input or force English output for improved comprehension and vocabulary growth.

What it does

LLM to Notes is a Chrome MV3 extension built with WXT, React, and TypeScript.
It extracts conversations from ChatGPT and Gemini and generates comprehensive, multilingual study notes entirely on-device using Chrome’s Built-in AI APIs.

Key features include:

  • 🧩 One-click extraction with a three-layer fallback system
    (content script → injected DOM → plain text)
  • 🔄 Gemini auto-scroll that loads all messages in long conversations automatically
  • 🧠 Intelligent chunking for large inputs (up to ~50,000 characters), combining partial summaries into a complete final note
  • 📖 Detail level control with two options:
    • Detailed Study Notes: Preserves code, commands, and technical examples for in-depth study
    • Concise Summary: Provides a shorter overview for quick review
  • 🌐 Dual translation workflow using Translator and LanguageDetector APIs for mixed-language content
  • ⚡ Streaming summarization with real-time progress updates
  • 🔒 Privacy-first design: All AI processing happens locally; no API keys, network calls, or cloud dependencies

What I Learned

Working on this project taught me how to:

  • Implement Chrome’s on-device AI APIs such as LanguageModel, Summarizer, Translator, and LanguageDetector
  • Design multi-layer extraction and auto-scrolling systems for dynamic AI chat pages
  • Build an intelligent chunking and synthesis pipeline that efficiently handles large conversations
  • Develop detail-aware prompting to preserve code snippets, commands, and technical accuracy
  • Use generative AI as a development assistant to accelerate learning, debugging, and architectural refinement while maintaining full creative ownership

Challenges

Some of the main challenges I faced included:

  • Experimenting with new Chrome Built-in AI APIs
  • Ensuring consistent extraction across ChatGPT and Gemini’s different page structures
  • Balancing performance and accuracy during local inference for large inputs
  • Designing language detection and translation heuristics such as looksCJK and isMostlyCJK for accurate multilingual handling
  • Optimizing chunk processing and synthesis to support up to 50K-character conversations without losing context

Summary

LLM to Notes enables learners to transform their AI conversations into structured, multilingual notes that can be easily reviewed, translated, and expanded.
It encourages users to build their own AI-powered knowledge ecosystems by connecting multiple conversations and summaries through tools like NotebookLM, turning ongoing dialogue into lasting understanding.
Everything runs locally and securely, giving learners full ownership of their insights while combining AI intelligence with human learning.

Built With

Share this project:

Updates