Parents ask questions in plain English. AI finds the answer from your school's official documents — with citations.
Under the hood
Every answer is built from your school's own source material — never invented. Here's how it works, end to end.
PDFs, Word, Excel, and PowerPoint flow through our ingestion pipeline. Scanned PDFs and image-based handouts are routed through our in-house vision model, which extracts text while preserving tables, headings, and structure.
Each document is split into overlapping chunks and embedded with our proprietary 768-dimensional embedding model. Vectors are stored in Postgres via pgvector for fast similarity search at query time.
Row-level security guarantees parents only retrieve chunks from documents their school has approved. Each question is embedded on the fly and matched against the corpus by vector similarity.
The top-ranked chunks are passed as context to our answer model, which composes a response with inline citations linking back to the exact source document and section — so every claim is verifiable.