RAGSuite is built as a triad: AI Connectors bring your knowledge in, AI Search returns citation-backed answers across it, and the AI Assistant turns that into grounded conversation — all on your own infrastructure. The three are designed together, which is why we never describe one without the others.
What each part does
Three parts, one loop
- AI Connectors — the front door: website crawl, document upload, Gmail, MCP, n8n (Beta), Connector Marketplace. All in the open Community core.
- AI Search — semantic, citation-backed answers; every result traceable to the document it came from.
- AI Assistant — Search in conversational form, sources on every reply, context across a dialogue.
Why the three belong together
Each part is weaker alone. Together they form a single loop: knowledge in, grounded answers out, sources on everything. And because the whole loop runs self-hosted and model-agnostic — OpenAI, Anthropic, Mistral, Gemini or local Ollama via Compare Models — you get the modern experience without giving up control. Explore each on the product page, or see the engine in how citation-backed RAG works.
Frequently asked questions
Are widgets a separate product?
No. Embeddable widgets are how you publish AI Search and the AI Assistant — a delivery mechanism, not a fourth pillar. Drop them into an intranet, help centre or customer-facing page.
Do I have to use all three?
They are designed together and reinforce each other — Connectors without grounding is just ingestion; Search without Connectors has nothing to search; an Assistant without citations is the thing enterprises are trying to avoid. Used together, you get the full loop.
Sources & further reading
- RAGSuite — platform overview — the triad in the product
- How citation-backed RAG works — the engine behind the triad