SurfableAI
Sign inSignup
All docs
Advanced3 min readUpdated May 28, 2026

FAQ

Common questions about semantic search, RAG, agents, MCP, and credits.

How does semantic search differ from keyword search?

Keyword search looks for exact word matches. Semantic search uses vector embeddings to understand the meaning behind a query — finding relevant content even when the exact words don't match. A user asking "how much does it cost?" will surface your pricing page even if that page never uses the word "cost".

What models do you use?

  • Generation: Gemini 3 Flash Preview (RAG, agent responses, and optimizer analysis)
  • Embeddings: Gemini Embedding 2 with 768-dimensional vectors
  • Voice: Gemini Live API/native-audio model for real-time voice mode

What is the difference between /rag and /agent/chat/stream?

/rag is a single-turn RAG response: one query → one grounded answer with citations. /agent/chat/stream is multi-step: the agent can chain searches, call tools, maintain session memory across turns, and complete multi-step reasoning tasks. Use /rag for a chatbot or FAQ widget. Use /agent/chat/stream when the user's task might require more than one search or a sequence of actions.

What is MCP and why does it matter?

Model Context Protocol (MCP) exposes your site's content and registered API endpoints as structured, machine-callable tools. AI agents — including ChatGPT, Claude, and Copilot — can discover and interact with your business through MCP without custom per-agent integrations.

Will AI answers hallucinate information?

Surfable uses strict RAG: every answer is retrieved from your ingested content first, then synthesised. If the answer isn't in your indexed content, the AI says so rather than guessing. Each response includes source citations you can verify.

How do credits work?

Credits are consumed per operation — crawled pages, search queries, RAG calls, and agent steps each have a defined cost. Check your current balance via GET /usage or the Dashboard → Usage tab. Credits reset monthly; see your plan limits on the pricing page.

How often is my site re-indexed?

Scheduled re-crawls are available on every plan and consume credits when changed content is crawled and embedded. You can also trigger a manual re-crawl at any time from Dashboard → Index → Re-crawl, or via POST /ingest through the API.

Can I index PDFs or other documents?

Yes. Upload PDFs directly via Dashboard → Index → Upload, or POST /content/upload. PDFs are extracted, chunked, and embedded the same way as web pages.

What happens when I update my website?

Surfable only reflects your site's content as of the last crawl. Re-index after publishing significant changes to keep answers accurate.

Next Steps