Seven formats
Plain text through PDF, Word and Excel — one intake for everything.
Knowledge
Company knowledge base
Feed manuals, pricing rules and FAQs to the agent — answers cite their sources.
What it does
Seven formats upload directly — txt, markdown, csv, html, pdf, docx, xlsx — parsed and chunked locally into the vector store. Retrieval is hybrid: vector recall reranked by keyword and phrase hits, and every answer carries source snippets you can click back to.
Embeddings default to a zero-cost local scheme and switch to a real embedding service in production; dimension mismatches fail loudly rather than silently mixing indexes; scanned PDFs and complex tables fall back to external OCR parsing.
Core capabilities
Hybrid retrieval
Vector plus keyword and phrase reranking; answers arrive with receipts.
OCR fallback
Scans and complex tables route to external parsing — no dead corners.
Feature checklist
- Qdrant vector store with overlap-window chunking
- Three agent tools: search, answer context, source listing
- Reindexing, chunk inspection and store stats built in
Want to see this module run?
Book a demo and we'll run this module on real data.