Skip to content
Demiton

Azure AI Search

Stable Connected tier

Azure AI Search is the vector store powering Ask Demiton’s document retrieval. Every document indexed here carries identity-scoped access control fields - the AI retrieves only documents the requesting user is authorised to see.

Authentication: OAuth 2.0 via Entra ID service principal
System type: AZURE_AI_SEARCH
API reference: Azure AI Search REST API

Supported resources

ResourceFETCHPUSHLOOKUPEXECUTE
Per-index (dynamic)✓✓✓
__indexes__✓
__stats__✓
__schema__✓
Index management✓

FETCH on a named index performs vector semantic search. PUSH ingests documents in batches of 32 with automatic embedding generation. LOOKUP retrieves a document by ID. EXECUTE supports CREATE_INDEX, CREATE_INDEX_FROM_TEMPLATE, and DELETE_INDEX.

Integration model

This connector is configured and managed by Demiton as part of your deployment.

Every document stored in the index is tagged with allowed_entra_users and allowed_entra_groups fields. All queries are filtered by the requesting user’s resolved Entra identity - users cannot retrieve documents outside their authorised scope. If the ACL check fails, zero results are returned (never partial or unfiltered results).

Document ingestion uses HNSW vector indexing with embedding generation via the platform’s embedding service. Batch ingestion includes retry logic and rate-limit awareness.

Security

Access type: Platform-managed - read and write (index management)

Index data is housed within your Azure deployment region and is not shared between customers. Every query is filtered by the requesting user’s Entra identity before results are returned.

Next steps