Streaming apps only recommend from their own catalogue, and only know what you watched on their service. Streamline reads the GDPR data exports from Netflix, Prime Video, Apple TV, Disney+ and Max, plus Plex plays, so it has your whole history.
How it works: each title is matched on TMDB and gets a short LLM-written description. From the titles you mark as loved, it builds a taste profile. The LLM tags each title once, and code groups them into themes, so the same ratings always give the same profile. A search turns your request into structured filters, gathers candidates from TMDB Discover and from LLM suggestions, removes everything you've seen, and ranks the rest against your request first and your taste second.
It's self-hosted (Docker, SQLite) and works with Claude, Gemini, OpenAI, or a local model. Searches cost about 6 cents with Claude. The docs screenshots are rebuilt by CI from a made-up library, so they never go stale and never show real data.
Comments URL: https://news.ycombinator.com/item?id=50013666
Points: 1
# Comments: 0
