Releases: awslabs/genai-idp-terraform
v0.4.8-tf.0
Release Notes — v0.4.8-tf.0
Upstream IDP Version: v0.4.8 | Previous: v0.3.18-tf.1
What's New
Agent Companion Chat
Session-based multi-turn AI assistant with real-time streaming, conversation history, and
multi-agent support (Analytics, Code Intelligence, Error Analyzer, General).
Enabled by default via enable_agent_companion_chat.
Test Studio
Web interface for managing test sets, running evaluations, and analyzing results with
interactive charts and side-by-side comparisons.
Enabled by default via enable_test_studio.
Error Analyzer
AI-powered document processing diagnostics using Claude Sonnet 4 + Strands. Surfaces failures
directly in the Web UI via a "Troubleshoot" button.
Enabled by default via enable_error_analyzer.
Docker Image Deployment for Pattern 1 & 3
BDA and SageMaker UDOP processors now deploy Lambda functions via ECR + CodeBuild (raising the package size limit from 250 MB to 10 GB).
⚠️ Expect ~15 min on first apply.
Evaluation in Step Functions
Evaluation is now an integrated state machine step across all three patterns. No stack
redeployment needed to toggle it — controlled via evaluation.enabled in configuration.
Post-Processing Decompressor
New Lambda that decompresses documents before invoking custom post-processing hooks.
External hook Lambdas no longer need to import idp_common.
MCP Integration (opt-in)
External access via Bedrock AgentCore Gateway with OAuth 2.0.
Enable via enable_mcp = true. Not available in GovCloud.
Breaking Changes
See docs/migration-v0.3.18-to-v0.4.8.md for full migration steps.
- Pattern 1 & 3 Docker migration — Lambda functions are replaced by Docker images; plan for
a deployment window - Evaluation EventBridge rules — existing rules/targets must be removed before upgrading
- Web UI env vars — CodeBuild environment variables renamed from
REACT_APP_*toVITE_*
v0.3.18-tf.1
Contributors
Assets 2
v0.3.18-tf.0
What's Changed
- feat: upgrade IDP from v0.3.8 to v0.3.18 in #11
BREAKING CHANGE: IAM policies updated with new permissions
Changes:
DynamoDB: full CRUD for tracking/configuration tables
SSM: GetParameter, PutParameter, GetParametersByPath
Bedrock: GetDataAutomationProject, GetBlueprint APIs
S3: DeleteObject, GetBucketLocation permissions
Conditional HITL policies using for_each patterns
Standardized environment variables across Lambdas
Upstream: aws-solutions-library-samples/accelerated-intelligent-document-processing-on-aws@v0.3.18
Full Changelog: v0.3.8...v0.3.18-tf.0
v0.3.8
GenAI IDP Accelerator for Terraform - Initial Release
This is the first public release of the Terraform implementation of the GenAI Intelligent Document Processing Accelerator, providing infrastructure-as-code management for intelligent document processing workflows on AWS.
What's Included
Modular Terraform Architecture
Complete Terraform implementation with 18+ reusable modules:
- Processing Patterns: BDA Processor, Bedrock LLM Processor, SageMaker UDOP Processor
- Core Infrastructure: Processing Environment, GraphQL API, Web UI
- Supporting Modules: Knowledge Base, User Identity, Human Review, Reporting, and more
Three Document Processing Patterns
- BDA Processor: Bedrock Data Automation for standard document types
- Bedrock LLM Processor: Flexible extraction using Claude/Nova foundation models
- SageMaker UDOP Processor: Custom model deployment for specialized documents
Key Features
- Serverless architecture (Lambda, Step Functions, AppSync)
- KMS encryption and IAM least privilege security
- Optional VPC deployment support
- React-based Web UI with CloudFront distribution
- Knowledge Base for RAG-based document querying
- CloudWatch dashboards and monitoring
Deployment Examples
Ready-to-use examples for each processing pattern in the examples/ directory.
Upstream Compatibility
Based on accelerated-intelligent-document-processing-on-aws v0.3.8
Getting Started
See the README for prerequisites and deployment instructions.