chore(knowledgebase): add support for vector type (#949)

* chore(knowledgebase): add support for vector type

---------

Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

.projenrc.ts

@@ -29,7 +29,7 @@ import {
 const GITHUB_USER = 'awslabs';
 const PUBLICATION_NAMESPACE = 'cdklabs';
 const PROJECT_NAME = 'generative-ai-cdk-constructs';
-const CDK_VERSION: string = '2.177.0';
+const CDK_VERSION: string = '2.178.0';
 
 function camelCaseIt(input: string): string {
   // Hypens and dashes to spaces and then CamelCase...

CHANGELOG.md

@@ -1,3 +1,7 @@
+# CDK Generative AI Constructs V0.1.293 (2025-02-10)
+
+Based on CDK library version 2.178.0
+
 # CDK Generative AI Constructs V0.1.291 (2025-01-26)
 
 Based on CDK library version 2.177.0

DEVELOPER_GUIDE.md

@@ -16,7 +16,7 @@ Default output format [None]: json
 ```
 
 - [Node](https://nodejs.org/en) >= v20.9.0
-- [AWS CDK](https://github.com/aws/aws-cdk/releases/tag/v2.177.0) >= 2.177.0
+- [AWS CDK](https://github.com/aws/aws-cdk/releases/tag/v2.178.0) >= 2.178.0
 - [Python](https://www.python.org/downloads/) >=3.9
 - [Projen](https://github.com/projen/projen) >= 0.91.5
 - [Yarn](https://classic.yarnpkg.com/lang/en/docs/cli/install/) >= 1.22.19

apidocs/namespaces/bedrock/README.md

@@ -35,6 +35,7 @@
 - [SharePointDataSourceAuthType](enumerations/SharePointDataSourceAuthType.md)
 - [SharePointObjectType](enumerations/SharePointObjectType.md)
 - [TransformationStep](enumerations/TransformationStep.md)
+- [VectorType](enumerations/VectorType.md)
 
 ## Classes
 

apidocs/namespaces/bedrock/classes/BedrockFoundationModel.md

@@ -65,6 +65,12 @@ The ARN of the Bedrock invokable abstraction.
 
 ***
 
+### supportedVectorType?
+
+> `readonly` `optional` **supportedVectorType**: [`VectorType`](../enumerations/VectorType.md)[]
+
+***
+
 ### supportsAgents
 
 > `readonly` **supportsAgents**: `boolean`

apidocs/namespaces/bedrock/enumerations/VectorType.md

@@ -0,0 +1,28 @@
+[**@cdklabs/generative-ai-cdk-constructs**](../../../README.md)
+
+***
+
+[@cdklabs/generative-ai-cdk-constructs](../../../README.md) / [bedrock](../README.md) / VectorType
+
+# Enumeration: VectorType
+
+The data type for the vectors when using a model to convert text into vector embeddings.
+The model must support the specified data type for vector embeddings. Floating-point (float32)
+is the default data type, and is supported by most models for vector embeddings. See Supported
+embeddings models for information on the available models and their vector data types.
+
+## Enumeration Members
+
+### BINARY
+
+> **BINARY**: `"BINARY"`
+
+`BINARY` convert the data to binary vector embeddings (less precise, but less costly).
+
+***
+
+### FLOATING\_POINT
+
+> **FLOATING\_POINT**: `"FLOAT32"`
+
+`FLOATING_POINT` convert the data to floating-point (float32) vector embeddings (more precise, but more costly).

apidocs/namespaces/bedrock/interfaces/BedrockFoundationModelProps.md

@@ -8,6 +8,14 @@
 
 ## Properties
 
+### supportedVectorType?
+
+> `readonly` `optional` **supportedVectorType**: [`VectorType`](../enumerations/VectorType.md)[]
+
+Embeddings models have different supported vector types
+
+***
+
 ### supportsAgents?
 
 > `readonly` `optional` **supportsAgents**: `boolean`

apidocs/namespaces/bedrock/interfaces/VectorKnowledgeBaseProps.md

@@ -149,3 +149,17 @@ type `VectorCollection`, `RedisEnterpriseVectorStore`,
 ```ts
 - A new OpenSearch Serverless vector collection is created.
 ```
+
+***
+
+### vectorType?
+
+> `readonly` `optional` **vectorType**: [`VectorType`](../enumerations/VectorType.md)
+
+The vector type to store vector embeddings.
+
+#### Default
+
+```ts
+- VectorType.FLOATING_POINT
+```

apidocs/namespaces/opensearch_vectorindex/interfaces/VectorIndexProps.md

@@ -32,6 +32,12 @@ The OpenSearch Vector Collection.
 
 ***
 
+### distanceType
+
+> `readonly` **distanceType**: `string`
+
+***
+
 ### indexName
 
 > `readonly` **indexName**: `string`
@@ -48,6 +54,12 @@ The metadata management fields.
 
 ***
 
+### precision
+
+> `readonly` **precision**: `string`
+
+***
+
 ### vectorDimensions
 
 > `readonly` **vectorDimensions**: `number`

lambda/opensearch-serverless-custom-resources/custom_resources/opensearch_index.py

@@ -51,6 +51,8 @@ class AnalyzerProperties(TypedDict):
 
 class VectorIndexProperties(TypedDict):
     Endpoint: str
+    Precision: str
+    DistanceType: str
     IndexName: str
     VectorField: str
     Dimensions: int | str
@@ -70,6 +72,10 @@ def validate_event(event: CustomResourceRequest[VectorIndexProperties]) -> bool:
             raise ValueError("VectorField is required")
         if event["ResourceProperties"]["Dimensions"] is None:
             raise ValueError("Dimensions is required")
+        if event["ResourceProperties"]["Precision"] is None:
+            raise ValueError("Precision is required")
+        if event["ResourceProperties"]["DistanceType"] is None:
+            raise ValueError("DistanceType is required")
         if isinstance(int(event["ResourceProperties"]["Dimensions"]), int) is False:
             raise ValueError("Dimensions must be an integer")
         if event["ResourceProperties"]["MetadataManagement"] is None:
@@ -125,16 +131,19 @@ def connect_opensearch(endpoint: str) -> OpenSearch:
 def create_mapping(
     vector_field: str,
     dimensions: int,
+    precision: str,
+    distance_type: str,
     metadata_management: Sequence[MetadataManagementField],
 ) -> dict:
     mapping = {
         "properties": {
             vector_field: {
                 "type": "knn_vector",
                 "dimension": dimensions,
+                "data_type": precision,
                 "method": {
                     "engine": "faiss",
-                    "space_type": "l2",
+                    "space_type": distance_type,
                     "name": "hnsw",
                     "parameters": {},
                 },
@@ -206,14 +215,16 @@ def handle_create(
     index_name: str,
     vector_field: str,
     dimensions: int,
+    precision: str,
+    distance_type: str,
     metadata_management: Sequence[MetadataManagementField],
     analyzer: AnalyzerProperties | None,
 ):
     if client.indices.exists(index_name):
         raise ValueError(f"Index {index_name} already exists")
 
     try:
-        mapping = create_mapping(vector_field, dimensions, metadata_management)
+        mapping = create_mapping(vector_field, dimensions, precision, distance_type, metadata_management)
         setting = create_setting(analyzer)
         create_index(client, index_name, mapping, setting)
     except Exception as e:
@@ -248,6 +259,8 @@ def on_create(
         event["ResourceProperties"]["IndexName"],
         event["ResourceProperties"]["VectorField"],
         int(event["ResourceProperties"]["Dimensions"]),
+        event["ResourceProperties"]["Precision"],
+        event["ResourceProperties"]["DistanceType"],
         event["ResourceProperties"]["MetadataManagement"],
         event["ResourceProperties"].get("Analyzer", None),
     )