fix(agent_utils.py): remove debug print statement for max_input_tokens to clean up code
refactor(anthropic_token_limiter.py): update state_modifier to use anthropic_trim_messages for better token management and maintain message structure
feat(anthropic_token_limiter.py): add convert_message_to_litellm_format function to standardize message format for litellm
fix(anthropic_token_limiter.py): update wrapped_token_counter to handle only BaseMessage objects and improve token counting logic
chore(anthropic_token_limiter.py): add debug print statements to track token counts before and after trimming messages
feat(main.py): add DEFAULT_MODEL constant to centralize model configuration
feat(main.py): enhance logging and error handling for better debugging
feat(main.py): implement state_modifier for managing token limits in agent state
feat(anthropic_token_limiter.py): create utilities for handling token limits with Anthropic models
feat(output.py): add print_messages_compact function for debugging message output
test(anthropic_token_limiter.py): add unit tests for token limit utilities and state management
refactor(agent_utils.py): import run_research_agent and run_web_research_agent from their respective modules to streamline the code structure and enhance clarity
This commit introduces a new `Trajectory` model to the database, which tracks the sequence of actions taken by agents, including tool executions and their results. The addition of the `TrajectoryRepository` allows for storing and retrieving these trajectories, enabling better analysis of agent behavior and debugging of issues.
Additionally, the commit refactors existing code to utilize the new repository and model, improving the overall architecture and maintainability of the codebase. This change is essential for enhancing the capabilities of the agent system and providing a more robust framework for future development.
style(agent_utils.py): format imports and code for better readability
refactor(agent_utils.py): standardize model name and cost calculation logic for clarity and maintainability
chore(anthropic_callback_handler.py): create a new file for the AnthropicCallbackHandler implementation and related functions
style(agent_utils.py): format imports and code for better readability
refactor(agent_utils.py): standardize model name and cost calculation logic for clarity and maintainability
chore(anthropic_callback_handler.py): create a new file for the AnthropicCallbackHandler implementation and related functions
Ariel Frischer
AI Christianson
Andrew I. Christianson