Skip to content

ChatGPT CLI v1.10.3

Choose a tag to compare

@kardolus kardolus released this 17 Jan 04:37
· 46 commits to main since this release

New Features

  • Agent Mode:
    • ReAct Mode: The new default mode that iteratively uses reasoning and acting (think→act→observe) to accomplish tasks.
    • Plan/Execute Mode: Generates an entire plan first and then executes the steps sequentially. This workflow is designed for more complex, multi-step tasks.
    • Agent Configuration: Enhanced configuration options to define the agent's behavior, budgets, and policies.
    • Logging and Reporting: Detailed execution logs are now generated for agent operations, stored in a dedicated directory for easy access and review.

Improvements

  • Enhanced Shell and File System Operations:
    • Expanded shell command and file operation capabilities and safety controls to prevent unintended access or modification of important data.
  • Performance Optimizations:
    • Quickened response time for iterative tasks through refined logic and improved resource handling.

Bug Fixes

  • Command Handling: Resolved issues where commands or file operations would fail under specific conditions due to policy limitations.
  • Template Parsing: Fixed errors in handling templates that affected task execution reliability.

Other Changes

  • Updated Smoke Test Scripts: More extensive automated testing has been introduced to ensure the robustness of new features, ensuring broader compatibility and reliability.
  • Refactoring for Maintainability: Code has been refactored to improve readability and maintainability, incorporating community feedback and code improvements.

Upgrade Notes

  • How to Update:
    • If you installed chatgpt-cli via Homebrew, simply run: brew upgrade chatgpt-cli.
    • Alternatively, download the binaries directly from Github Releases for your specific operating system.

This version adds significant features and improvements aimed at enhancing the usability and functionality of the software for complex task execution scenarios. We encourage users to explore the new agent modes and provide feedback on their experiences.