How shareai-lab/mini_kode works
Project type and purpose
• This is a minimal “mini Kode” CLI agent tool (an open-source tool) with two iterations:
• v1_basic_agent.py – model-as-agent loop with four core tools (bash, read_file, write_file, edit_text).
• v2_todo_agent.py – adds a structured Todo toolchain and system-reminder mechanism for explicit planning and self-supervision.
• Supplementary “articles” folder holds documentation drafts; “demo” folder contains standalone game demos (HTML/JS/Python) for showcasing graphics and physics utilities but is not part of the core CLI workflow.
File structure
• Top-level:
• README.md / README_en.md – project overview and architecture explanation.
• requirements.txt – Python dependencies (e.g., OpenAI SDK).
• v1_basic_agent.py & v2_todo_agent.py – main entry points for each version.
• “articles/” – Markdown documentation versions.
• “demo/” – independent web & Python demos illustrating UI, graphics, physics, AI in browser or CLI simulations.
• No separate “src” or “lib” folder; Python scripts encapsulate both core logic and tool implementations.
README for additional insights
• v1 workflow: CLI ↔ LLM via OpenAI API ↔ tool dispatcher ↔ filesystem & shell commands.
• Architectural patterns: tool-dispatching agent loop, self-supervising through structured external memory (Todo list), separation of prompts, tools, and I/O.