1# Azure AI Search — Python SDK Quick Reference
2 
3> Condensed from **azure-search-documents-py**. Full patterns (agentic retrieval, integrated vectorization, skillsets)
4> in the **azure-search-documents-py** plugin skill if installed.
5 
6## Install
7```bash
8pip install azure-search-documents azure-identity
9```
10 
11## Quick Start
12```python
13from azure.search.documents import SearchClient
14from azure.search.documents.indexes import SearchIndexClient, SearchIndexerClient
15from azure.search.documents.models import VectorizedQuery
16```
17 
18## Non-Obvious Patterns
19- `SearchIndexingBufferedSender` for batch uploads with auto-batching/retries
20- Vector field type: `Collection(Edm.Single)` with `vector_search_dimensions` + `vector_search_profile_name`
21- Async client: `from azure.search.documents.aio import SearchClient`
22- `KnowledgeBaseRetrievalClient` for agentic retrieval with LLM-powered Q&A
23 
24## Best Practices
251. Use hybrid search for best relevance combining vector and keyword
262. Enable semantic ranking for natural language queries
273. Index in batches of 100-1000 documents for efficiency
284. Use filters to narrow results before ranking
295. Configure vector dimensions to match your embedding model
306. Use HNSW algorithm for large-scale vector search
317. Create suggesters at index creation time (cannot add later)
328. Use `SearchIndexingBufferedSender` for batch uploads
339. Always define semantic configuration for agentic retrieval indexes
3410. Use `create_or_update_index` for idempotent index creation
3511. Close clients with context managers or explicit `close()`
36