Python: Weaviate vector store

python/.coveragerc

@@ -13,7 +13,7 @@ omit =
     semantic_kernel/connectors/memory/qdrant/qdrant_memory_store.py
     semantic_kernel/connectors/memory/redis/redis_memory_store.py
     semantic_kernel/connectors/memory/usearch/*
-    semantic_kernel/connectors/memory/weaviate/*
+    semantic_kernel/connectors/memory/weaviate/weaviate_memory_store.py
     semantic_kernel/reliability/*
     semantic_kernel/memory/*
 

python/.cspell.json

@@ -34,6 +34,7 @@
         "endregion",
         "genai",
         "generativeai",
+        "hnsw",
         "httpx",
         "huggingface",
         "kernelfunction",
@@ -58,7 +59,9 @@
         "skprompt",
         "templating",
         "vectordb",
+        "vectorizer",
         "vectorstoremodel",
-        "vertexai"
+        "vertexai",
+        "Weaviate"
     ]
 }

python/pyproject.toml

@@ -27,29 +27,23 @@ dependencies = [
     "pydantic ~= 2.0",
     "pydantic-settings ~= 2.0",
     "defusedxml ~= 0.7",
-
     # azure identity
     "azure-identity ~= 1.13",
-
     # embeddings
     "numpy >= 1.25.0; python_version < '3.12'",
     "numpy >= 1.26.0; python_version >= '3.12'",
-
     # openai connector
     "openai ~= 1.0",
-
     # openapi and swagger
     "openapi_core >= 0.18,<0.20",
-
     # OpenTelemetry
     "opentelemetry-api ~= 1.24",
     "opentelemetry-sdk ~= 1.24",
     "prance ~= 23.6.21.0",
-
     # templating
     "pybars4 ~= 0.9",
     "jinja2 ~= 3.1",
-    "nest-asyncio ~= 1.6"
+    "nest-asyncio ~= 1.6",
 ]
 
 ### Optional dependencies
@@ -109,7 +103,7 @@ usearch = [
     "pyarrow >= 12.0,<18.0"
 ]
 weaviate = [
-    "weaviate-client >= 3.18,<5.0"
+    "weaviate-client>=4.7,<5.0",
 ]
 pandas = [
     "pandas ~= 2.2"

python/samples/concepts/memory/new_memory.py

@@ -16,6 +16,7 @@
 from semantic_kernel.connectors.memory.qdrant import QdrantCollection
 from semantic_kernel.connectors.memory.redis import RedisHashsetCollection, RedisJsonCollection
 from semantic_kernel.connectors.memory.volatile import VolatileCollection
+from semantic_kernel.connectors.memory.weaviate.weaviate_collection import WeaviateCollection
 from semantic_kernel.data import (
     VectorStoreRecordCollection,
     VectorStoreRecordDataField,
@@ -79,6 +80,9 @@ class MyDataModelList:
 # - redis_hashset: Redis Hashset
 # - qdrant: Qdrant
 # - volatile: In-memory store
+# - weaviate: Weaviate
+#   Please either configure the weaviate settings via environment variables or provide them through the constructor.
+#   Note that embed mode is not supported on Windows: https://github.com/weaviate/weaviate/issues/3315
 #
 # This is represented as a mapping from the store name to a
 # function which returns the store.
@@ -109,6 +113,10 @@ class MyDataModelList:
         data_model_type=MyDataModel,
         collection_name=collection_name,
     ),
+    "weaviate": lambda: WeaviateCollection[MyDataModel](
+        data_model_type=MyDataModel,
+        collection_name=collection_name,
+    ),
 }
 
 

python/semantic_kernel/connectors/memory/azure_ai_search/utils.py

@@ -90,6 +90,7 @@ def data_model_definition_to_azure_ai_search_index(
 
     for field in definition.fields.values():
         if isinstance(field, VectorStoreRecordDataField):
+            assert field.name  # nosec
             if not field.property_type:
                 logger.debug(f"Field {field.name} has not specified type, defaulting to Edm.String.")
             type_ = TYPE_MAPPER_DATA[field.property_type or "default"]
@@ -119,6 +120,7 @@ def data_model_definition_to_azure_ai_search_index(
                 )
             )
         elif isinstance(field, VectorStoreRecordVectorField):
+            assert field.name  # nosec
             if not field.property_type:
                 logger.debug(f"Field {field.name} has not specified type, defaulting to Collection(Edm.Single).")
             if not field.index_kind:

python/semantic_kernel/connectors/memory/weaviate/README.md

@@ -0,0 +1,17 @@
+# Weaviate Memory Connector
+
+[Weaviate](https://weaviate.io/developers/weaviate) is an open source vector database. Semantic Kernel provides a connector to allow you to store and retrieve information for you AI applications from a Weaviate database.
+
+## Setup
+
+There are a few ways you can deploy your Weaviate database:
+- [Weaviate Cloud](https://weaviate.io/developers/weaviate/installation/weaviate-cloud-services)
+- [Docker](https://weaviate.io/developers/weaviate/installation/docker-compose)
+- [Embedded](https://weaviate.io/developers/weaviate/installation/embedded)
+- Other cloud providers such as [AWS](https://weaviate.io/developers/weaviate/installation/aws-marketplace) or [GCP](https://weaviate.io/developers/weaviate/installation/gc-marketplace).
+
+> Note that embedded mode is not supported on Windows yet: [GitHub issue](https://github.com/weaviate/weaviate/issues/3315)
+
+## Using the Connector
+
+Once the Weaviate database is up and running, and the environment variables are set, you can use the connector in your Semantic Kernel application. Please refer to this sample to see how to use the connector: [Weaviate Connector Sample](../../../../samples/concepts/memory/new_memory.py)

python/semantic_kernel/connectors/memory/weaviate/const.py

@@ -0,0 +1,15 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+from weaviate.classes.config import DataType
+
+TYPE_MAPPER_DATA = {
+    "str": DataType.TEXT,
+    "int": DataType.INT,
+    "float": DataType.NUMBER,
+    "bool": DataType.BOOL,
+    "list[str]": DataType.TEXT_ARRAY,
+    "list[int]": DataType.INT_ARRAY,
+    "list[float]": DataType.NUMBER_ARRAY,
+    "list[bool]": DataType.BOOL_ARRAY,
+    "default": DataType.TEXT,
+}