Merge branch 'develop' v0.7.12

aws-samples · Feb 21, 2025 · 9f89e57 · 9f89e57
2 parents b6bb5da + a75d4e6
commit 9f89e57
Show file tree

Hide file tree

Showing 20 changed files with 875 additions and 875 deletions.
diff --git a/.gitignore b/.gitignore
@@ -5,3 +5,6 @@ build
 /sample-data
 .vscode
 /pca-ui/src/witch/witch.zip
+.DS_Store
+.idea/
+pca-ssm/.DS_Store
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -7,6 +7,21 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.7.12] - 2025-02-21
+
+### Added
+- Added support for Amazon Nova models (amazon.nova-micro-v1:0, amazon.nova-lite-v1:0, amazon.nova-pro-v1:0) - PR #295
+- Updated Anthropic Claude models to the latest 3.5 versions (anthropic.claude-3-5-haiku-20241022-v1:0, anthropic.claude-3-5-sonnet-20241022-v2:0) - PR #295
+- Removed older versions of Amazon Titan and Anthropic Claude models (amazon.titan-text-express-v1, anthropic.claude-v1, anthropic.claude-instant-v1, anthropic.claude-v2) - PR #295
+- Refactored Bedrock calls to use the Converse API eliminating the need for custom model specific payloads - PR #295
+- Refactored all model invocation to use Inference Profiles. This is required for Nova models. It is also applied to Anthropic models for consistency and improved scalability. - PR #295
+- Added adaptive retry configuration to Bedrock api calls to add some tolerance for quota throttling exceptions (at the expense of latency) - PR #295
+
+### Fixed
+- Fixed: Only one PCA stack may be sucessfully deployed in a given account/region due to fixed SSM parameter names - PR #291, #293
+- Fixed: Current publish.sh does not work with ARM chipsets such as the Apple M series. Set x86_64 arch for local build - PR #288
+
+
 ## [0.7.11] - 2024-10-09
 
 ### Added
@@ -275,6 +290,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Initial release
 
 [Unreleased]: https://github.com/aws-samples/amazon-transcribe-post-call-analytics/compare/main...develop
+[0.7.12]: https://github.com/aws-samples/amazon-transcribe-post-call-analytics/releases/tag/v0.7.12
 [0.7.11]: https://github.com/aws-samples/amazon-transcribe-post-call-analytics/releases/tag/v0.7.11
 [0.7.10]: https://github.com/aws-samples/amazon-transcribe-post-call-analytics/releases/tag/v0.7.10
 [0.7.9]: https://github.com/aws-samples/amazon-transcribe-post-call-analytics/releases/tag/v0.7.9

diff --git a/VERSION b/VERSION
@@ -1 +1 @@
-0.7.11
+0.7.12
diff --git a/pca-main-nokendra.template b/pca-main-nokendra.template
@@ -1,6 +1,6 @@
 AWSTemplateFormatVersion: "2010-09-09"
 
-Description: Amazon Transcribe Post Call Analytics - PCA (v0.7.11) (uksb-1sn29lk73, SO9071)
+Description: Amazon Transcribe Post Call Analytics - PCA (v0.7.12) (uksb-1sn29lk73, SO9071)
 
 Parameters:
 
@@ -352,10 +352,10 @@ Parameters:
       - PROD
     Default: PROD
 
-  DatabaseName:
+  GlueDatabaseName:
     Type: String
-    Default: 'pca'
-    Description: Glue catalog database name used to contain tables/views for SQL integration.
+    AllowedPattern: "^$|[a-z0-9_-]+"
+    Description: (Optional) Glue catalog database name in lowercase characters used to contain tables/views for SQL integration.
 
   EnablePcaDashboards:
     Type: String
@@ -384,14 +384,17 @@ Parameters:
 
   GenAIQueryBedrockModelId:
     Type: String
-    Default: anthropic.claude-3-haiku-20240307-v1:0
+    Default: us.amazon.nova-lite-v1:0
     AllowedValues:
       - anthropic.claude-3-haiku-20240307-v1:0
       - anthropic.claude-3-sonnet-20240229-v1:0
-      - amazon.titan-text-express-v1
-      - anthropic.claude-v1
-      - anthropic.claude-instant-v1
-      - anthropic.claude-v2
+      - us.amazon.nova-micro-v1:0
+      - us.amazon.nova-lite-v1:0
+      - us.amazon.nova-pro-v1:0
+      - us.anthropic.claude-3-5-haiku-20241022-v1:0
+      - us.anthropic.claude-3-5-sonnet-20241022-v2:0
+      - eu.anthropic.claude-3-5-sonnet-20240620-v1:0
+      - apac.anthropic.claude-3-5-sonnet-20240620-v1:0
     Description: (Optional) If 'GenAIQuery' is BEDROCK, which Bedrock model to use.
 
   CallSummarization:
@@ -417,14 +420,17 @@ Parameters:
 
   SummarizationBedrockModelId:
     Type: String
-    Default: anthropic.claude-3-haiku-20240307-v1:0
+    Default: us.amazon.nova-lite-v1:0
     AllowedValues:
       - anthropic.claude-3-haiku-20240307-v1:0
       - anthropic.claude-3-sonnet-20240229-v1:0
-      - amazon.titan-text-express-v1
-      - anthropic.claude-v1
-      - anthropic.claude-instant-v1
-      - anthropic.claude-v2
+      - us.amazon.nova-micro-v1:0
+      - us.amazon.nova-lite-v1:0
+      - us.amazon.nova-pro-v1:0
+      - us.anthropic.claude-3-5-haiku-20241022-v1:0
+      - us.anthropic.claude-3-5-sonnet-20241022-v2:0
+      - eu.anthropic.claude-3-5-sonnet-20240620-v1:0
+      - apac.anthropic.claude-3-5-sonnet-20240620-v1:0
     Description: (Optional) If 'CallSummarization' is BEDROCK, which Bedrock model to use.
 
   TestBedrockModelId:
@@ -495,7 +501,7 @@ Metadata:
                 default: Advanced analytics and reporting
               Parameters:
                   - EnablePcaDashboards
-                  - DatabaseName
+                  - GlueDatabaseName
             - Label:
                 default: S3 Bucket Names and Retention Policy
               Parameters:
@@ -673,6 +679,19 @@ Resources:
           - ServerSideEncryptionByDefault:
               SSEAlgorithm: AES256
 
+  ########################################################
+  # Glue Database
+  ########################################################
+
+  PCADatabase:
+    Type: AWS::Glue::Database
+    Properties:
+      CatalogId: !Ref AWS::AccountId
+      DatabaseInput:
+        Name: !Ref GlueDatabaseName
+        Description: Post Call Analytics
+        LocationUri: !Sub 's3://${OutputBucket}/${OutputBucketParsedResults}/'
+
   ########################################################
   # If Bedrock Models are selected, verify that Bedrock 
   # is available in the region, and that models are enabled
@@ -701,6 +720,12 @@ Resources:
                 Resource:
                   - !Sub "arn:${AWS::Partition}:bedrock:*::foundation-model/*"
                   - !Sub "arn:${AWS::Partition}:bedrock:*:${AWS::AccountId}:custom-model/*"
+                  - !Sub "arn:aws:bedrock:${AWS::Region}:${AWS::AccountId}:inference-profile/*"
+              - Effect: Allow
+                Action:
+                  - "bedrock:GetInferenceProfile"
+                Resource:
+                  - !Sub "arn:aws:bedrock:${AWS::Region}:${AWS::AccountId}:inference-profile/*"
           PolicyName: BedrockPolicy
 
   TestBedrockModelFunction:
@@ -723,76 +748,38 @@ Resources:
           subprocess.check_call([sys.executable, '-m', 'pip', 'install', '--target', '/tmp', 'boto3'])
           sys.path.insert(0,'/tmp')
           import boto3
-
+          from botocore.config import Config
+
           # Defaults
           AWS_REGION = os.environ["AWS_REGION_OVERRIDE"] if "AWS_REGION_OVERRIDE" in os.environ else os.environ["AWS_REGION"]
           ENDPOINT_URL = os.environ.get("ENDPOINT_URL", f'https://bedrock-runtime.{AWS_REGION}.amazonaws.com')
           DEFAULT_MAX_TOKENS = 128
 
-          def get_request_body(modelId, parameters, prompt):
-              provider = modelId.split(".")[0]
-              request_body = None
-              if provider == "anthropic":
-                  if 'claude-3' in modelId:
-                      request_body = {
-                          "max_tokens": DEFAULT_MAX_TOKENS,
-                          "messages": [{"role": "user", "content": prompt}],
-                          "anthropic_version": "bedrock-2023-05-31"
-                      }
-                  else:    
-                      request_body = {
-                          "prompt": prompt,
-                          "max_tokens_to_sample": DEFAULT_MAX_TOKENS
-                      } 
-
-                  request_body.update(parameters)
-              elif provider == "ai21":
-                  request_body = {
-                      "prompt": prompt,
-                      "maxTokens": DEFAULT_MAX_TOKENS
-                  }
-                  request_body.update(parameters)
-              elif provider == "amazon":
-                  textGenerationConfig = {
-                      "maxTokenCount": DEFAULT_MAX_TOKENS
-                  }
-                  textGenerationConfig.update(parameters)
-                  request_body = {
-                      "inputText": prompt,
-                      "textGenerationConfig": textGenerationConfig
+          def get_generate_text(response):
+            return response["output"]["message"]["content"][0]["text"]
+
+          def call_llm(prompt, modelId):
+              client = boto3.client(
+                service_name='bedrock-runtime', 
+                region_name=AWS_REGION, 
+                endpoint_url=ENDPOINT_URL, 
+                config=Config(retries={'max_attempts': 50, 'mode': 'adaptive'})
+              )
+
+              message = {
+                  'role': 'user',
+                  'content': [{'text': prompt}]
+              }
+
+              response = client.converse(
+                  modelId=modelId,
+                  messages=[message],
+                  inferenceConfig={
+                      'maxTokens': DEFAULT_MAX_TOKENS,
+                      'temperature': 0
                   }
-              else:
-                  raise Exception("Unsupported provider: ", provider)
-              return request_body
-
-          def get_generate_text(modelId, response):
-              provider = modelId.split(".")[0]
-              generated_text = None
-              if provider == "anthropic":
-                  if 'claude-3' in modelId:
-                      response_raw = json.loads(response.get("body").read().decode())
-                      generated_text = response_raw.get('content')[0].get('text')
-
-                  else:
-                      response_body = json.loads(response.get("body").read().decode())
-                      generated_text = response_body.get("completion")
-              elif provider == "ai21":
-                  response_body = json.loads(response.get("body").read())
-                  generated_text = response_body.get("completions")[0].get("data").get("text")
-              elif provider == "amazon":
-                  response_body = json.loads(response.get("body").read())
-                  generated_text = response_body.get("results")[0].get("outputText")
-              else:
-                  raise Exception("Unsupported provider: ", provider)
-              return generated_text
-
-          def call_llm(parameters, prompt):
-              modelId = parameters.pop("modelId")
-              body = get_request_body(modelId, parameters, prompt)
-              print("ModelId", modelId, "-  Body: ", body)
-              client = boto3.client(service_name='bedrock-runtime', region_name=AWS_REGION, endpoint_url=ENDPOINT_URL)
-              response = client.invoke_model(body=json.dumps(body), modelId=modelId, accept='application/json', contentType='application/json')
-              generated_text = get_generate_text(modelId, response)
+              )
+              generated_text = get_generate_text(response)
               return generated_text
 
           def lambda_handler(event, context):
@@ -808,12 +795,8 @@ Resources:
                       # Test LLMModel
                       llmModelId = event['ResourceProperties'].get('LLMModelId', '')
                       modelId = llmModelId
-                      parameters = {
-                          "modelId": modelId,
-                          "temperature": 0
-                      }
                       print(f"Testing {modelId}")
-                      call_llm(parameters, prompt)            
+                      call_llm(prompt, modelId)               
                   except Exception as e:
                       status = cfnresponse.FAILED
                       reason = f"Exception thrown testing ModelId='{modelId}'. Check that Amazon Bedrock is available in your region, and that model is activated in your Amazon Bedrock account - {e}"
@@ -848,6 +831,7 @@ Resources:
     Properties:
       TemplateURL: pca-ssm/cfn/ssm.template
       Parameters:
+        StackName: !Ref 'AWS::StackName'
         BulkUploadBucketName:
           !If
           - ShouldCreateBulkUploadBucket
@@ -911,13 +895,15 @@ Resources:
           - 'None'
           - 'None'
         # WebUri: !GetAtt PCAUI.Outputs.WebUri
-        DatabaseName: !Ref DatabaseName
+        DatabaseName: !Ref PCADatabase
 
   PythonUtilsLayer:
     DependsOn: SSM
     Type: AWS::CloudFormation::Stack
     Properties:
       TemplateURL: pca-server/cfn/lib/python-utilities.template
+      Parameters:
+        ParentStackName: !Ref 'AWS::StackName'
 
   BedrockBoto3Layer:
     Type: AWS::CloudFormation::Stack
@@ -944,6 +930,8 @@ Resources:
     Properties:
       TemplateURL: pca-server/cfn/pca-server.template
       Parameters:
+        DatabaseName: !Ref PCADatabase
+        ParentStackName: !Ref 'AWS::StackName'
         ffmpegDownloadUrl: !Ref ffmpegDownloadUrl
         CallSummarization: !Ref CallSummarization
         SummarizationBedrockModelId: !Ref SummarizationBedrockModelId
@@ -966,6 +954,7 @@ Resources:
     Properties:
       TemplateURL: pca-ui/cfn/pca-ui.template
       Parameters:
+        ParentStackName: !Ref 'AWS::StackName'
         AdminUsername: !Ref AdminUsername
         AdminEmail: !Ref AdminEmail
         AllowedSignUpEmailDomain: !Ref AllowedSignUpEmailDomain
@@ -1005,7 +994,7 @@ Resources:
     Properties:
       TemplateURL: build/pca-dashboards.yaml
       Parameters:
-        PcaGlueCatalogDatabaseName: !Ref DatabaseName
+        PcaGlueCatalogDatabaseName: !Ref PCADatabase
         PcaOutputBucket: 
           !If
           - ShouldCreateOutputBucket
@@ -1023,6 +1012,7 @@ Resources:
     Properties:
       TemplateURL: pca-samples/copy-samples.template
       Parameters:
+        ParentStackName: !Ref 'AWS::StackName'
         DependsOnPCAServer: !Ref PCAServer
         DependsOnPCAUI: !If [ShouldEnableGui, !Ref PCAUI, !Ref AWS::NoValue]
         DependsOnPCADashboards: !If