Knowledge Base

Point the crawler at your help center URL and the knowledge base is active in 30 minutes. For corpora under 900K tokens the AI operates on the full document set with prompt caching — 90% cost reduction versus naive retrieval. Larger corpora automatically switch to hybrid BM25 + semantic RAG with cross-encoder re-ranking.

• ✓URL crawl ingestion: point at your help center URL, knowledge base active in 30 minutes
• ✓Full-context grounding with 90% cost reduction via prompt caching
• ✓pgvector RAG auto-activates when corpus exceeds 900K tokens
• ✓Source citations on every AI response — contacts can verify the information themselves