Add enableBedrockGlobalInference

README.md

@@ -568,12 +568,22 @@ The bot store feature allows users to share and discover custom bots. You can co
 - **enableBotStoreReplicas**: Controls whether standby replicas are enabled for the OpenSearch Serverless collection used by bot store (default: `false`). Setting it to `true` improves availability but increases costs, while `false` reduces costs but may affect availability.
   > **Important**: You can't update this property after the collection is already created. If you attempt to modify this property, the collection continues to use the original value.
 
-### Cross-region inference
+### Cross-region and Global inference
 
-[Cross-region inference](https://docs.aws.amazon.com/bedrock/latest/userguide/inference-profiles-support.html) allows Amazon Bedrock to dynamically route model inference requests across multiple AWS regions, enhancing throughput and resilience during peak demand periods. To configure, edit `cdk.json`.
+[Cross-region and Global inference](https://docs.aws.amazon.com/bedrock/latest/userguide/inference-profiles-support.html)
+allows Amazon Bedrock to dynamically route model inference requests across
+multiple AWS regions, enhancing throughput and resilience during peak demand
+periods. Global inference routes the requests to the optimal region based on
+latency and availability anywhere in the world, while cross-region inference
+routes requests within the same AWS region, for example, within the US. Some
+SCPs may restrict on or the other or both and therefore you can configure them
+independently. By default both are enabled.
+
+To configure change the following settings in `cdk.json` or `parameters.ts`:
 
 ```json
-"enableBedrockCrossRegionInference": true
+"enableBedrockGlobalInference": false,
+"enableBedrockCrossRegionInference": false,
 ```
 
 ### Lambda SnapStart

backend/app/bedrock.py

@@ -38,6 +38,9 @@
 logger.setLevel(logging.INFO)
 
 BEDROCK_REGION = os.environ.get("BEDROCK_REGION", "us-east-1")
+ENABLE_BEDROCK_GLOBAL_INFERENCE = (
+    os.environ.get("ENABLE_BEDROCK_GLOBAL_INFERENCE", "false") == "true"
+)
 ENABLE_BEDROCK_CROSS_REGION_INFERENCE = (
     os.environ.get("ENABLE_BEDROCK_CROSS_REGION_INFERENCE", "false") == "true"
 )
@@ -979,23 +982,25 @@ def get_regional_inference_profile_id(
 
 def get_model_id(
     model: type_model_name,
+    enable_global: bool = ENABLE_BEDROCK_GLOBAL_INFERENCE,
     enable_cross_region: bool = ENABLE_BEDROCK_CROSS_REGION_INFERENCE,
     bedrock_region: str = BEDROCK_REGION,
 ) -> str:
     base_model_id = BASE_MODEL_IDS.get(model)
     if not base_model_id:
         raise ValueError(f"Unsupported model: {model}")
 
-    if enable_cross_region:
-        # 1. First, try to use global inference profile if available
+    # 1. First, try to use global inference profile if enabled and available
+    if enable_global:
         global_profile_id = get_global_inference_profile_id(model, bedrock_region)
         if global_profile_id:
             logger.info(
                 f"Using global inference profile: {global_profile_id} for model '{model}'"
             )
             return global_profile_id
 
-        # 2. Fallback to regional cross-region inference profile if available
+    # 2. Fallback to regional cross-region inference profile if enabled and available
+    if enable_cross_region:
         regional_profile_id = get_regional_inference_profile_id(model, bedrock_region)
         if regional_profile_id:
             logger.info(
@@ -1007,6 +1012,6 @@ def get_model_id(
                 f"Region '{bedrock_region}' does not support cross-region inference for model '{model}'."
             )
 
-    # 3. No cross-region inference
+    # 3. Use standalone model (no global or cross-region inference)
     logger.info(f"Using local model ID: {base_model_id} for model '{model}'")
     return base_model_id

cdk/bin/bedrock-chat.ts

@@ -94,6 +94,7 @@ const chat = new BedrockChatStack(
     selfSignUpEnabled: params.selfSignUpEnabled,
     documentBucket: bedrockRegionResources.documentBucket,
     enableRagReplicas: params.enableRagReplicas,
+    enableBedrockGlobalInference: params.enableBedrockGlobalInference,
     enableBedrockCrossRegionInference: params.enableBedrockCrossRegionInference,
     enableLambdaSnapStart: params.enableLambdaSnapStart,
     alternateDomainName: params.alternateDomainName,

cdk/lib/bedrock-chat-stack.ts

@@ -45,6 +45,7 @@ export interface BedrockChatStackProps extends StackProps {
   readonly enableIpV6: boolean;
   readonly documentBucket: Bucket;
   readonly enableRagReplicas: boolean;
+  readonly enableBedrockGlobalInference: boolean;
   readonly enableBedrockCrossRegionInference: boolean;
   readonly enableLambdaSnapStart: boolean;
   readonly enableBotStore: boolean;
@@ -219,6 +220,8 @@ export class BedrockChatStack extends cdk.Stack {
       bedrockCustomBotProject: bedrockCustomBotCodebuild.project,
       usageAnalysis,
       largeMessageBucket,
+      enableBedrockGlobalInference:
+        props.enableBedrockGlobalInference,
       enableBedrockCrossRegionInference:
         props.enableBedrockCrossRegionInference,
       enableLambdaSnapStart: props.enableLambdaSnapStart,
@@ -269,6 +272,8 @@ export class BedrockChatStack extends cdk.Stack {
       bedrockRegion: props.bedrockRegion,
       largeMessageBucket,
       documentBucket: props.documentBucket,
+      enableBedrockGlobalInference:
+        props.enableBedrockGlobalInference,
       enableBedrockCrossRegionInference:
         props.enableBedrockCrossRegionInference,
       enableLambdaSnapStart: props.enableLambdaSnapStart,

cdk/lib/constructs/api.ts

@@ -39,6 +39,7 @@ export interface ApiProps {
   readonly apiPublishProject: codebuild.IProject;
   readonly bedrockCustomBotProject: codebuild.IProject;
   readonly usageAnalysis?: UsageAnalysis;
+  readonly enableBedrockGlobalInference: boolean;
   readonly enableBedrockCrossRegionInference: boolean;
   readonly enableLambdaSnapStart: boolean;
   readonly openSearchEndpoint?: string;
@@ -262,6 +263,8 @@ export class Api extends Construct {
           props.usageAnalysis?.ddbExportTable.tableName || "",
         USAGE_ANALYSIS_WORKGROUP: props.usageAnalysis?.workgroupName || "",
         USAGE_ANALYSIS_OUTPUT_LOCATION: usageAnalysisOutputLocation,
+        ENABLE_BEDROCK_GLOBAL_INFERENCE:
+          props.enableBedrockGlobalInference.toString(),
         ENABLE_BEDROCK_CROSS_REGION_INFERENCE:
           props.enableBedrockCrossRegionInference.toString(),
         GLOBAL_AVAILABLE_MODELS: props.globalAvailableModels 

cdk/lib/constructs/websocket.ts

@@ -23,6 +23,7 @@ export interface WebSocketProps {
   readonly websocketSessionTable: ITable;
   readonly largeMessageBucket: s3.IBucket;
   readonly accessLogBucket?: s3.Bucket;
+  readonly enableBedrockGlobalInference: boolean;
   readonly enableBedrockCrossRegionInference: boolean;
   readonly enableLambdaSnapStart: boolean;
 }
@@ -126,6 +127,8 @@ export class WebSocket extends Construct {
         LARGE_MESSAGE_BUCKET: props.largeMessageBucket.bucketName,
         LARGE_PAYLOAD_SUPPORT_BUCKET: largePayloadSupportBucket.bucketName,
         WEBSOCKET_SESSION_TABLE_NAME: props.websocketSessionTable.tableName,
+        ENABLE_BEDROCK_GLOBAL_INFERENCE:
+          props.enableBedrockGlobalInference.toString(),
         ENABLE_BEDROCK_CROSS_REGION_INFERENCE:
           props.enableBedrockCrossRegionInference.toString(),
       },

cdk/lib/utils/parameter-models.ts

@@ -32,6 +32,7 @@ const BaseParametersSchema = z.object({
 
   // Bedrock configuration
   bedrockRegion: z.string().default("us-east-1"),
+  enableBedrockGlobalInference: z.boolean().default(true),
   enableBedrockCrossRegionInference: z.boolean().default(true),
 });
 
@@ -232,6 +233,9 @@ export function resolveBedrockChatParameters(
       "publishedApiAllowedIpV6AddressRanges"
     ),
     enableRagReplicas: app.node.tryGetContext("enableRagReplicas"),
+    enableBedrockGlobalInference: app.node.tryGetContext(
+      "enableBedrockGlobalInference"
+    ),
     enableBedrockCrossRegionInference: app.node.tryGetContext(
       "enableBedrockCrossRegionInference"
     ),
@@ -303,6 +307,9 @@ export function resolveApiPublishParameters(): ApiPublishParameters {
     envName: getEnvVar("ENV_NAME"),
     envPrefix: getEnvVar("ENV_PREFIX"),
     bedrockRegion: getEnvVar("BEDROCK_REGION"),
+    enableBedrockGlobalInference: getEnvVar(
+      "ENABLE_BEDROCK_GLOBAL_INFERENCE"
+    ),
     enableBedrockCrossRegionInference: getEnvVar(
       "ENABLE_BEDROCK_CROSS_REGION_INFERENCE"
     ),