Detect ad segments in podcast audio using Whisper-powered transcription and an any open OpenAPI-compliant LLM backend .
AdSleuth transcribes podcast audio (local Whisper or remote OpenAI Whisper API), analyzes audio + transcript to detect ad segments (timestamps + confidence), and exposes OpenAPI-compatible endpoints for easy integration.
- Transcription via Whisper (local model or OpenAI Whisper API)
- Ad segment detection (timestamps, type, confidence)
- Backend: Flask
- Transcription: Whisper (local or remote)
- Optional: Docker / Docker Dev Container
jdbc:sqlite:file:\wsl$\Ubuntu-22.04\home\mateus\WSL\PROJETOS\local-IA-stack\whisper\podly_pure_podcasts_fork\src\instance\sqlite3.db?nolock=1
TODOs - TESTES - Check if the post is already processed in the database na rota process - os.environ["OLLAMA_API_BASE"] remover ? - Ajustar logs LiteLLM - Remover texto das Transcrições da DB.
python src/main.py flask --app ./src/main.py db init flask --app ./src/main.py db migrate -m "change" flask --app ./src/main.py db upgrade
QUANDO CRIAR A BASE DE DADOS PELO DEV CONTAINER ELE CRIA COM ROOT TEM Q AJUSTAR A PERMISSÃO DO ARQUIVO PARA EDITAR NO DBEAVER sudo chown mateus:mateus /home/mateus/WSL/PROJETOS/whisper/src/instance/sqlite3.db chmod 664 /home/mateus/WSL/PROJETOS/whisper/src/instance/sqlite3.db