Noundry Andy Standalone CLI

An AI-powered development assistant for the Noundry platform with embedded local LLM. Similar to Claude Code but specialized for Noundry web application development with offline capabilities.

Features

• Interactive Chat Mode: Engage in conversational development with Andy
• Local LLM Integration: Embedded CodeLlama model with Noundry-specific prompts
• File System Operations: Read, write, and edit files with intelligent context
• Command Execution: Run shell commands and scripts
• Offline Mode: Works without internet connection
• Rich CLI Interface: Beautiful terminal UI powered by Spectre.Console
• 20+ Templates: Comprehensive code generation templates for Noundry

Installation

From NuGet (when published):

dotnet tool install -g Noundry.Andy.Standalone

Build from Source:

cd Andy.Standalone
dotnet pack
dotnet tool install -g --add-source ./nupkg Noundry.Andy.Standalone

Model Setup

1. Download a compatible GGUF model:

   huggingface-cli download TheBloke/CodeLlama-7B-Instruct-GGUF \
     codellama-7b-instruct.Q4_K_M.gguf \
     --local-dir ./models
   

2. Configure the model path:

   nd config set model.path ./models/codellama-7b-instruct.Q4_K_M.gguf
   

Usage

Command: nd (Noundry Development)

Interactive Chat

nd chat

Run a Single Command

nd run "Create a contact form with validation"

Configure Settings

nd config                              # Show all settings
nd config set model.path /path/to/model.gguf
nd config set model.temperature 0.7
nd config list                         # List all settings

Commands in Chat Mode

• exit - Exit the chat session
• clear - Clear conversation context
• help - Show available commands
• /file <path> - Read and analyze a file
• /edit <path> - Edit a file
• /run <command> - Execute a shell command
• /search <pattern> - Search for files or content

Configuration

The tool can be configured through:

1. appsettings.json file
2. Environment variables (prefix with ANDY_)
3. Command-line options

Model Settings

• Model:Path - Path to the GGUF model file
• Model:ContextSize - Context window size (default: 4096)
• Model:Temperature - Generation temperature (default: 0.7)
• Model:MaxTokens - Maximum tokens per response (default: 2048)

Tool Settings

• Tools:EnableFileSystem - Enable file system operations
• Tools:EnableProcessExecution - Enable command execution
• Tools:MaxFileSize - Maximum file size for operations
• Tools:AllowedFileExtensions - List of allowed file extensions

Templates

Andy Standalone includes 20 comprehensive templates:

• Core: Forms, Data Tables, Services, Controllers
• Authentication: Login, Registration
• UI Components: Dashboards, Modals, Search, File Upload, Wizards, Navigation, Settings, Notifications, Profiles, Comments, Calendars, Reports
• Admin: Admin Panels, API Documentation

Templates use Noundry TagHelpers like <noundry-text-input>, <noundry-button>, <noundry-card>, etc.

GPU Acceleration

For faster generation with CUDA GPUs, modify Andy.Standalone.csproj:

<!-- Replace CPU backend with CUDA -->
<PackageReference Include="LLamaSharp.Backend.Cuda12" Version="0.13.0" />

Then set GPU layers:

nd config set model.gpulayers 35

Development

Project Structure

NoundryAndy/
├── Commands/       # CLI command implementations
├── Services/       # Core services (Model, FileSystem, Process)
├── Tools/          # Tool implementations
├── UI/             # UI components and services
├── Core/           # Core abstractions and models
└── Configuration/  # Configuration management

Adding New Tools

1. Create a new tool class in the Tools directory
2. Register it in ToolExecutor.cs
3. Add tool definition in GetAvailableTools()

Extending Model Capabilities

1. Modify LlamaModelService.cs for model interaction
2. Update ConversationService.cs for context management
3. Adjust prompts in appsettings.json

License

MIT License

Contributing

Contributions are welcome! Please submit pull requests with:

• Clear commit messages
• Updated documentation
• Unit tests for new features

Support

For issues or questions, please open an issue on GitHub.