v1.0.0 — Initial Release

Personal streaming recommendation engine built on real watch history.

Highlights

• Two-phase LLM pipeline — offline taste profiling from all watched titles, online natural language queries
• Multi-provider support — Anthropic (Claude) and Google (Gemini) with role-based model dispatch
• Hybrid candidate generation — TMDB Discover + LLM semantic suggestions
• Full taste profile — batched processing of 2000+ titles across 15 taste clusters
• Web UI — Flask + HTMX with search, taste profile dashboard, and poster archive
• Rich CLI — interactive REPL, conversational context, feedback system, usage stats
• Title matching — guessit classification, rapidfuzz dedup, manual overrides
• Streaming availability — annotated results with platform filtering
• Configurable — config.yaml for all settings, .env for secrets