Knowledge base & RAG
Verabase ingests your content and makes it queryable by AI agents through a hybrid vector + full-text search pipeline.
Supported sources
How ingestion works
Connect a source
Paste a URL, upload a file, or authenticate your Notion/Drive workspace. The system queues a background job immediately.
Chunking & vectorization
Content is split into logical chunks and converted into vector embeddings using your LLM provider. Duplicate content is detected automatically via content hashing.
Visual grounding for PDFs
Charts, diagrams, and images inside uploaded PDFs are analyzed by a vision model and indexed as descriptive text. Your agents can answer questions about graphs and infographics, not just written text.
Self-healing knowledge loop
When an agent can't answer a question, Verabase logs it as a knowledge gap. The system clusters similar unanswered questions, finds how your team resolved them historically, and drafts a suggested FAQ for you to review.
Review gaps in Knowledge Insights
Go to Knowledge → Insights to see clustered gaps ranked by frequency. Review the AI-drafted Q&A, edit if needed, and click Approve & Publish. The agent learns instantly.