Changelog
All notable changes to the Kubiya Workflow SDK will be documented here.[3.1.0] - 2024-01-26
Documentation Updates
- ๐ฏ Provider Clarity - Updated all documentation to accurately reflect ADK as the current orchestration provider
- ๐ณ Containerized Execution - Emphasized serverless Docker container execution throughout docs
- ๐ก๏ธ Application Layer Focus - New โWhy Kubiyaโ guide explaining safety vs god-mode AI systems
- ๐ DSL Reference - Comprehensive DSL reference with all workflow schema features
- ๐ Deterministic Workflows - New concept guide explaining determinism in workflows
Added
- Created
/concepts/why-kubiya.mdx
- Application-layer vs infrastructure-level AI orchestration - Created
/workflows/dsl-reference.mdx
- Complete DSL reference with AI-friendly explanations - Created
/concepts/deterministic-workflows.mdx
- Understanding deterministic execution
Changed
- Provider Overview - Rewritten to accurately show ADK as the orchestration provider
- Architecture Docs - Emphasized serverless, containerized execution model
- README - Updated to highlight containerized execution and ADK orchestration
- All Provider References - Removed misleading mentions of multiple providers
Key Messages
- Every workflow step runs in its own Docker container
- Workflows are stateless with zero pollution between executions
- ADK provides intelligent orchestration at the application layer
- No infrastructure god-mode permissions required
- Simple, deterministic schema that AI can easily generate
[3.0.0] - 2024-01-25
Added
- ๐ฏ Compose API - Unified interface for plan and act modes in ADK provider
- ๐ Execute Workflow - Complete implementation in ADK provider with streaming support
- ๐ Comprehensive ADK Docs - Agent architecture, streaming guide, API reference
- ๐ AI Automation Tutorial - Step-by-step guide for natural language workflows
- ๐ Default Runner Change - Changed from โautoโ to โkubiya-hostedโ for better reliability
- ๐ง Context Loading Fix - Direct platform context loading without MockContext issues
Changed
- ADK Provider now supports both SSE and Vercel AI SDK streaming formats
- Organization Info - Now uses client initialization parameter instead of API calls
- Workflow Generator - Includes smart runner selection logic based on user requirements
- Loop Orchestrator - Loads platform context using tools directly
- Config Handling - Fixed enum handling for model_provider configuration
Fixed
- MockContext missing ADK attributes (response_modalities, speech_config, etc.)
- Authentication error handling in ADK provider
- API endpoint issues with non-existent organization endpoint
- Streaming event formatting for both SSE and Vercel formats
Documentation
- Created
/providers/adk/agents.mdx
- Multi-agent architecture deep dive - Created
/providers/adk/streaming.mdx
- Comprehensive streaming guide - Created
/api-reference/compose.mdx
- Complete compose API reference - Created
/tutorials/ai-powered-automation.mdx
- Hands-on AI workflow tutorial - Updated
/workflows/architecture.mdx
- Added AI-powered generation section - Updated
README.md
- Prominent AI workflow generation section
[2.0.0] - 2024-01-20
Added
- ๐ค ADK Provider - AI-powered workflow generation using Googleโs Agent Development Kit
- ๐ SDK Server - Production-ready FastAPI server with REST endpoints
- ๐ก SSE Streaming - Real-time workflow execution updates
- ๐ Provider System - Pluggable architecture for AI providers
- ๐ Comprehensive Docs - New documentation site with MDX support
- ๐งช Jupyter Notebooks - Interactive examples for learning
- ๐ณ Docker Support - Complete Docker and Docker Compose configurations
- ๐ ๏ธ Makefile - Development automation commands
Changed
- Enhanced
Client
class with new platform methods - Fixed API endpoint format to use
operation=execute_workflow
- Added
native_sse=true
parameter for standard SSE streaming - Improved error handling and Python 3.9 compatibility
- Updated default runner to โautoโ for automatic selection
Fixed
- SSE streaming parsing for Kubiyaโs custom event format
- Byte/string handling in stream parser
- Response parsing for runners/integrations API
[1.5.0] - 2023-12-15
Added
- Workflow DSL with fluent interface
- Basic Kubernetes integration
- Testing framework with mocking support
Changed
- Improved workflow validation
- Better error messages
- Performance optimizations
Fixed
- Memory leaks in long-running workflows
- Race conditions in parallel execution
[1.0.0] - 2023-10-01
Added
- Initial release
- Core workflow engine
- Basic step execution
- Simple retry logic
The format is based on Keep a Changelog,
and this project adheres to Semantic Versioning.