[Feature Request] Add hybrid/vector search support to Query Index tool

[farzad528]

Describe the bug

The search_index_query tool only supports keyword (BM25) search. It passes a plain text string to SearchClient.SearchAsync with default SearchOptions and does not construct any vectorQueries in the request body. This means even when an index has a vectorizer configured, the tool cannot perform hybrid search (keyword + vector + RRF fusion). Customers who want hybrid search through the Azure MCP Server must create a Knowledge Base and use the knowledge_base_retrieve tool, which adds unnecessary overhead for simple hybrid query scenarios.

Expected behavior

When querying an index that has a vectorizer configured, the Query Index tool should automatically construct a hybrid search request by:

1. Detecting searchable vector fields and their associated vectorizer from the index schema
2. Adding a vectorQueries entry with kind: "text" alongside the existing keyword search parameter
3. Returning RRF-fused results

Optionally expose a --query-type parameter (simple | hybrid | semantic) to give users explicit control.

Actual behavior

IndexQueryCommand.cs registers only three options (Service, Index, Query) and calls searchService.QueryIndex(service, index, query, retryPolicy, ct) with no vector configuration. The vectorizer on the index is never invoked. Results are BM25 keyword-ranked only.

Reproduction Steps

1. Create an index with a vectorizer and both text + vector fields
2. Index documents with embeddings
3. Use the Azure MCP Server Query Index tool: "Search for 'machine learning' in index 'my-index' of service 'my-service'"
4. Compare results with a direct REST hybrid query against the same index
5. Observe that MCP results are keyword-ranked only (no vector component)

Environment

Environment:

• Hosting platform or OS: macOS / Windows
• Azure MCP Server version: @azure/mcp@latest (1.x GA)
• IDE: VS Code + GitHub Copilot Agent Mode / GitHub Copilot CLI

[farzad528]

Closing as @pablocastro pointed me to the logic that uses a vectorizer. Perhaps we could update the documentation here to specifically say that Hybrid Search is supported if an index has a vectorizer configured? Unfortunately, GH Copilot CLI did not pick up on this and told me that only "simple" search is supported.

https://learn.microsoft.com/en-us/azure/developer/azure-mcp-server/tools/azure-ai-search