pip install azure-search-documents azure-identity

AZURE_SEARCH_ENDPOINT=https://<resource>.search.windows.net   # Endpoint del servicio
AZURE_SEARCH_INDEX_NAME=products-catalog                       # Nombre del índice
AZURE_TOKEN_CREDENTIALS=prod                                   # Solo si DefaultAzureCredential en produccion
AZURE_SEARCH_API_KEY=<api-key>                                 # Solo para legacy API-key auth (ver abajo)

import os
from azure.identity import DefaultAzureCredential, ManagedIdentityCredential
from azure.search.documents import SearchClient
from azure.search.documents.indexes import SearchIndexClient

# Dev local: DefaultAzureCredential. Prod: AZURE_TOKEN_CREDENTIALS=prod
credential = DefaultAzureCredential(require_envvar=True)
# Produccion con credencial explicita:
# credential = ManagedIdentityCredential()

endpoint = os.environ["AZURE_SEARCH_ENDPOINT"]
index_name = os.environ["AZURE_SEARCH_INDEX_NAME"]

with SearchIndexClient(endpoint=endpoint, credential=credential) as index_client:
    names = [idx.name for idx in index_client.list_index_names()]
    print(f"Indices disponibles: {names}")

with SearchClient(endpoint=endpoint, index_name=index_name, credential=credential) as client:
    result = client.search(search_text="*", top=1)
    print("Conexion OK — primer doc:", next(iter(result), None))

from azure.core.credentials import AzureKeyCredential
from azure.search.documents import SearchClient

credential = AzureKeyCredential(os.environ["AZURE_SEARCH_API_KEY"])
with SearchClient(endpoint=endpoint, index_name=index_name, credential=credential) as client:
    ...

from azure.search.documents.indexes.models import (
    SearchIndex, SearchField, SearchFieldDataType,
    SimpleField, SearchableField, VectorSearch,
    HnswAlgorithmConfiguration, VectorSearchProfile,
    SemanticConfiguration, SemanticSearch, SemanticPrioritizedFields,
    SemanticField
)

fields = [
    SimpleField(name="id", type=SearchFieldDataType.String, key=True),
    SearchableField(name="nombre", type=SearchFieldDataType.String, analyzer_name="es.microsoft"),
    SearchableField(name="descripcion", type=SearchFieldDataType.String, analyzer_name="es.microsoft"),
    SimpleField(name="categoria", type=SearchFieldDataType.String, filterable=True, facetable=True),
    SimpleField(name="precio", type=SearchFieldDataType.Double, filterable=True, sortable=True),
    SearchField(
        name="vector_descripcion",
        type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
        searchable=True, vector_search_dimensions=1536,
        vector_search_profile_name="my-hnsw-profile"
    ),
]

vector_search = VectorSearch(
    algorithms=[HnswAlgorithmConfiguration(name="my-hnsw")],
    profiles=[VectorSearchProfile(name="my-hnsw-profile", algorithm_configuration_name="my-hnsw")]
)
semantic_config = SemanticConfiguration(
    name="my-semantic-config",
    prioritized_fields=SemanticPrioritizedFields(
        title_field=SemanticField(field_name="nombre"),
        content_fields=[SemanticField(field_name="descripcion")]
    )
)
index = SearchIndex(
    name=index_name,
    fields=fields,
    vector_search=vector_search,
    semantic_search=SemanticSearch(configurations=[semantic_config])
)

with SearchIndexClient(endpoint=endpoint, credential=credential) as index_client:
    created = index_client.create_or_update_index(index)
    print(f"Indice '{created.name}' listo — {len(created.fields)} campos")

from azure.search.documents.models import IndexDocumentsBatch

productos = [
    {
        "id": "001", "nombre": "Cafe de Especialidad Guatemala",
        "descripcion": "Granos arábica de altura, notas de chocolate y naranja",
        "categoria": "cafe", "precio": 18.50,
        "vector_descripcion": generar_embedding("Granos arábica de altura...")
    },
    {
        "id": "002", "nombre": "Kit Inicio Cultivo Microgreens",
        "descripcion": "Todo lo necesario para cultivar microgreens en casa en 10 dias",
        "categoria": "cultivo", "precio": 34.90,
        "vector_descripcion": generar_embedding("Todo lo necesario para cultivar...")
    },
]

with SearchClient(endpoint=endpoint, index_name=index_name, credential=credential) as client:
    results = client.upload_documents(documents=productos)
    success = [r for r in results if r.succeeded]
    print(f"Cargados: {len(success)}/{len(productos)}")

from azure.search.documents.models import VectorizedQuery, QueryType

query = "regalo especial para amante del cafe con notas frutales"
query_vector = generar_embedding(query)

with SearchClient(endpoint=endpoint, index_name=index_name, credential=credential) as client:
    results = client.search(
        search_text=query,
        query_type=QueryType.SEMANTIC,
        semantic_configuration_name="my-semantic-config",
        query_caption="extractive",
        vector_queries=[
            VectorizedQuery(
                vector=query_vector,
                k_nearest_neighbors=5,
                fields="vector_descripcion"
            )
        ],
        select=["id", "nombre", "descripcion", "precio"],
        top=3
    )
    for doc in results:
        score = doc.get("@search.score", 0)
        print(f"[{score:.3f}] {doc['nombre']} — €{doc['precio']}")