video1.0<iframe src="https://www.loom.com/embed/d0b6df37b0514febab2706f1488a3a84" frameborder="0" width="1670" height="1252" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12521670Loomhttps://www.loom.com12521670https://cdn.loom.com/sessions/thumbnails/d0b6df37b0514febab2706f1488a3a84-78fb52e1d553b060.gif340.647In this Loom I demo my Keyboard Manual Assistant, a FastAPI backend RAG app using QDurant for vector search, Sentence Transformers for embeddings, a local LLM, and LangFuse for observability. I upload the Montage manual PDF, convert it to text, chunk and embed it, then save it in the VectorData store. I ask a question like how do I layer sounds on a Montage keyboard and it returns detailed answers from retrieved chunks. In LangFuse I show the trace, retrieved chunks, and the response, plus an eval where the system did not behave exactly as expected regarding jazz voicings. No action is requested from viewers.