v0.3.11

[0.3.11]

Added

• Chat with Document now available at the bottom of the each Document Detail page.
• Anthropic Claude Opus 4.1 model available in configuration for all document processing steps
• Browser tab icon now features a blue background with a white "IDP"
• Experimental new classification method - multimodalPageBoundaryClassification - for detecting section boundaries during page level classification.

v0.3.10

[0.3.10]

Added

• Agent Analysis Feature for Natural Language Document Analytics

  • Added integrated AI-powered analytics agent that enables natural language querying of processed document data
  • Key Capabilities: Convert natural language questions to SQL queries, generate interactive visualizations and tables, explore database schema automatically
  • Secure Architecture: All Python code execution happens in isolated AWS Bedrock AgentCore sandboxes, not in Lambda functions
  • Multi-Tool Agent System: Database discovery tool for schema exploration, Athena query tool for SQL execution, secure code sandbox for data transfer, Python visualization tool for charts and tables
  • Example Use Cases: Query document processing volumes and trends, analyze confidence scores and extraction accuracy, explore document classifications and content patterns, generate custom charts and data tables
  • Sample W2 Test Data: Includes 20 synthetic W2 tax documents for testing analytics capabilities
  • Configurable Models: Supports multiple AI models including Claude 3.7 Sonnet (default), Claude 3.5 Sonnet, Nova Pro/Lite, and Haiku
  • Web UI Integration: Accessible through "Document Analytics" section with real-time progress display and query history

• Automatic Glue Table Creation for Document Sections

  • Added automatic creation of AWS Glue tables for each document section type (classification) during processing
  • Tables are created dynamically when new section types are encountered, eliminating manual table creation
  • Consistent lowercase naming convention for tables ensures compatibility with case-sensitive S3 paths
  • Tables are configured with partition projection for efficient date-based queries without manual partition management
  • Automatic schema evolution - tables update when new fields are detected in extraction results

v0.3.9

[0.3.9]

Added

• Optional Permissions Boundary Support for Enterprise Deployments
  • Added PermissionsBoundaryArn parameter to all CloudFormation templates for organizations with Service Control Policies (SCPs) requiring permissions boundaries
  • Comprehensive support for both explicit IAM roles and implicit roles created by AWS SAM functions and statemachines`
  • Conditional implementation ensures backward compatibility - when no permissions boundary is provided, roles deploy normally

Added

• IDP Configuration and Prompting Best Practices documentation doc

Changed

• Updated lending_package.pdf sample with more realistic driver's license image

Fixed

• Issue #27 - removed idp_common bedrock client region default to us-west-2 - PR #28

v0.3.8

[0.3.8]

Added

• Lending Package Configuration Support for Pattern-2

  • Added new lending-package-sample configuration to Pattern-2, providing comprehensive support for lending and financial document processing workflows
  • New default configuration for Pattern-2 stack deployments, optimized for loan applications, mortgage processing, and financial verification documents
  • Previous rvl-cdip-sample configuration remains available by selecting rvl-cdip-package-sample for the Pattern2Configuration parameter when deploying or updating stacks

• Text Confidence View for Document Pages

  • Added support for displaying OCR text confidence data through new TextConfidenceUri field
  • New "Text Confidence View" option in the UI pages panel alongside existing Markdown and Text views
  • Fixed issues with view persistence - Text Confidence View button now always visible with appropriate messaging when content unavailable
  • Fixed view toggle behavior - switching between views no longer closes the viewer window
  • Reordered view buttons to: Markdown View, Text Confidence View, Text View for better user experience

• Enhanced OCR DPI Configuration for PDF files

  • DPI for PDF image conversion is now configurable in the configuration editor under OCR image processing settings
  • Default DPI improved from 96 to 150 DPI for better default quality and OCR accuracy
  • Configurable through Web UI without requiring code changes or redeployment

Changed

• Converted text confidence data format from JSON to markdown table for improved readability and reduced token usage

  • Removed unnecessary "page_count" field
  • Changed "text_blocks" array to "text" field containing a markdown table with Text and Confidence columns
  • Reduces prompt size for assessment service while improving UI readability
  • OCR confidence values now rounded to 1 decimal point (e.g., 99.1, 87.3) for cleaner display
  • Markdown table headers now explicitly left-aligned using |:-----|:-----------| format for consistent appearance

• Simplified OCR Service Initialization

  • OCR service now accepts a single config dictionary parameter for cleaner, more consistent API
  • Aligned with classification service pattern for better consistency across IDP services
  • Backward compatibility maintained - old parameter pattern still supported with deprecation warning
  • Updated all lambda functions and notebooks to use new simplified pattern

• Removed fixed image target_height and target_width from default configurations, so images are processed in original resolution by default.

• Updated Default Configuration for Pattern1 and Pattern2

  • Changed default configuration for new stacks from "default" to "lending-package-sample" for both Pattern1 and Pattern2
  • Maintains backward compatibility for stack updates by keeping the parameter value "default" mapped to the rvl-cdip-sample for pattern-2.

• Reduce assessment step costs

  • Default model for granular assessment is now us.amazon.nova-lite-v1:0 - experimentation recommended
  • Improved placement of <> tags in assessment prompt to improve utilization of prompt caching

Fixed

• Fixed Image Resizing Behavior for High-Resolution Documents
  • Fixed issue where empty strings in image configuration were incorrectly resizing images to default 951x1268 pixels instead of preserving original resolution
  • Empty strings ("") in target_width and target_height configuration now preserve original document resolution for maximum processing accuracy
• Fixed issue where PNG files were being unnecessarily converted to JPEG format and resized to lower resolution with lost quality
• Fixed issue where PNG and JPG image files were not rendering inline in the Document Details page
• Fixed issue where PDF files were being downloaded instead of displayed inline
• Fixed pricing data for cacheWrite tokens for Amazon Nova models to resolve innacurate cost estimation in UI.

v0.3.7

[0.3.7]

Added

• Document Process Flow Visualization

  • Added interactive visualization of Step Functions workflow execution for document processing
  • Visual representation of processing steps with status indicators and execution details
  • Detailed step information including inputs, outputs, and error messages
  • Timeline view showing chronological execution of all processing steps
  • Auto-refresh capability for monitoring active executions in real-time
  • Support for Map state visualization with iteration details
  • Error diagnostics with detailed error messages for troubleshooting
  • Automatic selection of failed steps for quick issue identification

• Granular Assessment Service for Scalable Confidence Evaluation

  • New granular assessment approach that breaks down assessment into smaller, focused tasks for improved accuracy and performance
  • Key Benefits: Better accuracy through focused prompts, cost optimization via prompt caching, reduced latency through parallel processing, and scalability for complex documents
  • Task Types: Simple batch tasks (groups 3-5 simple attributes), group tasks (individual group attributes), and list item tasks (individual list items for maximum accuracy)
  • Configuration: Configurable batch sizes (simple_batch_size, list_batch_size) and parallel processing (max_workers) for performance tuning
  • Prompt Caching: Leverages LLM caching capabilities with cached base content (document context, images, OCR data) and dynamic task-specific content
  • Use Cases: Ideal for bank statements with hundreds of transactions, documents with 10+ attributes, complex nested structures, and performance-critical scenarios
  • Backward Compatibility: Maintains same interface as standard assessment service with seamless migration path
  • Enhanced Documentation: Comprehensive documentation in docs/assessment.md and example notebooks for both standard and granular approaches

• Reporting Database now has Document Sections Tables to enable querying across document fields

  • Added comprehensive document sections storage system that automatically creates tables for each section type (classification)
  • Dynamic Table Creation: AWS Glue Crawler automatically discovers new section types and creates corresponding tables (e.g., invoice, receipt, bank_statement)
  • Configurable Crawler Schedule: Support for manual, every 15 minutes, hourly, or daily (default) crawler execution via DocumentSectionsCrawlerFrequency parameter
  • Partitioned Storage: Data organized by section type and date for efficient querying with Amazon Athena

• Partition Projections for Evaluation and Metering tables

  • Automated Partition Management: Eliminates need for MSCK REPAIR TABLE operations with projection-based partition discovery
  • Performance Benefits: Athena can efficiently prune partitions based on date ranges without manual partition loading
  • Backward Compatibility Warning: The partition structure change from year=2024/month=03/day=15/ to date=2024-03-15/ means that data saved in the evaluation or metering tables prior to v0.3.7 will not be visible in Athena queries after updating. To retain access to historical data, you can either:
    • Manually reorganize existing S3 data to match the new partition structure
    • Create separate Athena tables pointing to the old partition structure for historical queries

• Optimize the classification process for single class configurations in Pattern-2

  • Detects when only a single document class is defined in the configuration
  • Automatically classifies all document pages as that single class
  • Creates a single section containing all pages
  • Bypasses the backend service calls (Bedrock or SageMaker) completely
  • Logs an INFO message indicating the optimization is active

• Skip the extraction process for classes with no attributes in Pattern 2/3

  • Add early detection logic in extraction class to check for empty/missing attributes
  • Return zero metering data and empty JSON results when no attributes defined

• Enhanced State Machine Optimization for Very Large Documents

  • Improved document compression to store only section IDs rather than full section objects
  • Modified state machine workflow to eliminate nested result structures and reduce payload size
  • Added OutputPath filtering to remove intermediate results from state machine execution
  • Streamlined assessment step to replace extraction results instead of nesting them
  • Resolves "size exceeding the maximum number of bytes service limit" errors for documents with 500+ pages

• Criteria Validation Service Class

  • New document validation service that evaluates documents against dynamic business rules using Large Language Models (LLMs)
  • Key Capabilities: Dynamic business rules configuration, asynchronous processing with concurrent criteria evaluation, intelligent text chunking for large documents, multi-file processing with summarization, comprehensive cost and performance tracking
  • Primary Use Cases: Healthcare prior authorization workflows, compliance validation, business rule enforcement, quality assurance, and audit preparation
  • Architecture Features: Seamless integration with IDP pipeline using common Bedrock client, unified metering with automatic token usage tracking, S3 operations using standardized file operations, configuration compatibility with existing IDP config system
  • Advanced Features: Configurable criteria questions without code changes, robust error handling with graceful degradation, Pydantic-based input/output validation with automatic data cleaning, comprehensive timing metrics and token usage tracking
  • Limitation: Python idp_common support only, not yet implemented within deployed pattern workflows.

Changed

• Default behavior for image attachment in Pattern-2 and Pattern3
  • If the prompt contains a {DOCUMENT_IMAGE} placeholder, keep the current behavior (insert image at placeholder)
  • If the prompt does NOT contain a {DOCUMENT_IMAGE} placeholder, do NOT attach the image at all
  • Previously, if the (classification or extraction) prompt did NOT contain a {DOCUMENT_IMAGE} placeholder, the image was appended at the end of the content array anyway
• Modified default assessment prompt for token efficiency
  • Removed confidence_reason from output to avoid consuming unnecessary output tokens
  • Refactored task_prompt layout to improve <> placement for efficiency when granular mode is enabled or disabled
• Enhanced .clinerules with comprehensive memory bank workflows
  • Enhanced Plan Mode workflow with requirements gathering, reasoning, and user approval loop

Fixed

• Fixed UI list deletion issue where empty lists were not saved correctly - #18
• Improve structure and clarity for idp_common Python package documentation
• Improved UI in View/Edit Configuration to clarify that Class and Attribute descriptions are used in the classification and extraction prompts
• Automate UI updates for field "HITL (A2I) Status" in the Document list and document details section.
• Fixed image display issue in PagesPanel where URLs containing special characters (commas, spaces) would fail to load by properly URL-encoding S3 object keys in presigned URL generation

v0.3.6

[0.3.6]

Fixed

• Update Athena/Glue table configuration to use Parquet format instead of JSON #20
• Cloudformation Error when Changing Evaluation Bucket Name #19

Added

• Extended Document Format Support in OCR Service
  • Added support for processing additional document formats beyond PDF and images:
    • Plain text (.txt) files with automatic pagination for large documents
    • CSV (.csv) files with table visualization and structured output
    • Excel workbooks (.xlsx, .xls) with multi-sheet support (each sheet as a page)
    • Word documents (.docx, .doc) with text extraction and visual representation
  • Key Features:
    • Consistent processing model across all document formats
    • Standard page image generation for all formats
    • Structured text output in formats compatible with existing extraction pipelines
    • Confidence metrics for all document types
    • Automatic format detection from file content and extension
  • Implementation Details:
    • Format-specific processing strategies for optimal results
    • Enhanced text rendering for plain text documents
    • Table visualization for CSV and Excel data
    • Word document paragraph extraction with formatting preservation
    • S3 storage integration matching existing PDF processing workflow

v0.3.5

[0.3.5]

Added

• Human-in-the-Loop (HITL) Support - Pattern 1
  • Added comprehensive Human-in-the-Loop review capabilities using Amazon SageMaker Augmented AI (A2I)
  • Key Features:
    • Automatic triggering when extraction confidence falls below configurable threshold
    • Integration with SageMaker A2I Review Portal for human validation and correction
    • Configurable confidence threshold through Web UI Portal Configuration tab (0.0-1.0 range)
    • Seamless result integration with human-verified data automatically updating source results
  • Workflow Integration:
    • HITL tasks created automatically when confidence thresholds are not met
    • Reviewers can validate correct extractions or make necessary corrections through the Review Portal
    • Document processing continues with human-verified data after review completion
  • Configuration Management:
    • EnableHITL parameter for feature toggle
    • Confidence threshold configurable via Web UI without stack redeployment
    • Support for existing private workforce work teams via input parameter
  • CloudFormation Output: Added SageMakerA2IReviewPortalURL for easy access to review portal
  • Known Limitations: Current A2I version cannot provide direct hyperlinks to specific document tasks; template updates require resource recreation
• Document Compression for Large Documents - all patterns
  • Added automatic compression support to handle large documents and avoid exceeding Step Functions payload limits (256KB)
  • Key Features:
    • Automatic compression (default trigger threshold of 0KB enables compression by default)
    • Transparent handling of both compressed and uncompressed documents in Lambda functions
    • Temporary S3 storage for compressed document state with automatic cleanup via lifecycle policies
  • New Utility Methods:
    • Document.load_document(): Automatically detects and decompresses document input from Lambda events
    • Document.serialize_document(): Automatically compresses large documents for Lambda responses
    • Document.compress() and Document.decompress(): Compression/decompression methods
  • Lambda Function Integration: All relevant Lambda functions updated to use compression utilities
  • Resolves Step Functions Errors: Eliminates "result with a size exceeding the maximum number of bytes service limit" errors for large multi-page documents
• Multi-Backend OCR Support - Pattern 2 and 3
  • Textract Backend (default): Existing AWS Textract functionality
  • Bedrock Backend: New LLM-based OCR using Claude/Nova models
  • None Backend: Image-only processing without OCR
• Bedrock OCR Integration - Pattern 2 and 3
  • Customizable system and task prompts for OCR optimization
  • Better handling of complex documents, tables, and forms
  • Layout preservation capabilities
• Image Preprocessing - Pattern 2 and 3
  • Adaptive Binarization: Improves OCR accuracy on documents with:
    • Uneven lighting or shadows
    • Low contrast text
    • Background noise or gradients
  • Optional feature with configurable enable/disable
• YAML Parsing Support for LLM Responses - Pattern 2 and 3
  • Added comprehensive YAML parsing capabilities to complement existing JSON parsing functionality
  • New extract_yaml_from_text() function with robust multi-strategy YAML extraction:
    • YAML in yaml and yml code blocks
    • YAML with document markers (---)
    • Pattern-based YAML detection using indentation and key indicators
  • New detect_format() function for automatic format detection returning 'json', 'yaml', or 'unknown'
  • New unified extract_structured_data_from_text() wrapper function that automatically detects and parses both JSON and YAML formats
  • Token Efficiency: YAML typically uses 10-30% fewer tokens than equivalent JSON due to more compact syntax
  • Service Integration: Updated classification service to use the new unified parsing function with automatic fallback between formats
  • Comprehensive Testing: Added 39 new unit tests covering all YAML extraction strategies, format detection, and edge cases
  • Backward Compatibility: All existing JSON functionality preserved unchanged, new functionality is purely additive
  • Intelligent Fallback: Robust fallback mechanism handles cases where preferred format fails (e.g., JSON requested as YAML falls back to JSON)
  • Production Ready: Handles malformed content gracefully, comprehensive error handling and logging
  • Example Notebook: Added notebooks/examples/step3_extraction_using_yaml.ipynb demonstrating YAML-based extraction with automatic format detection and token efficiency benefits

Fixed

• Enhanced JSON Extraction from LLM Responses (Issue #16)
  • Modularized duplicate _extract_json() functions across classification, extraction, summarization, and assessment services into a common extract_json_from_text() utility function
  • Improved multi-line JSON handling with literal newlines in string values that previously caused parsing failures
  • Added robust JSON validation and multiple fallback strategies for better extraction reliability
  • Enhanced string parsing with proper escape sequence handling for quotes and newlines
  • Added comprehensive unit tests covering various JSON formats including multi-line scenarios

v0.3.4

[0.3.4]

Added

• Configurable Image Processing and Enhanced Resizing Logic

  • Improved Image Resizing Algorithm: Enhanced aspect-ratio preserving scaling that only downsizes when necessary (scale factor < 1.0) to prevent image distortion
  • Configurable Image Dimensions: All processing services (Assessment, Classification, Extraction, OCR) now support configurable image dimensions through configuration with default 951×1268 resolution
  • Service-Specific Image Optimization: Each service can use optimal image dimensions for performance and quality tuning
  • Enhanced OCR Service: Added configurable DPI for PDF-to-image conversion (default: 300) and optional image resizing with dual image strategy (stores original high-DPI images while using resized images for processing)
  • Runtime Configuration: No code changes needed to adjust image processing - all configurable through service configuration
  • Backward Compatibility: Default values maintain existing behavior with no immediate action required for existing deployments

• Enhanced Configuration Management

  • Save as Default: New button to save current configuration as the new default baseline with confirmation modal and version upgrade warnings
  • Export Configuration: Export current configuration to local files in JSON or YAML format with customizable filename
  • Import Configuration: Import configuration from local JSON or YAML files with automatic format detection and validation
  • Enhanced Lambda resolver with deep merge functionality for proper default configuration updates
  • Automatic custom configuration reset when saving as default to maintain clean state

• Nested Attribute Groups and Lists Support

  • Enhanced document configuration schema to support complex nested attribute structures with three attribute types:
    • Simple attributes: Single-value extractions (existing behavior)
    • Group attributes: Nested object structures with sub-attributes (e.g., address with street, city, state)
    • List attributes: Arrays with item templates containing multiple attributes per item (e.g., transactions with date, amount, description)
  • Web UI Enhancements: Configuration editor now supports viewing and editing nested attribute structures with proper validation
  • Extraction Service Updates: Enhanced {ATTRIBUTE_NAMES_AND_DESCRIPTIONS} placeholder processing to generate formatted prompts for nested structures
  • Assessment Service Enhancements: Added support for nested structure confidence evaluation with recursive processing of group and list attributes, including proper confidence threshold application from configuration
  • Evaluation Service Improvements:
    • Implemented pattern matching for list attributes (e.g., Transactions[].Date maps to Transactions[0].Date, Transactions[1].Date)
    • Added data flattening for complex extraction results using dot notation and array indices
    • Fixed numerical sorting for list items (now sorts 0, 1, 2, ..., 10, 11 instead of alphabetically)
    • Individual evaluation methods applied per nested attribute (EXACT, FUZZY, SEMANTIC, etc.)
  • Documentation: Comprehensive updates to evaluation docs and README files with nested structure examples and processing explanations
  • Use Cases: Enables complex document processing for bank statements (account details + transactions), invoices (vendor info + line items), and medical records (patient info + procedures)

• Enhanced Documentation and Examples

  • New example notebooks with improved clarity, modularity, and documentation

• Evaluation Framework Enhancements

  • Added confidence threshold to evaluation outputs to enable prioritizing accuracy results for attributes with higher confidence thresholds

• Comprehensive Metering Data Collection

  • The system now captures and stores detailed metering data for analytics, including:
    • Which services were used (Textract, Bedrock, etc.)
    • What operations were performed (analyze_document, Claude, etc.)
    • How many resources were consumed (pages, tokens, etc.)

• Reporting Database Documentation

  • Added comprehensive reporting database documentation

Changed

• Pin packages to tested versions to avoid vulnerability from incompatible new package versions.
• Updated reporting data to use document's queued_time for consistent timestamps
• Create new extensible SaveReportingData class in idp_common package for saving evaluation results to Parquet format
• Remove save_to_reporting from evaluation_function and replace with Lambda invocation, for smaller Lambda packages and better modularity.
• Harden publish process and avoid package version bloat by purging previous build artifacts before re-building

Fixed

• Defend against non-numeric confidence_threshold values in the configuration - avoid float conversion or numeric comparison exceptions in Assessement step
• Prevent creation of empty configuration fields in UI
• Firefox browser issues with signed URLs (PR #14)
• Improved S3 Partition Key Format for Better Date Range Filtering:
  • Updated reporting data partition keys to use YYYY-MM format for month and YYYY-MM-DD format for day
  • Enables easier date range filtering in analytics queries across different months and years
  • Partition structure now: year=2024/month=2024-03/day=2024-03-15/ instead of year=2024/month=03/day=15/

v0.3.3

[0.3.3]

Added

• Amazon Nova Model Fine-tuning Support

  • Added comprehensive ModelFinetuningService class for managing Nova model fine-tuning workflows
  • Support for fine-tuning Amazon Nova models (Nova Lite, Nova Pro) using Amazon Bedrock
  • Complete end-to-end workflow including dataset preparation, job creation, provisioned throughput management, and inference
  • CLI tools for fine-tuning workflow:
    • prepare_nova_finetuning_data.py - Dataset preparation from RVL-CDIP or custom datasets
    • create_finetuning_job.py - Fine-tuning job creation with automatic IAM role setup
    • create_provisioned_throughput.py - Provisioned throughput management for fine-tuned models
    • inference_example.py - Model inference and evaluation with comparison capabilities
  • CloudFormation integration with new parameters:
    • CustomClassificationModelARN - Support for custom fine-tuned classification models in Pattern-2
    • CustomExtractionModelARN - Support for custom fine-tuned extraction models in Pattern-2
  • Automatic integration of fine-tuned models in classification and extraction model selection dropdowns
  • Comprehensive documentation in docs/nova-finetuning.md with step-by-step instructions
  • Example notebooks:
    • finetuning_dataset_prep.ipynb - Interactive dataset preparation
    • finetuning_model_service_demo.ipynb - Service usage demonstration
    • finetuning_model_document_classification_evaluation.ipynb - Model evaluation
  • Built-in support for Bedrock fine-tuning format with multi-modal capabilities
  • Data splitting and validation set creation
  • Cost optimization features including provisioned throughput deletion
  • Performance metrics and accuracy evaluation tools

• Assessment Feature for Extraction Confidence Evaluation (EXPERIMENTAL)

  • Added new assessment service that evaluates extraction confidence using LLMs to analyze extraction results against source documents
  • Multi-modal assessment capability combining text analysis with document images for comprehensive confidence scoring
  • UI integration with explainability_info display showing per-attribute confidence scores, thresholds, and explanations
  • Optional deployment controlled by IsAssessmentEnabled parameter (defaults to false)
  • Added e2e-example-with-assessment.ipynb notebook for testing assessment workflow

• Enhanced Evaluation Framework with Confidence Integration

  • Added confidence fields to evaluation reports for quality analysis
  • Automatic extraction and display of confidence scores from assessment explainability_info
  • Enhanced JSON and Markdown evaluation reports with confidence columns
  • Backward compatible integration - shows "N/A" when confidence data unavailable

• Evaluation Analytics Database and Reporting System

  • Added comprehensive ReportingDatabase (AWS Glue) with structured evaluation metrics storage
  • Three-tier analytics tables: document_evaluations, section_evaluations, and attribute_evaluations
  • Automatic partitioning by date and document for efficient querying with Amazon Athena
  • Detailed metrics tracking including accuracy, precision, recall, F1 score, execution time, and evaluation methods
  • Added evaluation_reporting_analytics.ipynb notebook for comprehensive performance analysis and visualization
  • Multi-level analytics with document, section, and attribute-level insights
  • Visual dashboards showing accuracy distributions, performance trends, and problematic patterns
  • Configurable filters for date ranges, document types, and evaluation thresholds
  • Integration with existing evaluation framework - metrics automatically saved to database
  • ReportingDatabase output added to CloudFormation template for easy reference

Fixed

• Fixed build failure related to pandas, numpy, and PyMuPDF dependency conflicts in the idp_common_pkg package
• Fixed deployment failure caused by CodeBuild project timeout, by raising TimeoutInMinutes property
• Added missing cached token metrics to CloudWatch dashboards
• Added Bedrock model access prerequisite to README and deployment doc.

Full Changelog: v0.3.2...v0.3.3

v0.3.2

[0.3.2]

Added

• Cost Estimator UI Feature for Context Grouping and Subtotals

  • Added context grouping functionality to organize cost estimates by logical categories (e.g. OCR, Classification, etc.)
  • Implemented subtotal calculations for better cost breakdown visualization

• DynamoDB Caching for Resilient Classification

  • Added optional DynamoDB caching to the multimodal page-level classification service to improve efficiency and resilience
  • Cache successful page classification results to avoid redundant processing during retries when some pages fail due to throttling
  • Exception-safe caching preserves successful work even when individual threads or the overall process fails
  • Configurable via cache_table parameter or CLASSIFICATION_CACHE_TABLE environment variable
  • Cache entries scoped to document ID and workflow execution ARN with automatic TTL cleanup (24 hours)
  • Significant cost reduction and improved retry performance for large multi-page documents

Fixed

• "Use as Evaluation Baseline" incorrectly sets document status back to QUEUED. It should remain as COMPLETED.