NotebookLM-Lite is a note-taking application inspired by Google's NotebookLM, aiming to fully replicate features such as document Q&A, podcast generation, and presentation creation. It is currently under development. For a simple Chinese podcast generation feature, please refer to the v0.1 branch.
The demo covers configuring model profiles, adding a source, asking grounded RAG questions with citations, generating interactive Mind Map, Flashcards/Quiz, Data Table, Podcast, and Video Overview placeholder artifacts. Model waiting time is shortened for README pacing.
- Unified LLM Interface: Text generation, embeddings, rerank, and speech profiles are managed through LiteLLM/OpenAI-compatible settings.
- Neural Audio Synthesis: High-quality, multi-speaker podcast generation using CosyVoice (via Dashscope).
- Document Intelligence: Docling parses source files, Chonkie chunks text, and SeekDB stores sources/chunks for hybrid RAG retrieval with citations.
- Interactive Player: Integrated audio player with playback controls and download capability.
- Frontend: React, Vite, TailwindCSS, Framer Motion (animations)
- Backend: Python, FastAPI, Uvicorn
- AI/ML: OpenAI API, Anthropic API, Dashscope (TTS)
- Python 3.10+
- Node.js 18+
- FFmpeg (Must be installed and added to system PATH)
Navigate to the project root:
# Install Python dependencies
pip install -r requirements.txt
# Create your local configuration file
cp config_example.yaml config.yamlEdit config.yaml and set the model credentials you want to use. Text,
embedding, optional rerank, and speech models are configured separately under
api.models. The default text path is LiteLLM/OpenAI-compatible, and
Anthropic or Gemini/GenAI-style model strings can be used through
LiteLLM-compatible configuration. Keep config.yaml local; it is ignored by
git and must not be committed with real keys.
# Start the API server
python backend/main.pyThe backend will run at http://localhost:8000.
Open a new terminal and navigate to the frontend directory:
cd frontend
# Install Node dependencies
npm install
# Start the development server
npm run devThe UI will be available at http://localhost:5173.
- Launch: Open
http://localhost:5173in your browser. - Configure:
- Configure the text model, embedding model, optional rerank model, and optional speech model in
config.yaml. - The Settings button also loads local defaults and can refresh runtime model profiles without exposing configured keys back to the browser.
- Use the text model thinking toggle when the selected backend supports it.
- Speech/TTS is optional for text artifact generation; podcast scripts can still be generated without audio credentials.
- Configure the text model, embedding model, optional rerank model, and optional speech model in
- Upload: Drag & drop a file or paste text into Sources. New uploads use Chonkie chunking and configured embeddings when available.
- Ask and Generate: Select sources, ask RAG questions with citations, or generate Studio text artifacts and podcast scripts.
- Listen: Use the built-in player to listen to your podcast or download it.
Apache 2.0