dao-ai 0.0.36__tar.gz → 0.1.1__tar.gz

dao_ai-0.1.1/CHANGELOG.md

@@ -0,0 +1,102 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [Unreleased]
+
+## [0.1.0] - 2025-12-19
+
+### Added
+- **DSPy-Style Assertion Middleware**: New middleware for output validation and refinement
+  - `AssertMiddleware`: Hard constraints with retry - enforces requirements or fails after max attempts
+  - `SuggestMiddleware`: Soft constraints with optional single retry - provides feedback without blocking
+  - `RefineMiddleware`: Iterative improvement - runs multiple iterations to optimize output quality
+  - Multiple constraint types: `FunctionConstraint`, `LLMConstraint`, `KeywordConstraint`, `LengthConstraint`
+  - Factory functions: `create_assert_middleware()`, `create_suggest_middleware()`, `create_refine_middleware()`
+
+- **Conversation Summarization**: Automatic summarization of long chat histories
+  - `LoggingSummarizationMiddleware`: Extends LangChain's `SummarizationMiddleware` with detailed logging
+  - Configurable via `chat_history` in YAML with `max_tokens`, `max_tokens_before_summary`, `max_messages_before_summary`
+  - Logs original and summarized message/token counts for observability
+  - New example config: `config/examples/conversation_summarization.yaml`
+
+- **GEPA-Based Prompt Optimization**: Replaced MLflow optimizer with GEPA (Generative Evolution of Prompts and Agents)
+  - `optimize_prompt()` function using DSPy's evolutionary optimization
+  - `DAOAgentAdapter` bridges DAO ResponsesAgent with GEPA optimizer
+  - Automatic prompt registration with comprehensive tags
+  - Reflective dataset generation for self-improvement
+
+- **Structured Input/Output Format**: New `configurable` and `session` structure
+  - `configurable`: Static configuration (thread_id, conversation_id, user_id, store_num)
+  - `session`: Accumulated runtime state (Genie conversation IDs, cache hits, follow-up questions)
+  - Backward compatible with legacy flat `custom_inputs` format
+
+- **conversation_id/thread_id Interchangeability**: Databricks-friendly naming
+  - Input accepts either `thread_id` or `conversation_id` (conversation_id takes precedence)
+  - Output includes both in `configurable` section with synchronized values
+  - Auto-generation of UUID if neither is provided
+
+- **In-Memory Memory Configuration**: Added to Genie example config
+  - Simplified setup for development and testing
+
+### Changed
+- **ChatHistoryModel Refinements**:
+  - Removed unused `max_summary_tokens` attribute
+  - Updated `max_tokens` default from 256 to 2048
+  - Added `gt=0` validation for numeric fields
+  - Improved docstrings
+
+- **CLI Thread ID Handling**:
+  - `--thread-id` now defaults to auto-generated UUID instead of "1"
+  - YAML configs no longer require hardcoded thread_id values
+
+- **Orchestration Package Refactoring**:
+  - Created `orchestration` package with `supervisor` and `swarm` submodules
+  - Shared code consolidated in `orchestration/__init__.py`
+  - Improved code organization and maintainability
+
+### Removed
+- MLflow `GepaPromptOptimizer` wrapper (replaced with direct GEPA integration)
+- `backend` and `scorer_model` fields from `PromptOptimizationModel`
+- Hardcoded `thread_id: "1"` from all example configurations
+
+### Fixed
+- Handoff issues in supervisor pattern with `Command.PARENT` graph reference
+- Pydantic serialization warnings suppressed for Context serialization
+- StopIteration error in Genie tests (upgraded databricks-ai-bridge to 0.11.0)
+- Message validation middleware now properly terminates with `@hook_config(can_jump_to=["end"])`
+
+### Dependencies
+- Added `dspy>=2.6.27` for assertion middleware patterns
+- Added `gepa` for prompt optimization
+- Updated `databricks-ai-bridge` to 0.11.0
+
+## [0.0.1] - 2025-06-19
+
+### Added
+- Initial release of DAO AI multi-agent orchestration framework
+- Support for Databricks Vector Search integration
+- LangGraph-based workflow orchestration
+- YAML-based configuration system
+- Multi-agent supervisor and swarm patterns
+- Unity Catalog integration
+- MLflow model packaging and deployment
+- Command-line interface (CLI)
+- Python API for programmatic access
+- Built-in guardrails and evaluation capabilities
+- Retail reference implementation
+
+### Features
+- **Multi-Modal Interface**: CLI commands and Python API
+- **Agent Lifecycle Management**: Create, deploy, and monitor agents
+- **Vector Search Integration**: Built-in Databricks Vector Search support
+- **Configuration-Driven**: YAML-based configuration with validation
+- **MLflow Integration**: Automatic model packaging and deployment
+- **Monitoring & Evaluation**: Built-in assessment capabilities
+
+[Unreleased]: https://github.com/natefleming/dao-ai/compare/v0.1.0...HEAD
+[0.1.0]: https://github.com/natefleming/dao-ai/compare/v0.0.1...v0.1.0
+[0.0.1]: https://github.com/natefleming/dao-ai/releases/tag/v0.0.1