V 0.7.3

What's Changed

• Fix/agnets by @Iamsdt in #91
• Fix failing sequential tool calls by @atharvacoolkni in #93
• Bump langchain-text-splitters from 0.3.11 to 1.1.2 by @dependabot[bot] in #101
• Vertex AI as a provider in the agent framework by @Mothilal-M in #96
• Bump authlib from 1.6.4 to 1.6.11 by @dependabot[bot] in #100
• Bump langsmith from 0.4.29 to 0.7.31 by @dependabot[bot] in #99
• Bump python-multipart from 0.0.20 to 0.0.26 by @dependabot[bot] in #98
• Enhance Vertex AI support in Agent class and provider detection logic by @Iamsdt in #103
• Add JSON output type for Google and OpenAI by @atharvacoolkni in #97

New Contributors

• @dependabot[bot] made their first contribution in #101
• @Mothilal-M made their first contribution in #96

Full Changelog: 0.7.0...0.7.3

0.7.0

What's Changed

• V4 by @Iamsdt in #46
• Refactor message routing and evaluation presets by @Iamsdt in #47
• Prashant lib by @prashant4654 in #48
• fixed evalution framework , fixed gen_ai_format function by @atharvacoolkni in #51
• reasoning_config by @atharvacoolkni in #56
• Refactor context trimming method and enhance error handling in Compil… by @Iamsdt in #59
• Add test for excluding Inject defaults and update dependencies by @Iamsdt in #60
• Enhance prompt injection detection by adding word boundaries for know… by @prashant4654 in #63
• 64 feature allow tools to return structured state updates by @Iamsdt in #65
• Fix/steam tool by @Iamsdt in #66
• Add 'is_non_streaming' key to StreamNodeHandler output for clarity by @Iamsdt in #70
• Enhance GoogleGenAIConverter to store and handle thought_signature in… by @Iamsdt in #71
• Implement retry and fallback mechanism in Agent with RetryConfig; add… by @Iamsdt in #72
• Add Skills system and A2A SDK integration by @atharvacoolkni in #67
• Skills by @Iamsdt in #73
• Enhance long-term memory integration and update GoogleEmbedding by @prashant4654 in #57
• Refactor code structure for improved readability and maintainability by @Iamsdt in #74
• Multimodal by @Iamsdt in #84
• Potential fix for code scanning alert no. 1: Workflow does not contain permissions by @Iamsdt in #85
• fix: Update skill documentation and improve resource handling instruc… by @Iamsdt in #86
• Re Arrange Folder by @Iamsdt in #87
• Fix/example by @Iamsdt in #88
• Multimodal fix by @Iamsdt in #89
• Example fix by @Iamsdt in #90

New Contributors

• @prashant4654 made their first contribution in #48
• @atharvacoolkni made their first contribution in #51

Full Changelog: 0.5.7...0.7.0

Release:0.5.7 Testing and Evulation

Agentflow Release Notes

PyPI Project: 10xscale-agentflow v0.5.7

Version: 0.5.7 (later)

Summary:
This release introduces comprehensive testing and evaluation frameworks for agents, enhances the @ tool decorator with additional capabilities, and includes bug fixes to improve stability and reliability. The addition of dedicated testing and evaluation modules provides developers with powerful tools for agent validation and performance assessment.

Key Enhancements

Testing Framework

• Introduced new testing module for unit testing agents
• Provides utilities for agent behavior validation and test assertions
• Enables developers to write comprehensive test suites for agent workflows

Evaluation Framework

• Added new evaluation module for evaluating agent performance
• Supports metrics collection and analysis for agent interactions
• Enables performance benchmarking and quality assessment of agent implementations

Enhanced @ tool Decorator

• Extended @ tool decorator for function-based tools with additional metadata support
• Added capability to specify tags for tools, enabling semantic categorization
• Added description parameter for better tool documentation
• Tags enable filtering and discovery of tools based on use cases

Improved Agent Class

• Enhanced Agent class now supports tool filtering by tags
• Allows selective tool inclusion in agent workflows based on tag-based criteria
• Improves modularity and tool management in complex agent systems

Google SDK Support for Converters

• Added Google SDK support for converters

Bug Fixes

• Fixed various stability issues in agent execution
• Improved error handling in tool execution pipelines
• Enhanced resource cleanup and management

Impact & Migration Notes

• New testing module provides optional testing utilities; existing tests can be refactored to use these utilities
• New evaluation module enables performance tracking; integration is optional
• Enhanced @tool decorator is backward compatible; existing tools continue to work as-is
• To use tag-based filtering, pass tags when decorating tools with @tool and filter in Agent initialization
• No breaking API changes; all changes are additive and backward compatible

New Modules

• agentflow.testing - Unit testing utilities and assertions for agents
• agentflow.evaluation - Evaluation framework for agent performance analysis

Recommended Actions

• Review and adopt the new testing framework for your agent test suites
• Integrate the evaluation module to track and improve agent performance
• Consider adding tags to existing tools for better organization and filtering
• Explore the new testing patterns and evaluation metrics for your use cases

Minor Release (Bug fixes)

Agentflow Release Notes

PyPI Project: 10xscale-agentflow v0.5.6

Version: 0.5.6 (later)

Summary:
This release brings major improvements to error handling, tool execution, agent creation, and database robustness. All notable changes and fixes from versions 0.5.1 through 0.5.5 are consolidated here for clarity and migration ease.

Key Enhancements

• Improved error handling and retry logic in PgCheckpointer, with actionable logging and safer thread updates (preserving thread names and returning creation status).
• Enhanced interrupt handling for MCP and Node execution, ensuring predictable pause/resume and correct event metadata.
• MCP Tool Execution now supports passing user information for richer context.
• Introduced the Agent class for simplified agent instantiation and LLM integration.
• Added tag-based tool filtering and @tool decorator for better tool discovery and schema generation.
• Support for handoff tools and remote tools, enabling agent transfer and external tool call handling.
• Parallel tool calls within nodes for improved throughput.
• ID generation and SQL type mapping fixes for numeric types (int, bigint).
• Improved logging with module-specific logger names.
• Fixed timestamp handling and thread reconstruction in PgCheckpointer, including updated_at field for accurate thread snapshots.

Impact & Migration Notes

• PgCheckpointer API now protects against NULL thread name updates and provides clearer error propagation for diagnostics.
• Interrupt handling is more deterministic; users relying on pause/resume should see improved behavior.
• To forward user details to MCP tools, instantiate ToolNode with pass_user_info_to_mcp=True (opt-in).
• Use tags to filter tools in agents and tool nodes; remote tool behavior requires frontends to handle RemoteToolCallBlock messages and resume graphs appropriately.
• Parallel tool execution may improve performance for workflows with multiple tool calls.
• Historical data and UI displays now incorporate updated_at for threads, improving ordering and accuracy.
• No breaking API changes, but migration is recommended to leverage new features and robustness improvements.

Minor Release

What's Changed

• Play pause by @Iamsdt in #1
• Replace with Injector by @Iamsdt in #2
• Fix/container by @Iamsdt in #3
• Fix/stream by @Iamsdt in #4
• Add cache control to system prompts in get_weather and enhance Inject… by @Iamsdt in #5
• Enhance README and code structure: Update weather agent example for a… by @Iamsdt in #6
• Fix/stream chunks by @Iamsdt in #7
• Feature/invoke by @Iamsdt in #8
• Feature/prebuilt by @Iamsdt in #9
• fix: Simplify coverage reporting in CI workflow and remove redundant steps by @Iamsdt in #10
• fix: Remove redundant blank line in README.md by @Iamsdt in #11
• feat: Implement stop functionality for running graphs and enhance thr… by @Iamsdt in #13
• Added thread aname generator by @Iamsdt in #14
• refactor: Remove stop_requested and run_id from ThreadInfo; update ex… by @Iamsdt in #15
• feat: Enhance PgCheckpointer with thread management and schema initia… by @Sarthak422004 in #16
• Feature/store by @suchith83 in #17
• Feature/document by @suchith83 in #19
• Feature/document by @Iamsdt in #20
• Feature/stream event2 by @Iamsdt in #21
• Project Name change by @Iamsdt in #22
• 23 handling tool result from custom mcp server by @Iamsdt in #24
• updated astore(), asearch() methods in mem0store by @suchith83 in #26
• Cleanup and Code Optimizations by @Iamsdt in #27
• Refactor state graph and execution state enums to use str Enum; remov… by @Iamsdt in #28
• Fix/docs by @Iamsdt in #29
• Feature/handoff by @Iamsdt in #30
• Fix/logger by @Iamsdt in #31
• Add tutorials for Embedding Services and Mem0Store by @Iamsdt in #32
• Fix timestamp handling in PgCheckpointer to avoid validation errors by @Iamsdt in #35
• Add updated_at field to ThreadInfo and PgCheckpointer for improved ti… by @Iamsdt in #36

New Contributors

• @Sarthak422004 made their first contribution in #16
• @suchith83 made their first contribution in #17

Full Changelog: https://github.com/10xHub/Agentflow/commits/0.5.2