AliBot
The assistant in the corner of this page. It answers questions about my work by retrieving the site's own content rather than recalling it: questions are embedded with bge-base and searched against Cloudflare Vectorize alongside BM25, the two rankings are fused, and a cross-encoder reranker picks what the model may use. Answers stream from Llama 3.3 70B on a Cloudflare Worker and cite the card they came from. Below a confidence threshold it declines instead of guessing.














