🧠 Inspiration

While researching and studying online, I constantly found myself juggling multiple PDFs, academic articles, and notes — all open in different tabs, scattered and disorganized. Copying citations, rewriting complex passages, and keeping track of relevant information felt exhausting. ClarityNote was born from that frustration: a single space where you can collect, simplify, translate, and align ideas — all on your device, without ever sending your data to the cloud.

⚙️ What It Does

ClarityNote is an on-device Chrome AI web app that turns reading and note-taking into a structured, intelligent workflow. Users can:

Highlight or extract text from PDFs and add it to a side panel of smart note cards.

Simplify or translate notes using Chrome’s built-in Summarizer, Translator, Writer, and Rewriter APIs.

Align selected notes with their main draft using LanguageModel and Prompt APIs, generating clear insights like supports, contradicts, or neutral.

Choose writing modes — Essay, Research Paper, or Presentation — to guide the structure and tone of AI rewriting.

Export everything into a clean TXT or DOCX file. All processing happens locally through Gemini Nano, ensuring total privacy.

🛠️ How We Built It

Frontend: React + TailwindCSS

Storage: Firestore / Local Storage for user persistence

AI Integration: Chrome’s built-in APIs — Prompt / LanguageModel, Summarizer, Translator, Writer, Rewriter

Design: Minimal three-column layout (Notes | Editor | Imports) for focus and usability

Browser: Chrome Canary (v143+) with on-device Gemini Nano enabled

The AI runs entirely in the page world, and the extension bridge allows direct communication with the built-in APIs without external servers.

🚧 Challenges We Ran Into

Audio transcription: Multimodal API access for audio varied between Chrome builds, making it unreliable for demo use.

On-device model setup: Getting Gemini Nano to download and initialize consistently required fine-tuning Canary flags and origin trials.

Prompt API permissions: Content script isolation made connecting page-world AI to the UI tricky — solved through a custom postMessage bridge.

Time constraints: Building a full academic workflow within two weeks required strict prioritization.

🏆 Accomplishments That We’re Proud Of

Built a complete on-device AI workflow (Summarize → Rewrite → Translate → Align) with zero cloud calls.

Designed a clean, academic-friendly interface that feels familiar and practical.

Implemented a working alignment engine that checks each note’s relevance to the thesis — with Accept/Reject suggestions.

Created a system that genuinely improves clarity and organization for researchers and students.

📚 What We Learned

How to fully utilize Chrome’s Prompt API and LanguageModel features in the page world.

Techniques for handling on-device model availability and progress events.

UX lessons: users value simplicity and structure more than raw features.

The future of AI-assisted learning lies in private, local-first tools — not cloud chatbots.

🔮 What’s Next for ClarityNote

Enable working audio transcription and in-browser recording using the multimodal LanguageModel API.

Expand academic templates with Dissertation, Case Study, and Technical Report modes.

Add citation auto-formatting (APA, MLA, Harvard) and drag-and-drop bibliography export.

Integrate collaborative workspaces for students to co-edit and align research in real time

Built With

  • chrome
  • css-frameworks-&-libraries:-react-18
  • css3
  • gemini-nano
  • git
  • github
  • html
  • javascript
  • localstorage
  • promptapi
  • react
  • rewriteapi
  • summerizerapi
  • tailwind
  • tailwindcss-browser-platform:-google-chrome-(canary-v143+)-with-gemini-nano-enabled-ai-apis:-prompt-/-languagemodel-api-summarizer-api-translator-api-writer-api-rewriter-api-storage-&-hosting:-firestore-+-firebase-hosting-(for-demo)
  • translatorapi
  • vite
  • writeapi
Share this project:

Updates