![\"Codemini](\"./codemini-web/codemini_logo.png\"){kind=logo}
\r\n\r\n \r\n
\r\n An extremely restrained coding + tasks CLI.
\r\n Every platform. Every terminal. Minimal by design.\r\n
\r\n ![\"npm](\"https://img.shields.io/npm/v/codemini-cli?style=flat-square&logo=npm\")
![\"node](\"https://img.shields.io/badge/node-%3E%3D22.19.0-339933?style=flat-square&logo=nodedotjs&logoColor=white\"){kind=icon}
![\"license\"](\"https://img.shields.io/badge/license-MIT-blue?style=flat-square\"){kind=icon}
![\"web](\"https://img.shields.io/badge/Web_UI-included-7c3aed?style=flat-square\"){kind=icon}
![\"quick](\"https://img.shields.io/badge/quick_start-5_commands-0ea5e9?style=flat-square\"){kind=icon}
\r\n English ·\r\n 简体中文 ·\r\n Web UI ·\r\n Operator Guide ·\r\n Deployment\r\n
\r\n\r\n---\r\n\r\n## English\r\n\r\n### What is Codemini?\r\n\r\nCodemini is a **coding + tasks CLI** for Windows, macOS, and Linux. It talks to OpenAI-compatible and Anthropic APIs, and keeps sessions, project state, skills, memories, and indexes on your machine. Model requests go only to the provider you configure.\r\n\r\nIt can refactor a codebase, drive a Git workflow, run a multi-step pipeline, research a technical question, or process local files.\r\n\r\nCodemini provides two interfaces powered by the same runtime:\r\n\r\n- A terminal TUI for keyboard-first work.\r\n- A browser Web UI for sessions, files, diffs, terminals, research, CodeWiki, skills, hooks, and settings.\r\n\r\n### Design\r\n\r\nCodemini keeps the prompt small instead of making every feature part of every turn.\r\n\r\n- **Managed context.** Compaction, prompt preflight, and tool-result spill keep long sessions usable.\r\n- **Lazy-loaded skills.** Routing metadata stays small; the full `SKILL.md` is loaded only when a skill is selected.\r\n- **Project-aware retrieval.** Symbols, AST structure, dependencies, and the project graph locate evidence before large files are read.\r\n- **Proportional approvals.** Read-only work can proceed while risky actions remain visible and reviewable.\r\n- **Local persistence.** Sessions, memory, indexes, and checkpoints live on your machine.\r\n- **Provider choice.** Use an OpenAI-compatible or Anthropic endpoint; configure a separate fast model when useful.\r\n\r\n### Quick Start\r\n\r\nRequires Node.js 22.19 or newer.\r\n\r\n```bash\r\nnpm install -g codemini-cli\r\n\r\ncodemini config set gateway.base_url http://127.0.0.1:8000/v1\r\ncodemini config set gateway.api_key your_api_key\r\ncodemini config set model.name your_model_name\r\n\r\ncodemini doctor\r\ncodemini\r\n```\r\n\r\nStart the Web UI instead:\r\n\r\n```bash\r\ncodemini --web\r\n```\r\n\r\nOr choose a project and port explicitly:\r\n\r\n```bash\r\ncodemini web --project . --port 3456\r\n```\r\n\r\n### Beyond Code: Automated Tasks\r\n\r\n`codemini run` turns a natural-language request into a one-off local workflow.\r\n\r\n```bash\r\n# Interactive session\r\ncodemini \"Summarize the recent Git history and update CHANGELOG.md\"\r\n\r\n# One-off task\r\ncodemini run \"Find stale dependencies and explain the upgrade risks\"\r\n\r\n# Role-constrained harness\r\ncodemini run --harness reviewer \"Review the current changes\"\r\n\r\n# Sequential pipeline\r\ncodemini run --pipeline \"Run tests, fix failures, and summarize the result\"\r\n```\r\n\r\n| Command | Purpose |\r\n| --- | --- |\r\n| `codemini` / `codemini chat` | Start an interactive terminal session. |\r\n| `codemini run