Refactor agent.py and graph.py to enhance agent functionality and logging. Introduce Configuration class for managing parameters, improve state handling in AgentRunner, and update agent graph to support step logging and user interaction. Add new tests for agent capabilities and update requirements for code formatting tools.

Fix logging in AgentNode's step callback in graph.py to display action directly instead of action step, improving clarity in debug output.

Refactor AgentNode in graph.py to utilize CodeAgent instead of ToolCallingAgent, update prompt template loading to use code_agent.yaml, and enhance step callback logging for improved debugging.

Enhance logging in AgentNode's __call__ method in graph.py to track state before and after processing, including detailed information about messages, question, and answer.

Add max_steps parameter to ToolCallingAgent in graph.py for improved agent control

Enhance graph.py to load prompt templates from local file and configure LiteLLM; add new prompt files for tool calling and code agent

Add initial implementation of AgentRunner and agent graph; include .gitignore and update requirements