FAQ

Standard RAG re-derives an answer from raw documents on every query. Vessios accumulates knowledge in a wiki: a single ingest creates 10–15 pages, and good query answers are written back into the wiki, so the system gets better with use instead of staying static. See Introduction.

Text, Markdown, PDF, .docx, .xlsx, and images (via Claude vision or local OCR). URLs are also supported, with SSRF protection. See Source ingestion.

Every ingest goes through duplicate detection, prompt-injection screening, and PII flagging before it touches the wiki. On top of that, the Lint job (Pro and above) runs weekly to detect contradictions and stale content — see Automatic maintenance.

Yes. Every tenant’s pages, sources, and knowledge graph live in isolated Firestore subcollections, and vector search isolation is enforced deterministically by tenant, not by a hash function. See Security.

Yes — as llms.txt / llms-full.txt (for feeding into downstream LLM workflows) or graphml (for graph visualization), via POST /export. See Integrations.

Pro or Enterprise. Free and Starter tenants can still read the same underlying wiki-health signals, but the scheduled weekly Lint and confidence-decay jobs require Pro or above. See Plans & billing.

Yes — Vessios can post notifications to Slack and answer questions asked directly in Slack. See Integrations.