Python: Bug: AzureAISearchCollection.hybrid_search sends declared field names instead of storage_name aliases #14684

Describe the bug

AzureAISearchCollection.hybrid_search() sends the model's declared field names instead of their storage_name aliases. When a field declares storage_name, the index is created with the storage names, so the hybrid request references fields the index does not contain and Azure AI Search rejects it (400 unknown field).

Index field names created by ensure_collection_exists: ['doc_id', 'body', 'vec']
hybrid_search search_fields   : ['content']
hybrid_search vector fields   : embedding
vector search vector fields   : vec

Root cause: the hybrid branch uses field.name / vector_field.name (python/semantic_kernel/connectors/azure_ai_search.py:602, :616, :623), while the vector branch (:572, :582, :589) correctly uses storage_name or name. The index creation path also uses the storage names (:235, :269).

To Reproduce

With a model whose content field declares storage_name="body" and whose embedding field declares storage_name="vec":

1. Call ensure_collection_exists() — the index is created with doc_id, body, vec.
2. Call hybrid_search(...) — the emitted request uses search_fields=['content'] and vector_queries[0].fields='embedding'.
3. Call search(vector=...) — the vector-only request correctly uses vec.

A self-contained script that captures the request objects (no live service) prints the three lines above; run from python/:

.venv/Scripts/python.exe -I repro_azure_ai_search_hybrid_storage_names.py

Requires azure-search-documents>=11.6.0b4,<13.0.0 (the connector's own dependency).

Expected behavior

hybrid_search() should use storage_name or name for search_fields and vector fields, matching both the index definition and the vector-only search path (body/vec above).

Platform

• Language: Python
• Source: main branch at 4bb64ee (2026-10-28), semantic-kernel 1.46.0
• AI model: n/a (request construction only)
• OS: Windows 11

Additional context

• Searched open and closed issues/PRs for storage_name, azure_ai_search, and hybrid search: no duplicate; no open PR touches azure_ai_search.py. Python: Chroma, MongoDB, Pinecone and Weaviate filters raise NotImplementedError for 'in' with a list literal (and Pinecone 'not' raises TypeError) #14557 covers list-literal filters in Chroma/Mongo/Pinecone/Weaviate only.
• Suggested fix: use storage_name or name in the hybrid branch, and add a test asserting hybrid and vector paths agree when aliases are configured.

Reported with AI assistance (analysis and reproduction script); the reporter ran the reproduction and verified the output above.