Yail.Ragretrieval · embeddings · chat
Ask your stuff
get an answer.
Local-first embeddings on your hardware, Claude for generation, hybrid retrieval with citations, source connectors via Yail.Connect. Built on the YAIL-OS module template.
At a glance
Will fill out once you're signed in.
Jump to
The retrieval stack, end to end.
Collections
corpora · embeddings · mode
Group documents by purpose. Each collection has its own embedding model, chunking strategy, and retrieval mode (dense, lexical, or hybrid).
Chat
Claude · citations · streaming
Conversational interface over your corpus. Every claim comes back with the source chunks it drew from, so answers are auditable.
Search
raw retrieval · top-k
Hybrid pgvector + BM25 retrieval, no LLM in the loop. Useful for evals, debugging chunking, or embedding in other apps via the Rag API.
Sources
Connect · webhook · cron
Wire up Google Drive, GitHub, IMAP — anything Yail.Connect can reach. Changes stream into the index on a schedule or webhook.
Evals
golden sets · regression
Pin a golden set of questions + expected citations. Re-run after any chunking, model or prompt change — catch regressions before users do.
Widget
embed anywhere
Drop a single <script> tag onto any page to embed the chat widget — scoped to a collection, themed via Yail.Branding, gated by API key.