abaja/notes-taker/incoming_audio - Storage Bucket

🎓 Autonomous Academic Lecture & Notes Assistant

Production-ready, local-first & cloud-deployable AI study copilot for graduate-level STEM coursework.

---

📌 Overview

The Autonomous Academic Lecture Assistant automates the entire lifecycle of lecture ingestion, transcription, structured academic synthesis, LaTeX mathematical derivation, exam preparation, and cross-device synchronization:

               ┌──────────────────────────────────────────────┐
               │    Audio Recording / PDF / Slides / Word     │
               └──────────────────────┬───────────────────────┘
                                      │
           ┌──────────────────────────┼──────────────────────────┐
           ▼                          ▼                          ▼
┌─────────────────────┐    ┌─────────────────────┐    ┌─────────────────────┐
│ 📁 ./incoming_audio │    │ 🤖 Telegram Bot     │    │ 💻 Streamlit Web UI │
│ (Watcher Daemon)    │    │ (@note_taker_bot)   │    │ (Direct Upload)     │
└──────────┬──────────┘    └──────────┬──────────┘    └──────────┬──────────┘
           │                          │                          │
           └──────────────────────────┼──────────────────────────┘
                                      ▼
                      ┌──────────────────────────────┐
                      │    Gemini Flash Engine       │
                      │ (Auto-fallback Architecture) │
                      └──────────────┬───────────────┘
                                     │ Generates structured Markdown & LaTeX
                                     ▼
                      ┌──────────────────────────────┐
                      │   📚 Obsidian Vault Vault    │
                      │      (./lectures/*.md)       │
                      └──────────────┬───────────────┘
                                     │
           ┌─────────────────────────┴─────────────────────────┐
           ▼                                                   ▼
┌─────────────────────────────┐             ┌─────────────────────────────────────┐
│    Obsidian on Your Mac     │             │        GitHub Vault Repository      │
│  (Auto-sync via Git Engine) │             │  (my-obsidian-notes.git)            │
└─────────────────────────────┘             └─────────────────────────────────────┘

---

✨ Key Features

• Multi-Format Ingestion: Supports .m4a, .mp3, .wav, .aac, .ogg, .flac, .pdf, .docx, .doc, .pptx, .ppt, .txt, .md, and Live Browser Mic.
• Tiered Model Routing: Routes audio and dense math proofs to gemini-3.6-flash and typed slides/docs to ultra-fast gemini-3.1-flash-lite.
• Obsidian Native Integration: Automatically attaches YAML frontmatter, [[Wikilinks]], #course/... tags, and > [!WARNING] callouts.
• Obsidian Maps of Content (MOC): Auto-compiles chronological syllabus tables, master theorem indexes, and exam pitfalls.
• Anki Flashcard Deck Exporter: Auto-compiles native .apkg flashcard decks for spaced repetition.
• Semester-Wide ChromaDB Vector Search: Semantic RAG queries across all past course notes and proofs.
• Automated Git Sync Engine: Commits and pushes notes to your remote GitHub vault automatically.
• Streamlit Web Dashboard:
  • Live in-browser microphone recording.
  • Multi-subject daily briefings.
  • Date-filtered syllabus exam tutor & semester semantic search.
• Asynchronous Telegram Bot:
  • Send audio/PDFs on-the-go.
  • Interactive /menu button interface.
  • Export Anki decks directly to mobile via /anki.
  • Semantic vector search via /search.
  • High-res LaTeX math image rendering via /latex.

---

📁 Repository Structure

.
├── .github/workflows/         # CI/CD auto-sync workflow for Hugging Face Spaces
├── .env.example               # Environment variables configuration template
├── .gitignore                 # Standard Python, Obsidian, and IDE ignore rules
├── Dockerfile                 # Container image specification (UID 1000, Port 7860)
├── LICENSE                    # MIT License
├── README.md                  # Project documentation with HF Spaces metadata
├── anki_exporter.py           # Anki .apkg deck generation engine
├── app.py                     # Streamlit web dashboard with mic & password gate
├── audio_optimizer.py         # 32kbps mono AAC audio compressor
├── bot.py                     # Telegram bot service
├── core_engine.py             # Shared reasoning, date filtering, and LaTeX repair
├── docker-compose.yml         # Container orchestration configuration
├── git_sync.py                # Automated Git synchronization engine
├── incoming_audio/            # Watched folder for recordings/documents
├── ingest_audio.py            # Universal audio and document ingestion pipeline
├── lectures/                  # Generated structured Markdown study notes
├── main.py                    # Unified launcher with startup vault hydration & GC
├── metadata_db.py             # Offline SQLite metadata database
├── requirements.txt           # Pinned project dependencies
├── vector_store.py            # ChromaDB vector store & semantic search
└── watcher.py                 # File system watcher daemon

---

🛠️ Quick Start (Local Setup)

1. Clone & Setup Virtual Environment

git clone https://github.com/AkhilBaja3005/notes-taker.git
cd notes-taker

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

2. Configure .env

Copy .env.example to .env and fill in your credentials:

cp .env.example .env

GEMINI_API_KEY=your_gemini_api_key_here
AUDIO_MODEL=gemini-3.6-flash
DOC_MODEL=gemini-3.1-flash-lite
TELEGRAM_BOT_TOKEN=your_telegram_bot_token_here
ALLOWED_TELEGRAM_USER_IDS=
WATCH_DIR=./incoming_audio
LECTURES_DIR=./lectures

# Obsidian Git Sync Configuration
ENABLE_GIT_SYNC=true
GIT_VAULT_REPO_URL=https://<GITHUB_PAT>@github.com/yourusername/my-obsidian-notes.git
GIT_BRANCH=main
STREAMLIT_PASSWORD=

3. Launch All Services

python main.py

---

🤗 Hugging Face Spaces Deployment

Space URL: https://huggingface.co/spaces/abaja/notes-taker

In your Hugging Face Space settings, add the following secrets:

• GEMINI_API_KEY: Your Gemini API Key
• TELEGRAM_BOT_TOKEN: Your Telegram Bot Token (from @BotFather)
• GIT_VAULT_REPO_URL: https://<GITHUB_PAT>@github.com/yourusername/my-obsidian-notes.git
• ENABLE_GIT_SYNC: true
• STREAMLIT_PASSWORD: (Optional password gate)

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.