Skip to content
AgenticRAG
  • Discover
  • Marketplace
  • Console
  • Atlas
  • Protocol
  • Analytics
—credits
Agentic RAG Paywall

A knowledge API priced per answer. Agents discover the price from a 402, settle on Stellar, and get a cited response.

Live on Stellar testnet

Product

  • Console
  • Knowledge library
  • Analytics

Protocol

  • x402 walkthrough
  • Integration guide
  • OpenAPI reference

Stack

  • Next.js 16 · React 19
  • FastAPI · Python 3.12
  • Gemini · ChromaDB
  • Stellar SDK · x402

© 2026 Agentic RAG Paywall · MIT licensed

Source

Embedding space

The corpus, as the retriever sees it

Every chunk is a 3072-dimensional vector. This projects them onto their two principal components — and because that projection is linear, a live query lands in exactly the same space.

▸Visualising the semantic relationships between documents and the queries that retrieve them.

Try:Retrieval-only — costs nothing

Documents

Click to isolate a source

    How to read this

    • Each dot is one indexed passage. Colour is its source document.
    • Proximity is semantic: passages that mean similar things sit together, even with no words in common.
    • Project a query and it appears as a glowing marker, linked to the passages the retriever chose.

    Projection

    • MethodPCA · 2 components
    • Variance kept0%
    • PC1 / PC2
    • Source dimensions—

    PCA is linear, so new queries project into the same basis. t-SNE and UMAP look tidier but cannot place a new point without refitting — which would make this overlay a lie.