HyperChat

# HyperChat > 🌟 **Local AI Agent Platform** - Pioneering the concept of AI as Code, giving each project its own dedicated AI brain HyperChat is a revolutionary **local AI Agent platform** that allows AI capabilities to be fully localized, transferable, and version-controlled through **configuration file-driven** approach. Say goodbye to cloud dependencies and have your own project-level AI expert. ### 🎯 Key Features - 🏠 **Fully Localized**: Data stays local, privacy and security are under control - 🧠 **Agent Memory**: AI Agents understand and remember your project context - 📁 **Configuration as Code**: All AI capabilities are configured through files, supporting Git management - 🔧 **Deep Tool Integration**: MCP protocol support, can directly operate the local file system - 📦 **One-Click Migration**: Complete `.hyperchat/` configuration directory, migrate with the project - 📂 **Multiple @File Paths**: Intelligently identify multiple file references, supporting complex file operations ✨**New Feature** - 🎯 **Agent Custom Commands**: Markdown template-driven shortcut command system ✨**New Feature** [](https://github.com/BigSweetPotatoStudio/HyperChat/actions/workflows/build.yml) [](https://www.npmjs.com/package/@dadigua/hyperchat) [](https://npm-stat.com/charts.html?package=@dadigua/hyperchat) ## 🎯 Project Vision **[HyperChat 1.0](./archive/README.1.md)** is a fully hand-written project and is being migrated to 2.0. **HyperChat 2.0** Let's **Vibe Coding** together, welcome to use Claude Code and GitHub Copilot and other AI tools to develop together. ## 🎯 Two-Tier Architecture Design > **Innovative Dual-Mode Architecture** - Optimizing the user experience based on different usage scenarios ### 🌐 Web Frontend: Multi-Workspace Collaboration Center **Design Concept**: Project-level collaboration, unified resource management - **🗂️ Multi-Workspace Tab Management**: Open multiple project workspaces at the same time, switch with one click - **👥 Team Collaboration Optimization**: Workspace-level Agent collection, MCP service pool - **📊 Visual Management Interface**: Graphical configuration, real-time monitoring, data statistics - **💼 Applicable Scenarios**: Project development, team collaboration, workspace management, visual operation ### 💻 CLI Frontend: Agent-First Rapid Interaction **Design Concept**: Agent-centric, quickly launch a single intelligent agent - **⚡ Quick Start**: Directly select an Agent, load MCP tools from the workspace as needed - **🎯 Focus on Dialogue**: Agent-exclusive memory, context, chat history - **🔧 Flexible Toolchain**: Agent built-in MCP tools, fall back to workspace shared resources - **🚀 Applicable Scenarios**: Quick dialogue, automated scripts, command-line workflows, CI/CD integration ### 📋 Architecture Comparison | Feature | 🌐 Web Multi-Workspace Mode | 💻 CLI Agent-First Mode | |---------|---------------------|----------------------| | **Core Concept** | | Design Center | 🗂️ Workspace Collaboration Center | 🤖 Agent Direct Interaction | | Resource Management | Workspace Unified MCP Pool | Agent On-Demand MCP Startup | | Usage | Multi-Tab Concurrent Management | Single Agent Quick Start | | **Interface Experience** | | Interface Style | 🖥️ Modern Web Interface | 📟 Command Line + 🎨 TUI | | Interaction Mode | Mouse Click + Form Operation | Keyboard Input + Command Parameters | | Real-Time Update | ✅ SSE Streaming Push | ✅ Terminal Streaming Output | | **Applicable Scenarios** | | Main Use | Project Development, Team Collaboration | Quick Dialogue, Script Integration | | Usage Environment | Desktop Browser, Development IDE | Terminal, Server, CI/CD | | Workflow | Long-Term Project Management | Temporary Dialogue Processing | ## 🚀 Quick Experience ### ⌨️ Command Line Quick Start ```bash # Global installation npm install -g @dadigua/hyperchat # Or run directly npx -y @dadigua/hyperchat ``` **Quickly configure environment variables**: ```bash # Basic configuration - Set the default AI model export HyperChat_API_KEY=your-api-key # API key export HyperChat_API_URL=your-api-url # API endpoint URL export HyperChat_AI_Provider=openai # AI provider (openai/claude/gemini/kimi/qwen, etc.) export HyperChat_AI_Model=gpt-4o # Default model name # Then use directly hyperchat "Hello, world!" # Use the configured default model ``` ### 🌐 Web Multi-Workspace Mode Usage Example ```bash # Start the multi-workspace Web interface hyperchat serve # Access: http://localhost:16100 # Web interface features: # ✅ Multi-workspace tab management # ✅ Each tab has an independent Agent collection, MCP service, chat history # ✅ Visual configuration and real-time monitoring # ✅ Team collaboration and project management ``` ### 💻 CLI Agent-First Mode Usage Example ```bash # 🚀 Agent-first quick start - Core features hyperchat agent list # Discover available Agents (global + workspace) hyperchat agent mybot "Hello" # 🎯 Directly start the Agent, load MCP on demand hyperchat agent mybot chat # 🎯 Agent exclusive dialogue session # Quick AI chat (using the default Agent) hyperchat "Hello, how are you today?" # Chat directly with the default model hyperchat chat "Write a Python script" # Chat command hyperchat chat # Interactive chat mode # 📁 Intelligent file processing - Multiple @ symbol support ✨New Feature hyperchat "Analyze the code quality of @./src/index.ts" hyperchat "Compare @./package.json and @./yarn.lock" hyperchat "Please compare the structure of @./src/components/ and @./docs/" # 🎯 Agent custom commands - Quickly enter professional prompts ✨New Feature hyperchat agent coder "/bug-fix @./src/login.ts Login function is abnormal" hyperchat agent coder "/review @./src/api/user.js" hyperchat agent coder "/optimize This code has poor performance" # Agent Management (in the current workspace or globally) hyperchat agent create mybot # Create a new Agent hyperchat agent delete mybot # Delete Agent # Workspace Management hyperchat workspace create # Create a workspace in the current directory # Global options and workspace specification hyperchat chat --workspace /path/to/project # Use a specific workspace hyperchat --verbose chat "Hello" # Detailed log hyperchat --help # Show help # CLI mode advantages: # ⚡ Agent starts directly, no need to switch interfaces # 🔧 Load MCP tools from the workspace on demand # 💾 Agent exclusive memory and chat history # 🚀 Suitable for script integration and automation ``` ### 🔄 Dual-Mode Collaborative Work ```bash # Scenario 1: Web during development hyperchat serve # Start the Web interface to manage the project # Scenario 2: CI/CD automation uses CLI hyperchat agent test-runner "Run all tests and generate a report" # Scenario 3: Team collaboration uses Web # Manage multiple project workspaces in the Web interface, configure shared Agents and MCP services ``` ### 🔧 Environment Variable Configuration HyperChat 2.0 implements a powerful **5-layer priority environment variable system** to make configuration management more flexible: **Priority Order** (from low to high): 1. **Default Value** - Built-in default configuration in the code 2. **process.env** - System environment variables 3. **Global .env** - `~/Documents/HyperChat/.env` 4. **Workspace .env** - `.env` file in the project directory 5. **CLI Parameters** - Parameters passed in from the command line (highest priority) **Supported Core Environment Variables**: ```bash # Quick configuration - Default AI model HyperChat_API_KEY=your-api-key # Default API key HyperChat_API_URL=your-api-url # Default API endpoint HyperChat_AI_Provider=openai # Default AI provider HyperChat_AI_Model=gpt-4o # Default model name # Service configuration HYPERCHAT_WEB_PASSWORD=your-web-password # Web interface access password HYPERCHAT_PORT=16100 # Web service port HYPERCHAT_HOST=localhost # Service binding address # Interface configuration HYPERCHAT_LANGUAGE=zh # Interface language (zh/en) HYPERCHAT_LOG_LEVEL=info # Log level # Custom API endpoint HYPERCHAT_OPENAI_BASE_URL=https://api.openai.com/v1 HYPERCHAT_CLAUDE_BASE_URL=https://api.anthropic.com ``` **Usage Example**: ```bash # Method 1: Quickly configure the default model export HyperChat_API_KEY=sk-1234567890 export HyperChat_AI_Provider=openai export HyperChat_AI_Model=gpt-4o hyperchat "Hello" # Use the default configuration directly # Method 2: Web service configuration export HYPERCHAT_WEB_PASSWORD=mypassword hyperchat serve # Method 3: Project .env file echo "HyperChat_API_KEY=your-key" > .env echo "HyperChat_AI_Provider=claude" >> .env hyperchat chat # Method 4: CLI parameters (highest priority) hyperchat serve --password=clipass # Method 5: Global configuration file echo "HyperChat_API_KEY=global-key" > ~/Documents/HyperChat/.env echo "HyperChat_AI_Provider=gemini" >> ~/Documents/HyperChat/.env ``` ## 🛠️ Technical Architecture ### 🎯 Two-Tier Architecture Technical Implementation HyperChat 2.0 adopts a **two-tier architecture design** to provide the most optimized experience based on different usage scenarios: #### 🌐 Web Layer: Workspace-Centric Architecture ```typescript // Web Frontend: Workspace Unified MCP Management const workspace = workspaceManager.get(workspacePath); const mcpManager = workspace.getMcpManager(); const client = mcpManager.getClient(clientName); ``` **Technical Features**: - 🗂️ **Multi-Workspace Concurrent Management**: `WorkspaceManager-enhanced` supports multiple workspace instance caching - 🔌 **Workspace-Level MCP Service Pool**: Unified MCP client management - 📊 **Real-Time Data Synchronization**: SSE streaming push, multi-tab independent state management - 🔄 **Configuration Merging Mechanism**: Intelligent merging of global configuration + workspace configuration #### 💻 CLI Layer: Agent-First Architecture ```typescript // CLI Frontend: Agent Exclusive MCP Access const agentInstance = workspace.getAgentInstance(agentName); const client = agentInstance.getMCPClient(clientName); // Fall back to workspace shared MCP (if needed) ``` **Technical Features**: - ⚡ **Agent Direct Startup**: Omit workspace initialization, directly access Agent instance - 🔧 **Toolchain On-Demand Loading**: Progressive loading from Agent built-in tools to workspace shared resources - 💾 **Exclusive Context**: Agent independent memory, chat history, configuration management - 🚀 **Ultra-Fast Response**: No UI rendering overhead, suitable for scripts and automation ``` HyperChat Dual-Layer Architecture/ ├── packages/ │ ├── shared/ # Shared code and type definitions │ ├── core/ # Node.js core service + CLI Agent-first layer │ │ ├── src/cli/ # 💻 CLI Agent direct interaction layer │ │ ├── src/workspace/ # Dual-layer workspace management system │ │ │ ├── workspaceManager-enhanced.mts # 🌐 Multi-workspace concurrent management │ │ ├── src/mcp/ # MCP protocol implementation │ │ └── src/commands/ # Web API command system │ ├── web/ # 🌐 React Web multi-workspace layer │ │ ├── src/pages/workspace/ # Multi-workspace management components │ │ │ ├── WorkspaceManage.tsx # Multi-tab manager │ │ │ └── workspace.tsx # Single workspace instance component │ │ └── src/hooks/useChatStream.ts # workspacePath parameterization │ └── electron/ # Electron application packaging └── docs/ # Dual-layer architecture documentation ``` ### 💼 Dual-Layer Architecture Configuration Management ``` 🌐 Web Multi-Workspace Mode Configuration: Project Directory/ ├── .hyperchat/ # Workspace unified configuration directory │ ├── mcp.json # 🔌 Workspace-level MCP service pool │ ├── ai_models.json # 🤖 Workspace AI model configuration │ ├── .env # Workspace environment variables │ └── agents/ # 🗂️ Workspace Agent collection │ ├── project-assistant/ # Project-specific Agent │ │ ├── agent.yaml # Agent configuration │ │ ├── memory.md # Project context memory │ │ └── chatlogs/ # Team dialogue history │ └── code-reviewer/ # Code review Agent ├── .git/ # Version control └── package.json # Project configuration 💻 CLI Agent-First Mode Access: Global Configuration/ ~/Documents/HyperChat/ .hyperchat/ ├── mcp.json # 🌍 Global MCP service pool ├── .env # Global environment variables └── agents/ # 🚀 Global Agent library ├── personal-assistant/ # Personal assistant Agent │ ├── agent.yaml # Agent configuration │ ├── memory.md # Personal context memory │ └── chatlogs/ # Personal dialogue history └── code-expert/ # Professional code Agent ``` **Core Advantages of Dual-Layer Configuration**: - 🌐 **Web Mode**: Workspace-centric, suitable for project-level management and team collaboration - 💻 **CLI Mode**: Agent-centric, supports the use of personal tools across projects - 🔄 **Intelligent Merging**: 5-layer priority management of global configuration + workspace configuration - 💾 **Data Isolation**: Project-level and personal-level data are completely separated and do not interfere with each other ``` ## 🌟 Revolutionary Advantages of AI as Code ### 🔄 Intelligent Management Under Dual-Layer Architecture #### 🌐 Web Mode: Team Collaboration Version Control ```bash # 📁 Project-level AI configuration version management git add .hyperchat/agents/project-assistant/ git commit -m "Add project-specific AI assistant" git push origin feature/project-ai # 👥 Team AI best practice sharing git clone https://github.com/team/ai-workspace-templates.git cp -r ai-workspace-templates/react-fullstack/.hyperchat ./ # 🔄 Project AI configuration rollback git checkout HEAD~1 -- .hyperchat/ ``` #### 💻 CLI Mode: Personal Tool Management ```bash # 🚀 Quickly deploy personal Agent hyperchat agent create personal-coder --template ~/ai-templates/ ``` # 🌍 Use Personal Agent Across Projects cd /project-a && hyperchat agent personal-coder "分析这个项目" cd /project-b && hyperchat agent personal-coder "分析这个项目" # 💾 Agent Memory and Context Auto-Switching # personal-coder will remember the characteristics and context of different projects ``` ### Agent Configuration and Custom Commands ✨**New Feature** #### Agent Basic Configuration ```yaml # .hyperchat/agents/project-assistant/agent.yaml name: "项目助手" description: "Specifically designed for this project React + Node.js full-stack assistant" modelKey: "claude-3-5-sonnet" isConfirmCallTool: false # 🔌 Use workspace-level MCP service pool allowMCPs: ["filesystem", "git", "npm", "database"] prompt: | You are the exclusive AI assistant for this project, familiar with: - Project architecture: React + TypeScript + Node.js - Business domain: E-commerce platform development - Team specifications: ESLint + Prettier + Jest Please provide professional advice based on the project context. tags: ["project", "fullstack", "ecommerce"] ``` #### Agent Custom Command Directory Structure ✨**New Feature** ``` .hyperchat/agents/project-assistant/ ├── agent.yaml # Agent Configuration ├── memory.md # Agent Memory ├── commands/ # 🎯 Custom Command Directory │ ├── bug-fix.md # Bug Fix Command │ ├── review.md # Code Review Command │ ├── explain.md # Code Explanation Command │ ├── test.md # Test Case Command │ ├── optimize.md # Code Optimization Command │ └── document.md # Documentation Generation Command └── chatlogs/ # Chat Logs ``` #### Command Template Example (Pure Markdown) ```markdown # commands/bug-fix.md Please help me fix the bug in the following code: $ARG Requirements: 1. Carefully analyze the root cause of the problem 2. Provide a detailed repair plan 3. Give the complete code after the fix 4. Explain the principle and best practices of the fix 5. Provide suggestions for preventing similar problems ``` ```markdown # commands/review.md Please conduct a comprehensive code review of the following code: $ARG Review points: 1. **Code Quality**: Readability, maintainability, naming conventions 2. **Performance Issues**: Algorithm efficiency, memory usage, potential bottlenecks 3. **Security Risks**: Input validation, permission control, data leakage risks 4. **Best Practices**: Design patterns, architectural principles, team specifications 5. **Improvement Suggestions**: Specific optimization plans and alternative implementations Please provide specific modification suggestions and code examples. ``` #### Usage Example ```bash # Use a custom command to fix a bug hyperchat agent project-assistant "/bug-fix @./src/login.ts 用户登录后跳转异常" # Actual content sent to AI: # Please help me fix the bug in the following code: # @./src/login.ts 用户登录后跳转异常 # Requirements: # 1. Carefully analyze the root cause of the problem # 2. Provide a detailed repair plan # 3. Give the complete code after the fix # 4. Explain the principle and best practices of the fix # 5. Provide suggestions for preventing similar problems # Use the code review command hyperchat agent project-assistant "/review @./src/components/UserProfile.tsx" # Use in interactive chat hyperchat agent project-assistant chat > /bug-fix This function has a memory leak problem > /optimize @./src/utils/dataProcessor.js ``` ```yaml # ~/Documents/HyperChat/.hyperchat/agents/personal-coder/agent.yaml name: "个人编程助手" description: "Cross-project general programming assistant, supports multiple technology stacks" modelKey: "gpt-4o" isConfirmCallTool: true # ⚡ Agent Exclusive MCP Tools + Fallback to Workspace allowMCPs: ["terminal", "browser", "calculator"] prompt: | You are my personal programming assistant, good at: - Multi-language development: Python, JavaScript, Go, Rust - Architecture design and code review - Rapid prototype development and problem solving Adjust the answering style according to the context of different projects. tags: ["personal", "general", "cross-project"] ``` ### 🤖 AI Model Recommendation | Model Provider | Recommendation Level | Features | |------------|----------|----------| | Claude | ⭐⭐⭐⭐⭐⭐ | 最强 | | Kimi k2 | ⭐⭐⭐⭐⭐ | 很不错 | ## 🔧 Development Guide ### 💻 Local Development ```bash # Clone the project git clone https://github.com/BigSweetPotatoStudio/HyperChat.git cd HyperChat # Install dependencies npm install cd packages/electron && npm install cd packages/web && npm install cd ../.. # Start the development server npm run dev ## 🤝 Community Communication - [Telegram](https://t.me/dadigua001) - [QQ 群](https://qm.qq.com/cgi-bin/qm/qr?k=KrNWdu5sp7H3ves3ZPSd7ppKjQiPrAvZ&jump_from=webapi&authKey=xnW+Lcgk5KLh5NPh3lU0ddz9CFDbXgvjEy35wsYipUrCsqXFcqlvM5Yorh6jkGim) ## 📄 Disclaimer This project is for learning and communication purposes only. The project developers are not responsible for any operations (such as web crawling) performed using this project. ## 📜 License This project uses an open-source license. See the [LICENSE](LICENSE) file for details.