v0.4.2
Added
-
Stickler-Based Evaluation System for Enhanced Comparison Capabilities
- Migrated evaluation service from custom comparison logic to AWS Labs Stickler library for structured object evaluation
- Field Importance Weights: New capability to assign business criticality weights to fields (e.g., shipment ID weight=3.0 vs notes weight=0.5)
- Enhanced Configuration: Added
x-aws-idp-evaluation-*extensions for evaluation configuration - Backward compatible: Maintained API compatibility - all existing code works unchanged
- Enhanced Comparators: Leverages Stickler's optimized comparison algorithms (Exact, Levenshtein, Numeric, Fuzzy, Semantic) with LLM evaluation preserved through custom wrapper
- Better List Matching: Hungarian algorithm via Stickler for optimal list comparisons regardless of order
-
UI: Evaluation Configuration in Document Schema UI
- Added evaluation weight, threshold (with conditional display), and document-level match threshold fields for complete Stickler configuration control
- Added LEVENSHTEIN and HUNGARIAN evaluation methods with auto-populated threshold defaults based on selected method
-
IDP CLI Force Delete All Resources Option
- Added
--force-delete-allflag toidp-cli deletecommand for comprehensive stack cleanup - Post-CloudFormation Cleanup: Analyzes resources after CloudFormation deletion completes to identify retained resources (DELETE_SKIPPED status)
- Use Cases: Complete test environment cleanup, CI/CD pipelines requiring full teardown, cost optimization by removing all retained resources
- Added
Changed
-
Containerized Pattern-1 and Pattern-3 Deployment Pipelines
- Migrated Pattern-1 and Pattern-3 Lambda functions to Docker image deployments (following Pattern-2 approach from v0.3.20)
- Builds and pushes all Lambda images via CodeBuild with automated ECR cleanup
- Increases Lambda package size limit from 250 MB (zip) to 10 GB (Docker image) to accommodate larger dependencies
-
Agent Companion Chat - Chat History Feature
- Added chat history feature from Agent Analysis back into Agent Companion Chat
- Users can now load and view previous chat sessions with full conversation context
- Chat history dropdown displays recent sessions with timestamps and message counts
Fixed
-
Agent Companion Chat - Session Persistence and input control
- Agent Companion Chat in-session memory now persists even when user changes pages
- Prompt input is disabled during active streaming responses to prevent concurrent requests
- Fixed issue where charts in loaded chat history were not displaying
-
GovCloud Template Generation errors
- Fixed CloudFormation deployment error
Fn::GetAtt references undefined resource GraphQLApiwhen deploying GovCloud templates
- Fixed CloudFormation deployment error
-
Example Notebook error fixed
- Example notebooks updated to work with new v0.4.0+ JSON schema
Templates
- us-west-2:
https://s3.us-west-2.amazonaws.com/aws-ml-blog-us-west-2/artifacts/genai-idp/idp-main_0.4.2yaml - us-east-1:
https://s3.us-east-1.amazonaws.com/aws-ml-blog-us-east-1/artifacts/genai-idp/idp-main_0.4.2.yaml - eu-central-1:
https://s3.eu-central-1.amazonaws.com/aws-ml-blog-eu-central-1/artifacts/genai-idp/idp-main_0.4.2.yaml