NoteWave is an AI-powered "Second Brain" and immersive research ecosystem. It transforms static PDF documents into dynamic knowledge graphs, interactive research studios, and voice-enabled learning environments.
Built with Next.js 16 (Turbopack), NoteWave utilizes high-integrity LLM orchestration, vector search, and real-time audio processing to redefine document interaction.
It is strongly recommended to run this application locally.
The publicly deployed version is intended as a portfolio demo. Because it relies on free-tier API keys (Groq & ElevenLabs), it may be rate-limited, become unavailable, or run out of credits without notice. Running it locally gives you full control and guarantees data privacy.
NoteWave is organized into specialized "Studios," each designed for a specific cognitive task:
- 📄 RAG-Powered Chat: Professional document analysis with semantic source citing using Pinecone and Groq (Llama 3.3 70B).
- 🎙️ Podcast Studio: Generates an engaging audio deep-dive conversation between AI hosts. Supports MP3 downloads and real-time script tracking.
- 🧠 Flashcard Studio: AI-driven concept extraction with a 3D flip UI and "Creator Mode" for manual additions.
- ⚡ Knowledge Graph: A 3D Force-Directed Graph that visualizes the relationship between concepts found in your research.
- ⚖️ Agentic Debate: A multi-persona research environment where Dr. Skeptic, The Weaver, and Veritas debate the core thesis of your files.
- 🛡️ Verified Vault: An integrity auditor that scans documents for bias, logical fallacies, and "hallucination" scores.
- 📝 Quiz Studio: Adaptive learning module that generates custom assessments and provides detailed mastery reports.
- 🎤 Voice Immersion: A hands-free mode using Deepgram Nova-2 for speech-to-text and OpenAI TTS-1 for neural voice replies.
- Adaptive Settings Studio: Houses a Bio-Adaptive Profile that tracks cognitive load and suggests learning styles (Kinesthetic, Visual, etc.).
- Focus Mode: A UI transformation that dims background elements and simplifies the workspace for deep work.
- Command Orchestration: A global
/command palette with full keyboard navigation for instant studio switching. - Local Persistence: All session settings and document metadata are synced to
localStoragefor privacy-first continuity.
The interface features a centered chat with resizable sidebars for source management and learning studios.
A professional directory that guides users through the available AI specialized tools.
- Core: Next.js 16, TypeScript, Tailwind CSS, Shadcn/UI
- Intelligence: Llama 3.3 70B via Groq
- Vector Engine: Pinecone (Serverless)
- Embeddings: Hugging Face (
all-MiniLM-L6-v2) - Voice STT: Deepgram Nova-2 (Sub-300ms latency)
- Voice TTS: ElevenLabs
- Spatial UI:
react-force-graph-3d& Three.js
- Ingestion Pipeline: PDF text is semantically split, vectorized, and stored in Pinecone with strict filename filtering.
- Hardened Chat Logic: The AI assistant uses a high-integrity system prompt to prevent hallucinations and maintain academic rigor.
- Flashcard Generation: AI identifies key concepts and definitions from context to generate interactive study decks.
- Command Orchestration: A custom
/command palette with keyboard navigation (Up/Down/Enter) allows users to swap studios instantly.
- Streaming Synchronization: Solved "text stuttering" in streaming responses by implementing an accumulator pattern and
TextDecoderstream handling. - Layout Fluidity: Engineered a resizable sidebar system that maintains perfectly straight header margins across three columns.
- Conversational Awareness: Fine-tuned the system prompt to handle professional greetings (e.g., "Hello") while maintaining strict document-only research integrity.
-
Clone & Install:
git clone [https://github.com/samarthsaxena2004/note-wave.git](https://github.com/samarthsaxena2004/note-wave.git) cd note-wave npm install -
Environment Variables (
.env.local):GROQ_API_KEY=your_key PINECONE_API_KEY=your_key PINECONE_INDEX_NAME=note-wave HUGGINGFACE_API_KEY=your_key DEEPGRAM_API_KEY=your_key OPENAI_API_KEY=your_key ELEVENLABS_API_KEY=your_key
-
Install Radix Primitives (If missing):
npx shadcn-ui@latest add switch progress dialog badge
-
Run:
npm run dev
Open http://localhost:3000 in your browser.
Troubleshooting: If the application fails due to a retired model or deprecated API call, please refer to the comprehensive debugging steps in the
MAINTENANCE_GUIDE.mdfile in the repository root.
- Phase 3 (Personalization): Quiz Studio, Mastery Tracking, and Settings Studio.
- Phase 4 (Immersion): Functional Voice-to-Concept (V2C) and Neural TTS.
- Phase 5 (Synthesis): Recursive multi-document summaries and Insight Export (PDF/Markdown).
- Phase 6 (Sovereignty): Cloud Sync via Supabase and user authentication.
- Phase 7 (Neural Performance): Edge deployment and local LLM fallback (Ollama).
Contributions are welcome! If you have ideas for features (e.g., persistent database storage, user authentication), feel free to fork the repo and submit a PR.
- Fork the Project
- Create your Feature Branch (
git checkout -b feature/AmazingFeature) - Commit your Changes (
git commit -m 'Add some AmazingFeature') - Push to the Branch (
git push origin feature/AmazingFeature) - Open a Pull Request
- Project Blueprint: Inspired by Google NotebookLM's core research and application design.
- Architectural Guidance: Strategic design, debugging, and quality assurance provided by the Gemini 2.5 Flash AI assistant.