[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-80655":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":16,"stars30d":17,"stars90d":16,"forks30d":16,"starsTrendScore":16,"compositeScore":18,"rankGlobal":10,"rankLanguage":10,"license":19,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":10,"pushedAt":10,"updatedAt":28,"readmeContent":29,"aiSummary":30,"trendingCount":16,"starSnapshotCount":16,"syncStatus":31,"lastSyncTime":32,"discoverSource":33},80655,"AutoRun","Gao-Ruilin\u002FAutoRun","Gao-Ruilin","An AI agent for coding and others","",null,"Python",102,10,6,3,0,52,48.32,"MIT License",false,"main",true,[24,25,26,27],"ai","ai-agent","ai-agents","code-generation","2026-06-12 04:01:29","\u003C!-- Language Switcher -->\n\u003Cp align=\"right\">\n  \u003Ca href=\"#english\">English\u003C\u002Fa> |\n  \u003Ca href=\"#chinese\">中文\u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\n\u003Ca id=\"english\">\u003C\u002Fa>\n\n# AutoRUN v1\n\nA universal AI coding assistant supporting OpenAI and Anthropic compatible APIs.\n\n## Installation\n\nRequires Python 3.8+.\n\n### Quick Install (recommended)\n\n**Windows:**\n```cmd\ninstall.bat\n```\n\n**macOS \u002F Linux:**\n```bash\n.\u002Finstall.sh\n```\n\nThis creates a virtual environment, installs dependencies, and sets up the `autorun` command.\n\n### Manual Install\n\n```bash\ncd AutoRUN_v1\npip install -e .\n```\n\nAfter installation, the `autorun` command is available system-wide:\n\n```bash\nautorun --version\n```\n\n### Alternative: pip + requirements.txt\n\n```bash\ncd AutoRUN_v1\npip install -r requirements.txt\npython cli.py\n```\n\n## Quick Start\n\n```bash\nautorun\n```\n\nOn first run, AutoRUN will guide you through API configuration interactively:\n\n```\n╔══════════════════════════════════════════╗\n║     AutoRUN v1.0 — 首次设置             ║\n╚══════════════════════════════════════════╝\n\n选择 API 类型 [1=OpenAI, 2=Anthropic] [默认 openai]: 1\nAPI URL [默认 https:\u002F\u002Fapi.openai.com]:\nAPI Key: sk-xxxxxxxxxxxx\n模型名称 [默认 gpt-4o]:\n\n✓ 配置已保存！\n```\n\nConfig is stored at `~\u002F.autorun\u002Fconfig.json`.\n\n## API Configuration\n\nThree methods, listed by priority:\n\n### Method 1: First-run Setup Wizard\n\n```bash\nautorun\n```\nFollow the interactive prompts on first run, or run `autorun --setup` to reconfigure anytime.\n\n### Method 2: Environment Variables\n\n**Windows (PowerShell):**\n```powershell\n$env:AUTORUN_API_TYPE = \"openai\"\n$env:AUTORUN_API_URL  = \"https:\u002F\u002Fapi.openai.com\"\n$env:AUTORUN_API_KEY  = \"sk-xxxxxxxxxxxx\"\n$env:AUTORUN_MODEL    = \"gpt-4o\"\nautorun\n```\n\n**Windows (CMD):**\n```cmd\nset AUTORUN_API_TYPE=openai\nset AUTORUN_API_URL=https:\u002F\u002Fapi.openai.com\nset AUTORUN_API_KEY=sk-xxxxxxxxxxxx\nset AUTORUN_MODEL=gpt-4o\nautorun\n```\n\n**macOS \u002F Linux:**\n```bash\nexport AUTORUN_API_TYPE=openai\nexport AUTORUN_API_URL=https:\u002F\u002Fapi.openai.com\nexport AUTORUN_API_KEY=sk-xxxxxxxxxxxx\nexport AUTORUN_MODEL=gpt-4o\nautorun\n```\n\n### Method 3: Configure inside REPL\n\n```bash\nautorun\n```\n\nInside the REPL:\n```\n\u002Fapi type openai                          # API type: openai or anthropic\n\u002Fapi url https:\u002F\u002Fapi.openai.com           # API base URL\n\u002Fapi key sk-xxxxxxxxxxxx                  # Your API key\n\u002Fmodel gpt-4o                             # Model name\n```\n\n## Usage\n\n### CLI REPL (default interactive mode)\n\n```bash\nautorun\n```\n\n### Web UI\n\n```bash\nautorun --web\n```\n\nDefault: http:\u002F\u002F127.0.0.1:8765\n\n```bash\n# Custom port and host\nautorun --web --port 8080 --host 0.0.0.0\n```\n\n### Pipe Mode\n\n```bash\necho \"Explain this code\" | autorun --print\nautorun --print \"Explain this code\"\n```\n\n### Reconfigure API\n\n```bash\nautorun --setup\n```\n\n## CLI Options\n\n| Flag | Description |\n|------|-------------|\n| `--version`, `-V`, `-v` | Show version |\n| `--setup` | Re-run API setup wizard |\n| `--web` | Start Web UI server |\n| `--print`, `-p` | Pipe\u002Fprint mode (non-interactive) |\n| `--port \u003CN>` | Web UI port (default: 8765) |\n| `--host \u003CHOST>` | Web UI host (default: 127.0.0.1) |\n| `-m \u003Cmodel>`, `--model \u003Cmodel>` | Override default model |\n| `-d \u003Cpath>`, `--dir \u003Cpath>` | Working directory |\n| `--context \u003CN>` | Context window size (tokens) |\n\n## REPL Commands\n\n| Command | Description |\n|---------|-------------|\n| `\u002Fhelp` | Show help |\n| `\u002Fapi` | Show current API config |\n| `\u002Fapi type \u003Copenai\\|anthropic>` | Set API type |\n| `\u002Fapi url \u003CURL>` | Set API base URL |\n| `\u002Fapi key \u003CKEY>` | Set API key |\n| `\u002Fmodel \u003Cname>` | Set model name |\n| `\u002Fcontext [tokens]` | Show\u002Fset context window size |\n| `\u002Fstatus` | Show session status |\n| `\u002Fclear` | Clear conversation history |\n| `\u002Fcompact` | Compact conversation context |\n| `\u002Fmemory` | Show memory status |\n| `\u002Fskill` | List available skills |\n| `\u002Ftodos` | Show todo list |\n| `\u002Ffast` | Toggle fast mode |\n| `\u002Fexit` | Exit |\n\n## Supported APIs\n\n- **OpenAI Compatible** — OpenAI official API & all `\u002Fv1\u002Fchat\u002Fcompletions` compatible services (OpenAI, Azure OpenAI, DeepSeek, Groq, etc.)\n- **Anthropic Compatible** — Anthropic official API & all `\u002Fv1\u002Fmessages` compatible services (Claude series)\n\n## Configuration Storage\n\n```\n~\u002F.autorun\u002F\n├── config.json       # API key, URL, type, model, context window\n├── history           # Input history\n├── memory\u002F           # Memory system\n└── skills\u002F           # User skills\n```\n\n- Environment variables take priority over config.json\n- API key is stored as plain text (same as `.env` files)\n\n## OCR (Optical Character Recognition)\n\nAutoRUN includes a built-in OCR tool based on [Falcon-OCR](https:\u002F\u002Fhuggingface.co\u002Ftiiuae\u002FFalcon-OCR), a lightweight 0.3B vision-language model that runs entirely on your local machine.\n\n### Features\n\n- **Plain OCR** — Full-page text extraction for documents, screenshots, receipts, slides\n- **Layout-aware OCR** — Detects text regions (tables, formulas, headers, captions) and extracts per-region text. Best for complex multi-column documents and academic papers.\n- **Fully local** — All inference runs on your GPU\u002FCPU. No image data leaves your machine.\n- **Token saving** — Non-multimodal models can \"see\" images by calling OCR first, then analyzing the extracted text.\n- **Auto-download** — Model (~600MB) is downloaded on first use. Supports Chinese mirrors (`HF_ENDPOINT=https:\u002F\u002Fhf-mirror.com`).\n\n### Usage\n\nThe OCR tool is automatically available in the REPL and Web UI. The AI will call it when you ask it to read text from an image:\n\n```\n> 帮我把这张截图里的文字提取出来\n```\n\nOr use with explicit mode:\n\n```\n> 用 layout 模式 OCR 这篇论文的截图\n```\n\n### Configuration\n\n| Environment Variable | Description | Default |\n|----------------------|-------------|---------|\n| `HF_ENDPOINT` | HuggingFace mirror (set to `https:\u002F\u002Fhf-mirror.com` for China) | `https:\u002F\u002Fhuggingface.co` |\n| `AUTORUN_OCR_LOCAL_DIR` | Local model directory (skip download) | Auto-download |\n| `AUTORUN_OCR_DEVICE` | Inference device (`cuda`\u002F`cpu`) | Auto-detect |\n| `AUTORUN_OCR_DTYPE` | Model dtype (`float32`\u002F`float16`\u002F`bfloat16`) | `float32` |\n\n### Requirements\n\n- CUDA GPU recommended (CPU works but slower)\n- Dependencies auto-installed on first use: `torch`, `transformers`, `torchvision`, `huggingface_hub`, `safetensors`, `tokenizers`, `Pillow`\n\n## Project Structure\n\n```\nAutoRUN_v1\u002F\n├── cli.py              # Entry point (autorun command)\n├── main.py             # CLI argument parsing & routing\n├── commands.py         # Slash command system\n├── query.py            # Core query loop\n├── query_engine.py     # High-level conversation orchestrator\n├── pyproject.toml      # Package metadata & dependencies\n├── requirements.txt    # Pip dependencies (legacy)\n├── install.bat         # Windows one-click installer\n├── install.sh          # macOS\u002FLinux one-click installer\n├── api\u002F                # API clients (OpenAI \u002F Anthropic)\n├── context\u002F            # Context building (git, environment, etc.)\n├── messages\u002F           # Message types & utilities\n├── prompts\u002F            # System prompts\n├── services\u002F           # Compaction, LSP, etc.\n├── skills\u002F             # Skill discovery & loading\n├── state\u002F              # Session state management\n├── tools\u002F              # Tool registry & execution\n│   └── ocr_engine\u002F     # Falcon-OCR inference engine (Apache-2.0, from TII)\n├── ui\u002F                 # CLI \u002F Web UI\n│   ├── cli\u002F            # prompt_toolkit \u002F Textual REPL\n│   └── web\u002F            # FastAPI Web server + frontend\n└── utils\u002F              # Config, tokens, env utils\n```\n\n---\n\n\u003Ca id=\"chinese\">\u003C\u002Fa>\n\n# AutoRUN v1\n\n通用 AI 编程助手，支持 OpenAI 和 Anthropic 兼容 API。\n\n## 安装\n\n需要 Python 3.8+。\n\n### 一键安装（推荐）\n\n**Windows:**\n```cmd\ninstall.bat\n```\n\n**macOS \u002F Linux:**\n```bash\n.\u002Finstall.sh\n```\n\n自动创建虚拟环境、安装依赖，并配置 `autorun` 命令。\n\n### 手动安装\n\n```bash\ncd AutoRUN_v1\npip install -e .\n```\n\n安装后，`autorun` 命令在终端中全局可用：\n\n```bash\nautorun --version\n```\n\n### 备选：pip + requirements.txt\n\n```bash\ncd AutoRUN_v1\npip install -r requirements.txt\npython cli.py\n```\n\n## 快速开始\n\n```bash\nautorun\n```\n\n首次运行会自动引导你配置 API：\n\n```\n╔══════════════════════════════════════════╗\n║     AutoRUN v1.0 — 首次设置             ║\n╚══════════════════════════════════════════╝\n\n选择 API 类型 [1=OpenAI, 2=Anthropic] [默认 openai]: 1\nAPI URL [默认 https:\u002F\u002Fapi.openai.com]:\nAPI Key: sk-xxxxxxxxxxxx\n模型名称 [默认 gpt-4o]:\n\n✓ 配置已保存！\n```\n\n配置保存在 `~\u002F.autorun\u002Fconfig.json`。\n\n## API 配置\n\n三种配置方式，按优先级排序：\n\n### 方式一：首次运行向导\n\n```bash\nautorun\n```\n\n首次运行自动触发，或通过 `autorun --setup` 随时重新配置。\n\n### 方式二：环境变量\n\n**Windows (PowerShell):**\n```powershell\n$env:AUTORUN_API_TYPE = \"openai\"\n$env:AUTORUN_API_URL  = \"https:\u002F\u002Fapi.openai.com\"\n$env:AUTORUN_API_KEY  = \"sk-xxxxxxxxxxxx\"\n$env:AUTORUN_MODEL    = \"gpt-4o\"\nautorun\n```\n\n**Windows (CMD):**\n```cmd\nset AUTORUN_API_TYPE=openai\nset AUTORUN_API_URL=https:\u002F\u002Fapi.openai.com\nset AUTORUN_API_KEY=sk-xxxxxxxxxxxx\nset AUTORUN_MODEL=gpt-4o\nautorun\n```\n\n**macOS \u002F Linux:**\n```bash\nexport AUTORUN_API_TYPE=openai\nexport AUTORUN_API_URL=https:\u002F\u002Fapi.openai.com\nexport AUTORUN_API_KEY=sk-xxxxxxxxxxxx\nexport AUTORUN_MODEL=gpt-4o\nautorun\n```\n\n### 方式三：启动后在 REPL 中设置\n\n```bash\nautorun\n```\n\n进入 REPL 后输入：\n```\n\u002Fapi type openai                          # API 类型: openai 或 anthropic\n\u002Fapi url https:\u002F\u002Fapi.openai.com           # API 基础 URL\n\u002Fapi key sk-xxxxxxxxxxxx                  # 你的 API 密钥\n\u002Fmodel gpt-4o                             # 模型名称\n```\n\n配置保存在 `~\u002F.autorun\u002Fconfig.json`，下次启动自动加载。\n\n## 运行方式\n\n### CLI REPL（默认交互模式）\n\n```bash\nautorun\n```\n\n### Web UI\n\n```bash\nautorun --web\n```\n\n默认访问 http:\u002F\u002F127.0.0.1:8765\n\n```bash\n# 自定义端口和地址\nautorun --web --port 8080 --host 0.0.0.0\n```\n\n### 管道模式\n\n```bash\necho \"解释一下这段代码\" | autorun --print\nautorun --print \"解释一下这段代码\"\n```\n\n### 重新配置 API\n\n```bash\nautorun --setup\n```\n\n## 命令行选项\n\n| 选项 | 说明 |\n|------|------|\n| `--version`, `-V`, `-v` | 显示版本号 |\n| `--setup` | 重新运行 API 设置向导 |\n| `--web` | 启动 Web UI 服务器 |\n| `--print`, `-p` | 管道\u002F打印模式（非交互） |\n| `--port \u003CN>` | Web UI 端口（默认: 8765） |\n| `--host \u003CHOST>` | Web UI 地址（默认: 127.0.0.1） |\n| `-m \u003Cmodel>`, `--model \u003Cmodel>` | 覆盖默认模型 |\n| `-d \u003Cpath>`, `--dir \u003Cpath>` | 工作目录 |\n| `--context \u003CN>` | 上下文窗口大小（tokens） |\n\n## REPL 常用命令\n\n| 命令 | 说明 |\n|------|------|\n| `\u002Fhelp` | 显示帮助 |\n| `\u002Fapi` | 查看当前 API 配置 |\n| `\u002Fapi type \u003Copenai\\|anthropic>` | 设置 API 类型 |\n| `\u002Fapi url \u003CURL>` | 设置 API 基础 URL |\n| `\u002Fapi key \u003CKEY>` | 设置 API 密钥 |\n| `\u002Fmodel \u003Cname>` | 设置模型名称 |\n| `\u002Fcontext [tokens]` | 显示\u002F设置上下文窗口大小 |\n| `\u002Fstatus` | 显示会话状态 |\n| `\u002Fclear` | 清除对话历史 |\n| `\u002Fcompact` | 压缩对话上下文 |\n| `\u002Fmemory` | 显示记忆系统状态 |\n| `\u002Fskill` | 列出可用 skill |\n| `\u002Ftodos` | 显示 todo 列表 |\n| `\u002Ffast` | 切换快速模式 |\n| `\u002Fexit` | 退出 |\n\n## 支持的 API\n\n- **OpenAI 兼容** — OpenAI 官方 API 及所有兼容 `\u002Fv1\u002Fchat\u002Fcompletions` 的服务（OpenAI, Azure OpenAI, DeepSeek, 硅基流动, Groq 等）\n- **Anthropic 兼容** — Anthropic 官方 API 及所有兼容 `\u002Fv1\u002Fmessages` 的服务（Claude 系列）\n\n## 配置存储\n\n```\n~\u002F.autorun\u002F\n├── config.json       # API key, URL, type, model, context window\n├── history           # 输入历史\n├── memory\u002F           # 记忆系统\n└── skills\u002F           # 用户 skill\n```\n\n- 环境变量优先级高于 config.json\n- API key 明文存储（与 `.env` 文件方式相同）\n\n## OCR (光学字符识别)\n\nAutoRUN 内置基于 [Falcon-OCR](https:\u002F\u002Fhuggingface.co\u002Ftiiuae\u002FFalcon-OCR) 的 OCR 工具，0.3B 参数的轻量视觉语言模型，完全在本地运行。\n\n### 功能\n\n- **全页 OCR (plain)** — 全文提取，适合文档、截图、收据、幻灯片\n- **布局感知 OCR (layout)** — 自动检测文字区域（表格、公式、标题、页眉等）并分别提取。适合复杂多栏文档和学术论文\n- **完全本地** — 所有推理在 GPU\u002FCPU 上完成，图片数据不会离开本机\n- **节省 Token** — 非多模态模型可先调用 OCR 提取文字，再对文字分析推理\n- **自动下载** — 首次使用时自动下载模型（约 600MB），支持国内镜像加速（`HF_ENDPOINT=https:\u002F\u002Fhf-mirror.com`）\n\n### 使用方式\n\nOCR 工具在 REPL 和 Web UI 中自动可用。当你要求 AI 从图片中提取文字时，AI 会自动调用：\n\n```\n> 帮我把这张截图里的文字提取出来\n```\n\n或显式指定模式：\n\n```\n> 用 layout 模式 OCR 这篇论文的截图\n```\n\n### 配置\n\n| 环境变量 | 说明 | 默认值 |\n|---------|------|--------|\n| `HF_ENDPOINT` | HuggingFace 镜像（国内设为 `https:\u002F\u002Fhf-mirror.com`） | `https:\u002F\u002Fhuggingface.co` |\n| `AUTORUN_OCR_LOCAL_DIR` | 本地模型目录（跳过下载） | 自动下载 |\n| `AUTORUN_OCR_DEVICE` | 推理设备 (`cuda`\u002F`cpu`) | 自动检测 |\n| `AUTORUN_OCR_DTYPE` | 模型数据类型 (`float32`\u002F`float16`\u002F`bfloat16`) | `float32` |\n\n### 依赖\n\n- 推荐 CUDA GPU（CPU 可用但较慢）\n- 依赖在首次使用时自动安装：`torch`, `transformers`, `torchvision`, `huggingface_hub`, `safetensors`, `tokenizers`, `Pillow`\n\n## 项目结构\n\n```\nAutoRUN_v1\u002F\n├── cli.py              # 入口点（autorun 命令）\n├── main.py             # CLI 参数解析和路由\n├── commands.py         # 斜杠命令系统\n├── query.py            # 核心查询循环\n├── query_engine.py     # 高层对话编排\n├── pyproject.toml      # 包元数据和依赖声明\n├── requirements.txt    # Pip 依赖（兼容旧方式）\n├── install.bat         # Windows 一键安装脚本\n├── install.sh          # macOS\u002FLinux 一键安装脚本\n├── api\u002F                # API 客户端 (OpenAI \u002F Anthropic)\n├── context\u002F            # 上下文构建 (git, 环境等)\n├── messages\u002F           # 消息类型和工具\n├── prompts\u002F            # 系统提示词\n├── services\u002F           # 压缩、LSP 等服务\n├── skills\u002F             # 技能加载\n├── state\u002F              # 会话状态\n├── tools\u002F              # 工具注册和执行\n│   └── ocr_engine\u002F     # Falcon-OCR 推理引擎 (Apache-2.0, from TII)\n├── ui\u002F                 # CLI \u002F Web UI\n│   ├── cli\u002F            # prompt_toolkit \u002F Textual REPL\n│   └── web\u002F            # FastAPI Web 服务器 + 前端\n└── utils\u002F              # 配置、token 等工具\n```\n\n---\n\n## 致谢 \u002F Acknowledgments\n\n本项目的 OCR 功能基于 [Falcon-Perception](https:\u002F\u002Fgithub.com\u002Ftiiuae\u002FFalcon-Perception) 项目，使用 [Falcon-OCR](https:\u002F\u002Fhuggingface.co\u002Ftiiuae\u002FFalcon-OCR) 模型。\n\nFalcon-Perception 和 Falcon-OCR 由 [Technology Innovation Institute (TII)](https:\u002F\u002Ftii.ae\u002F), UAE 开发，以 Apache-2.0 许可证开源发布。\n\n> **引用 \u002F Citation:**\n> Bevli et al., *Falcon-Perception*, arXiv:2603.27365, 2026.\n\nThe OCR engine source code from Falcon-Perception is integrated under `tools\u002Focr_engine\u002F`, licensed under Apache-2.0.\n\n---\n\n\u003C!-- Language Switcher -->\n\u003Cp align=\"center\">\n  \u003Ca href=\"#english\">\u003Cb>English\u003C\u002Fb>\u003C\u002Fa> |\n  \u003Ca href=\"#chinese\">\u003Cb>中文\u003C\u002Fb>\u003C\u002Fa>\n\u003C\u002Fp>\n","AutoRUN 是一个支持 OpenAI 和 Anthropic 兼容 API 的通用 AI 编码助手。该项目使用 Python 语言开发，通过命令行或 Web 界面提供代码生成和解释等功能。用户可以通过简单的安装脚本快速部署，并通过交互式向导、环境变量或直接在 REPL 中配置所需的 API 信息。AutoRUN 适用于需要自动化代码生成、理解和优化的场景，如软件开发辅助、代码审查以及技术文档编写等。其简洁的命令行选项设计使得无论是初学者还是经验丰富的开发者都能轻松上手。",2,"2026-06-01 03:51:50","CREATED_QUERY"]