Version History

This section documents the MATIH Platform versioning strategy, release process, changelog format conventions, and historical release notes. Use this reference to understand what changed between versions, plan upgrades, and review the evolution of platform capabilities.

Versioning Strategy

The MATIH Platform uses Semantic Versioning (SemVer) for all platform releases, individual service versions, SDK versions, and Helm chart versions.

Version Format

MAJOR.MINOR.PATCH[-PRERELEASE][+BUILD]

Version Increment Rules

Platform vs Service Versioning

The MATIH Platform has a top-level version that represents the overall release. Individual services, SDKs, and Helm charts maintain their own version numbers:

Release Cadence

Support Policy

Changelog Format

All changelogs follow the Keep a Changelog (opens in a new tab) format with MATIH-specific extensions.

Entry Categories

Changelog Entry Template

## [1.3.0] - 2026-02-01
 
### Added
- **AI Service:** Added support for AWS Bedrock as an LLM provider (ADR-0010)
- **BI Workbench:** Added pivot table widget type with drill-down support
- **Connector SDK:** Added Salesforce connector with CDC support
- **ML Service:** Added model A/B testing with configurable traffic splitting
 
### Changed
- **Query Engine:** Upgraded Trino to 450 for improved Iceberg performance
- **Semantic Layer:** Metric definitions now support parameterized filters
 
### Deprecated
- **AI Service:** The `/api/v1/sql/generate` endpoint is deprecated; use `/api/v1/text-to-sql` instead (removal in 2.0.0)
 
### Fixed
- **IAM Service:** Fixed token refresh race condition under high concurrency (MATIH-IAM-7823)
- **Tenant Service:** Fixed DNS zone cleanup during tenant deprovisioning (MATIH-TENANT-7956)
- **Data Workbench:** Fixed SQL editor autocomplete for tables with special characters
 
### Security
- **All Services:** Updated Spring Boot to 3.2.5 to address CVE-2026-XXXXX
- **AI Service:** Added input sanitization for SQL injection prevention in text-to-SQL
 
### Performance
- **Query Engine:** Reduced query planning latency by 40% through plan caching
- **AI Service:** Reduced LLM token usage by 25% through prompt optimization
 
### Infrastructure
- **Helm Charts:** Added Vertical Pod Autoscaler (VPA) support for all services
- **CD Pipeline:** Added parallel service deployment in Phase 12 (3x faster)

Release Notes: Platform v1.0.0

Release Date: 2025-06-01 Codename: Foundation

This is the initial General Availability (GA) release of the MATIH Enterprise Platform.

Highlights

• Control Plane: 10 microservices providing IAM, tenant management, configuration, auditing, billing, notifications, observability, infrastructure management, platform registry, and API gateway
• Data Plane: 14 microservices covering AI conversational analytics, ML model lifecycle, query execution, BI dashboards, catalog, semantic layer, pipelines, data quality, governance, ontology, ops agents, rendering, and coordination
• Frontend: 8 workbench applications (BI, ML, Data, Agentic, Control Plane UI, Data Plane UI, Ops, Onboarding)
• AI Service: LangGraph multi-agent orchestrator with text-to-SQL, RAG, streaming responses, and DNN Builder
• Infrastructure: 55+ Helm charts, multi-cloud Terraform (Azure, AWS, GCP), Strimzi Kafka, Trino, ClickHouse, StarRocks

Known Issues

Release Notes: Platform v1.1.0

Release Date: 2025-07-01 Codename: Velocity

Highlights

• Google Vertex AI (Gemini) and AWS Bedrock support in the AI Service
• Pivot table widget in the BI Workbench
• Flink CDC pipeline templates
• Improved query performance with plan caching

Added

• AI Service: Google Vertex AI provider with Workload Identity support
• AI Service: AWS Bedrock provider with IRSA support
• AI Service: Agent performance metrics dashboard
• BI Workbench: Pivot table widget with drag-and-drop dimensions
• BI Workbench: Scheduled report delivery via email and Slack
• Data Workbench: Flink CDC pipeline visual builder
• ML Service: Automated hyperparameter tuning with Ray Tune
• Catalog Service: OpenLineage integration for automatic lineage capture
• Connector SDK: HubSpot CRM connector
• Templates: 5 new BI dashboard templates, 3 new ML experiment templates

Changed

• Query Engine: Enabled query plan caching (30% latency reduction for repeated query patterns)
• Semantic Layer: Metric definitions now support time grain parameters
• AI Service: Upgraded LangGraph to 0.5 for improved state management

Fixed

• MATIH-1001: Dashboard PDF export timeout for large dashboards (fixed with chunked rendering)
• MATIH-1002: ML model serving cold start reduced to under 15 seconds
• IAM Service: Fixed SCIM group membership update not propagating to JWT claims

Release Notes: Platform v1.2.0

Release Date: 2025-08-01 Codename: Context

Highlights

• Context Graph module for semantic understanding
• Ops Agent Service with AI-powered incident management
• ChromaDB vector store for RAG
• Enhanced multi-cloud Terraform modules

Added

• Context Graph: Full context graph module with Dgraph storage, semantic feature flags, and agent thinking service
• Ops Agent Service: AI-powered operations management with incident detection, root cause analysis, and automated remediation agents
• Data Infrastructure: ChromaDB deployment for vector storage (RAG, semantic search)
• AI Service: Prompt A/B testing for optimizing agent prompts
• AI Service: Quality drift detection for monitoring AI response quality over time
• Governance Service: Automated PII detection and classification
• Templates: Ontology templates for 6 industry domains

Changed

• Terraform: Refactored multi-cloud modules for Azure, AWS, and GCP parity
• Helm Charts: Standardized health probe configuration across all services
• AI Service: Store backend now supports PostgreSQL for durable state persistence

Fixed

• MATIH-1003: Ontology import OOM for large files (streaming parser implemented)
• Tenant Service: Fixed ingress controller deployment race condition during parallel tenant provisioning
• Query Engine: Fixed Iceberg table metadata refresh not detecting new partitions

Security

• All Services: Implemented network policies for all data plane services
• IAM Service: Added brute force protection (account lockout after 5 failed attempts)

Release Notes: Platform v1.3.0

Release Date: 2025-09-01 Codename: Scale

Highlights

• Horizontal autoscaling for AI inference
• Multi-region deployment support
• Enhanced DNN Builder with VibeTensor framework
• Real-time collaborative dashboards

Added

• AI Service: Prometheus-based custom metric autoscaling (requests/sec, latency P95, token usage)
• AI Service: VibeTensor framework support in DNN Builder
• BI Workbench: Real-time collaborative dashboard editing (WebSocket-based)
• BI Workbench: Dashboard versioning with diff and rollback
• ML Service: Model fairness monitoring templates
• Infrastructure: Multi-region Terraform modules with geo-replication
• Connector SDK: Slack and Jira connectors

Changed

• Autoscaling: All services now support VPA recommendations in "Off" mode (recommend only)
• Kafka: Upgraded to Strimzi 0.42 with KRaft mode (ZooKeeper-less)
• Trino: Upgraded to Trino 450 with improved Iceberg support

Deprecated

• AI Service: The vllm.baseUrl configuration key is deprecated; use vllm.base_url (snake_case) for consistency
• Query Engine: The /api/v1/query/execute synchronous endpoint is deprecated for queries expected to exceed 30 seconds; use /api/v1/query/execute/async instead

Performance

• AI Service: LLM token usage reduced by 25% through optimized schema embedding prompts
• Query Engine: Iceberg metadata caching reduces plan time by 40%
• BI Service: Dashboard loading time reduced by 50% through server-side rendering precomputation

Upgrade Guides

Upgrading Between Patch Versions

Patch upgrades (e.g., 1.2.3 to 1.2.4) are zero-downtime and require no manual steps:

# Update Helm chart versions and deploy
./scripts/cd-new.sh services dev

Upgrading Between Minor Versions

Minor upgrades (e.g., 1.2.x to 1.3.0) are backward-compatible and typically require:

1. Review the changelog for new features and deprecation notices
2. Update Helm chart values if new configuration options are needed
3. Run the CD pipeline to deploy updated services
4. Verify health checks pass for all services

# Pull latest code
# Review changelog
# Update values files as needed
./scripts/cd-new.sh all dev
./scripts/disaster-recovery/health-check.sh

Upgrading Between Major Versions

Major upgrades (e.g., 1.x.x to 2.0.0) may include breaking changes and require a migration guide. Each major release includes:

1. Migration Guide: Detailed document listing every breaking change and the required migration steps
2. Migration Scripts: Automated scripts for database migrations, configuration updates, and data transformations
3. Compatibility Matrix: Table showing which SDK versions are compatible with which platform versions
4. Support Timeline: Dates for end-of-support for the previous major version

Compatibility Matrix

Release Notes: Platform v1.4.0

Release Date: 2025-10-01 Codename: Intelligence

Highlights

• Context Graph RBAC and security hardening
• Agent thinking traces for explainable AI
• Enhanced Ops Agent with runbook automation
• Connector SDK v2 with async support

Added

• Context Graph: RBAC-based access control for context graph nodes and edges
• Context Graph: Agent thinking embeddings for semantic search over reasoning traces
• AI Service: Thinking trace visualization in the Agentic Workbench
• Ops Agent Service: Runbook automation with step-by-step execution and approval gates
• Ops Agent Service: Integration with PagerDuty and OpsGenie for incident escalation
• Connector SDK: v2 with async data reading, progress callbacks, and cancellation support
• ML Service: Model card generation with automated documentation
• BI Workbench: Map widget with geospatial data overlay and choropleth support
• Templates: 4 new ontology templates (manufacturing, logistics, education, media)
• SDKs: Added OpenTelemetry tracing propagation to all SDKs

Changed

• Helm Charts: Migrated all services to use topologySpreadConstraints for better pod distribution
• Kafka: Topic naming convention standardized to matih.{domain}.{entity}.{event} format
• AI Service: Default LLM provider changed from OpenAI to Azure OpenAI for enterprise deployments
• Security: All inter-service communication now requires mTLS (previously optional in dev)

Fixed

• AI Service: Fixed WebSocket reconnection dropping conversation context (MATIH-AI-8234)
• Tenant Service: Fixed concurrent tenant provisioning causing namespace conflicts (MATIH-TENANT-8456)
• ML Service: Fixed model artifact upload failing for files larger than 2GB (MATIH-ML-8123)
• Query Engine: Fixed Trino session properties not being reset between tenant queries (MATIH-QUERY-8567)
• BI Service: Fixed dashboard sharing permissions not propagating to cloned dashboards (MATIH-BI-8345)

Security

• IAM Service: Added brute force IP blocking with configurable thresholds
• All Services: Upgraded all base images to address CVE-2025-XXXXX series
• Governance Service: Added automated PII scanning for new connector data sources

Performance

• AI Service: Schema embedding retrieval improved 3x through hierarchical caching
• Catalog Service: Metadata search latency reduced by 60% with Elasticsearch query optimization
• Render Service: Server-side chart rendering parallelized, reducing dashboard export time by 70%

Infrastructure

• Terraform: Added GCP Artifact Registry module for multi-cloud container registry support
• CD Pipeline: Added canary deployment support for data plane services
• Monitoring: Added SLO-based alerting for all services (error rate, latency P99, availability)

Release Notes: Platform v1.5.0

Release Date: 2025-11-15 Codename: Enterprise

Highlights

• HIPAA compliance mode for healthcare tenants
• Data mesh support with federated governance
• Advanced agent evaluation framework
• Multi-region active-active deployment

Added

• Governance Service: HIPAA compliance mode with automated audit controls and data handling restrictions
• Governance Service: Data mesh federated governance with domain-level ownership and policy delegation
• AI Service: Agent evaluation framework with automated accuracy, relevance, and safety scoring
• AI Service: Prompt versioning with rollback and A/B testing
• Infrastructure: Multi-region active-active deployment with CockroachDB for control plane (optional)
• ML Service: Feature importance visualization in model monitoring dashboards
• Data Workbench: Visual pipeline builder with drag-and-drop task composition
• Connector SDK: Google BigQuery and Snowflake connectors
• Templates: Spark streaming templates for real-time feature computation

Changed

• Database: Alembic migrations now run as Kubernetes init containers (previously in-process)
• AI Service: Default agent timeout increased from 30s to 60s for complex queries
• Billing Service: Introduced granular metering for LLM token usage by model and provider

Deprecated

• AI Service: Module-level feature flags (MODULE_BI_ENABLED, etc.) are deprecated; use the Config Service feature flag system instead (removal in 2.0.0)
• Connector SDK v1: The synchronous DataReader interface is deprecated; use the async AsyncDataReader from SDK v2

Fixed

• AI Service: Fixed agent trace not capturing all retry attempts in SQL generation (MATIH-AI-8901)
• Tenant Service: Fixed certificate renewal failing when DNS zone has CNAME conflicts (MATIH-TENANT-9012)
• Pipeline Service: Fixed Temporal workflow history exceeding retention period causing orphaned workflows (MATIH-PIPE-9023)

Dependency Version Matrix

The following table documents the versions of key third-party dependencies used in each platform release:

Reporting Issues

If you encounter a bug or regression after an upgrade:

1. Check the Known Issues section of the relevant release notes
2. Search existing issues in the platform issue tracker
3. If the issue is new, report it with:
   • Platform version and service versions (from ./scripts/tools/platform-status.sh)
   • Steps to reproduce
   • Expected vs actual behavior
   • Relevant log snippets (redact any sensitive data)
   • Environment details (cloud provider, Kubernetes version, node pool configuration)

Contributing to the Changelog

When contributing changes to the platform, include a changelog entry in your pull request:

1. Add an entry to the CHANGELOG.md file in the root of the repository under the [Unreleased] section
2. Use the appropriate category (Added, Changed, Fixed, etc.)
3. Reference the service name in bold at the start of the entry
4. Include the issue/ticket number in parentheses at the end
5. Keep entries concise (one line per change)

Example:

## [Unreleased]
 
### Added
- **AI Service:** Added support for streaming agent traces via SSE (MATIH-AI-9234)
 
### Fixed
- **Tenant Service:** Fixed DNS zone delegation failing for subdomains with hyphens (MATIH-TENANT-9345)