PodPulse

Local-first podcast intelligence engine for macOS Apple Silicon.

Monorepo layout

• backend/ FastAPI + pipeline + SQLite
• frontend/ React + Vite + TypeScript UI
• data/ local storage (DB, audio, logs, style profiles)

Prerequisites

• Python 3.11+
• Node.js 20+
• ffmpeg
• yt-dlp
• Optional: ollama, mlx-whisper (fallback only)

Backend setup

# if needed on macOS:
# brew install python@3.11

cd /Users/simon/Documents/PulsePod/backend
python3.11 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cd /Users/simon/Documents/PulsePod
cp .env.example .env
python -m podpulse.migrations
python -m podpulse.scripts.seed_feeds --file docs/feed_seed_vc_ai.json
python -m podpulse.main

Frontend setup

cd /Users/simon/Documents/PulsePod/frontend
npm install
npm run dev

Standard commands

• Backend run: python -m podpulse.main
• Frontend dev: npm run dev
• DB migrate/init: python -m podpulse.migrations
• Seed default podcast list: python -m podpulse.scripts.seed_feeds --file docs/feed_seed_vc_ai.json
• Tests: cd backend && pytest -q
• Pipeline run once: python -m podpulse.workers.pipeline

Notes

• API docs: http://127.0.0.1:8000/docs
• Frontend expects VITE_API_BASE_URL (defaults to http://127.0.0.1:8000)
• Transcription defaults to parakeet-mlx (local Apple Silicon ASR) on Python 3.11+.
• Override Parakeet model with PARAKEET_MODEL in .env if you want a different local ASR model.
• Long episodes are chunked for Parakeet by default (PARAKEET_CHUNK_DURATION_SECONDS, PARAKEET_OVERLAP_SECONDS) to avoid Metal memory errors.