I’m building a small personal automation system that captures YouTube and Instagram videos from my phone and processes them on a Mac mini. The system will: 1. Receive a video URL from an iPhone Shortcut 2. XXXXXX 3. Retrieve or generate a transcript 4. Generate a short summary 5. Store everything in Dropbox using a clean file structure The goal is a simple, robust pipeline, not a large framework. ⸻ Architecture Input iPhone Shortcut → POST URL to webhook Processing machine: Mac mini running a small Python service. Storage: Dropbox folder. Output files (for each task): yyyymmdd-title.xxx yyyymmdd-title.txt (transcript) yyyymmdd-title.json (metadata + summary) Components to implement 1. Webhook API Small HTTP endpoint to receive URLs from the shortcut. 2. Job Queue Simple file-based queue with human-readable JSON files. [ job_id, url, source yt/IG, status, created, started, completed, title, duration,xxxxxx, transcript_file, summary_file, error] 3. Video Must support: YouTube Instagram reels / posts Responsibilities: • resolve canonical video URL • retrieve metadata • detect subtitles if available 4. Transcript system a. Use YouTube captions if available b. Otherwise transcribe audio Transcription can use paid API (eg OpenAI Whisper) 5. Summary generation Use LLM API to generate: • short summary • names appearing • names mentioned • topic tags 6. Dropbox storage Files written to one flat folder: • xxxx • 20260313-video-title.txt • 20260313-video-title.md 7. Worker process Background worker that: • scans queue • processes queued jobs • updates status • logs progress Crash-safe behavior is important.