{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/796076cf00834195b62403e6b505cdeb\" frameborder=\"0\" width=\"1920\" height=\"1440\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1440,"width":1920,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1440,"thumbnail_width":1920,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/796076cf00834195b62403e6b505cdeb-bd63ee4c10cf8571.gif","duration":456.896,"title":"Agentic Music Recommender with RAG Workflow 🎵","description":"I built an agentic music recommender with RAG over Spotify song lyrics and metadata, plus feature filtering using Kaggle data. In a simple Streamlit UI, you type a plain English vibe like energetic and happy, about the beauty of life, and the agents extract preferences, plan retrieval, run lyric RAG, metadata search, and audio feature filtering, then verify, critique, revise if needed, and explain why each song was chosen. Results show confidence around 0.55, and I traced everything with LangSmith. I ran evaluation on 10 prompts and a 20 task blind human comparison, plus an LLM judge, to compare full vs ablated systems. There was no explicit action requested from viewers."}