Skip to content

Releases: mx-Liu123/AgentCommander

Initial Public Release - Agent Commander

20 Jan 05:05
@mx-Liu123 mx-Liu123
v0.1.0
c88534a

Choose a tag to compare

View all tags
Initial Public Release - Agent Commander Pre-release
Pre-release

Key Features

*   **Hierarchical Visual Workflow Editor with AI Assistant**: Adopts a sophisticated two-layer architecture to balance granular control with high-level evolution.
*   **Inner Subloop (Experiment Lifecycle)**: Encapsulates the rigorous execution logic within a single experiment folder (e.g., Hypothesis Generation $\to$ Code Implementation $\to$ Evaluation $\to$ Result Analysis).
*   **Outer Control Plane (Evolutionary Strategy)**: Orchestrates the macro-level logic *between* experiments. This layer manages **lineage inheritance** (cloning and mutating the best strategies), **meta-learning** (extracting lessons from failed branches), and **external exploration** (injecting fresh ideas via web search), ensuring the system evolves continuously rather than just iterating blindly.
*   **AI-Assisted Design**: Modify this complex topology directly via natural language commands (e.g., "Add a shell check after step 2").

Workflow Editor

  • Multi-Model CLI Integration: Deeply integrated with both Gemini CLI and Qwen CLI for powerful, prompt-driven code generation and analysis. Choose the backend that best fits your needs directly from the UI.
  • Inherited CLI Skills: Because it sits on top of the CLI ecosystem, AgentCommander inherits all the native capabilities of the underlying CLI tools. Any "skill" supported by Gemini/Qwen CLI (e.g., web search, file management, system commands) is automatically available to your agents within the workflow.
  • Infinite Iteration & Advanced Learning: Create self-improving loops where the agent experiments, learns from failures, and refines its strategy indefinitely. Advanced features like the "Lesson" mechanism (to learn from past errors) and online search integration (for inspiration) are available in example workflows to boost continuous improvement.
  • ML & Symbolic Regression: Specifically tailored to assist in discovering mathematical formulas and optimizing ML models through iterative experimentation.
  • Experiment Management & Evolutionary Tree: Automatically track and visualize experiment history, metrics, and branches as an "evolutionary tree", where each experiment node connects to its parent.
    Progress Tree
  • Dynamic Configuration: Manage global variables and system settings through a centralized UI.
Loading