A RAG (Retrieval-Augmented Generation) system for DataCommit podcast episodes. Downloads audio from YouTube, transcribes with Whisper, and enables Q&A using Haystack, ChromaDB, and Gemini.

DataCommit is a Turkish podcast series where data science experts share their career journeys, technical knowledge, and industry experiences. 🎙️ Watch all episodes on YouTube

DataCommit Banner

• Audio Download: yt-dlp
• Speech-to-Text: Local Whisper-Turbo
• Audio Processing: FFmpeg, librosa, K-Means
• Text Correction: Gemini 2.5 Flash Agent

• Backend: Python, Flask
• RAG Framework: Haystack 2.22
• Vector Database: ChromaDB
• LLM: Google Gemini 3 Flash
• Embeddings: Sentence Transformers (all-MiniLM-L6-v2)
• Frontend: HTML, CSS, JavaScript

Preprocessing architecture

• Python 3.10+
• Google Gemini API Key

git clone https://github.com/enesmanan/DataCommit.git
cd DataCommit
python -m venv venv
venv\Scripts\activate # Windows
pip install -r requirements.txt

Create a .env file in the project root:

GEMINI_API_KEY=your_gemini_api_key_here

python create_database.py

This will:

• Load all episode transcripts from data/Final/
• Split them into chunks with metadata
• Create embeddings and store in ChromaDB

To rebuild database: delete chroma_db/ folder and run again.

python app.py

Open your browser at: http://localhost:5000

For audio preprocessing (YouTube to transcript), see /preprocessing

DataCommit/
├── app.py # Flask web server
├── rag_pipeline.py # RAG pipeline & Gemini integration
├── create_database.py # Vector database creation
├── data/ # Episode transcripts
├── chroma_db/ # Vector database (auto-generated)
├── static/ # Frontend assets (CSS, JS, images)
├── templates/ # HTML templates
└── preprocessing/ # Audio-to-text scripts

Enes Fehmi Manan

• 🔗 LinkedIn: linkedin.com/in/enesfehmimanan
• 🐙 GitHub: github.com/enesmanan
• 📧 Email: enesmanan768@gmail.com

Made with ❤️ for the Turkish Data Science Community