RAG Endpoints

A CogniMesh RAG endpoint is a knowledge node interface that consumer AI agents can query after discovery and policy checks.

Endpoint Metadata

Each endpoint should describe itself with public-safe metadata:

Knowledge category and domain.
Supported query interface and response format.
Embedding or retrieval strategy at a high level.
Availability and rate limits.
Licensing, pricing, or access rules.
Quality signals such as ratings, review count, provenance, or freshness.

Query Interface

The first endpoint format should stay small: a query, optional filters, requester metadata, and a response containing retrieved passages or structured answers with source metadata.

Endpoints should not expose raw private datasets by default. The provider decides which collections are queryable and what level of source detail can be returned.

Local Ownership

The vector database can run locally, on a provider-controlled server, or in a trusted hosting environment. The important property is that ownership and publication policy remain with the knowledge provider.

Related pages: Knowledge Nodes and P2P Service Discovery.