Index every video into a queryable SQLite database

Steps

1. Install peepshow

   npm install -g peepshow

2. Run with the SQLite sink

   Schema auto-creates on first run.

   peepshow ./demo.mp4 --sink sqlite

3. Query the archive

   sqlite3 ~/.peepshow/sinks/sqlite/peepshow.db 'SELECT * FROM frames LIMIT 5'

4. Expose with Datasette

   Optional but ridiculous-fun — instant searchable UI.

   pip install datasette
   datasette ~/.peepshow/sinks/sqlite/peepshow.db

5. Auto-sink for every run

   peepshow sinks add sqlite

Why it works

SQLite is the perfect long-term archive — one file, zero ops, queryable by every tool that speaks SQL. peepshow's SQLite sink builds a normalised schema (`runs`, `frames`, `transcripts`, `tags`, `sinks`) so you can answer 'which clips contain the word X' or 'show me every frame with motion above N' in plain SQL.

When it helps

• Long-term personal video archives — searchable forever.
• Research datasets where every clip's metadata needs SQL access.
• Compliance / audit workflows where 'show me every frame where Y happened' is a regular query.

Pitfalls

• Default DB lives in `~/.peepshow/sinks/sqlite/peepshow.db` — set `PEEPSHOW_SQLITE_PATH` to relocate.
• Very high frame volumes (>1M frames total) — VACUUM occasionally.

Works with these LLMs

Pairs with these sinks