v0.3.5 - Enhanced Error Handling

What's New

• ExecutionResult wrapper: Better structured execution outcomes
• Configurable error feedback: New detailed_error_feedback
  parameter
• Improved streaming parser: Fixed buffer flushing and code block
  detection
• Better output limiting: Enhanced handling of long execution results

Enhanced Execution Context & Output Management - v0.3.4

Improvements

• Add max_execution_result_length parameter to prevent LLM context overflow
• Clarify Jupyter-like persistent session behavior in prompts

v0.3.3 - Configurable Python Block Identifier

✨ Features

• Configurable Python Block Identifier: Customize code block identifier (default: "python")
• Code Snippet Tracking: Track executed code in agent responses
• Enhanced Agent Identity: Improved prompt structure and naming

🔧 Improvements

• Better function description formatting
• Added litellm and openai dev dependencies
• Refactored role_definition to agent_identity

v0.3.2 - Enhanced Debugging & Message Management

This release improves PyCallingAgent's observability and conversation management:

• Enhanced Debug Logging: Added comprehensive debug logging for final responses, code snippets, execution results, system prompts, and history tracking
• Improved Message History: Refined conversation history trimming logic with better handling of system vs non-system messages
• Better Observability: Enhanced debugging capabilities throughout the agent execution pipeline

The changes focus on making the agent's execution flow more transparent and easier to debug during development and production use.

v0.3.1 - Security Validation & Enhanced Code Execution Safety

• Adds tests for security violation and syntax error detection in code execution.

v0.3.0 - Major Refactor

🚀 Release v0.3.0 - Major Framework Refactor

✨ What's New

Complete framework overhaul with new architecture and enhanced capabilities.

🔧 Key Features

• 🤖 Abstract Model Interface: Support for multiple LLM providers (OpenAI, LiteLLM)
• ⚡ Real-time Streaming: Event streaming
• 🛡️ Security: AST-based code validation and execution limits
• 💾 State Management: Runtime state persistence across conversations

version 0.2.3

fix: streamline agent execution flow and fix inconsistencies.

version 0.2.2

Complete project restructuring:

• Reorganized directory structure
• Updated dependency management
• Added streaming response

version 0.1.0

PyAgent v0.1.0

PyAgent is a tool-augmented agent framework that enables function-calling through LLM code generation. This is our first release with core functionality implemented.

Features

Code Generation Based Function Calling

• Use Python code generation instead of JSON schemas
• Full access to Python language features
• Better error messages through actual code execution

State Management

• Maintain variables and objects across interactions
• Direct access to runtime state
• Support for complex Python objects

Tool Integration

• Use Python classes and functions as tools
• Automatic docstring and type hint integration
• Seamless class method support

Installation

pip install llm-py-agent

Breaking Changes

• None (First Release)

Bug Fixes

• None (First Release)

Documentation

• Added comprehensive README with examples
• Included docstrings for all public APIs
• Added example scripts demonstrating key features