LLM CLI

A command-line interface for interacting with various Large Language Models (LLMs) with built-in tool support for development tasks.

#LLM #DeepSeek #Ollama #ChatGPT #CLI #Tools

Overview

LLM-CLI is a TypeScript-based command-line interface that allows you to interact with multiple LLM providers (DeepSeek, Ollama, ChatGPT) through a unified interface. The tool features built-in capabilities for file management, workspace operations, and development tasks through an extensible tool system.

Features

• Multi-LLM Support: Switch between DeepSeek, Ollama, and ChatGPT
• Tool Integration: Built-in tools for file operations, npm package management, weather info, and more
• Streaming Responses: Real-time response streaming with reasoning display
• Conversation Management: Maintain conversation context with temperature control
• File Operations: Read, write, edit, and list files in your workspace
• Development Tools: Install npm packages, manage dependencies, and more

Installation

Prerequisites

• Node.js (v16 or higher)
• npm or yarn
• For Ollama: Ollama installed and running locally

Local Development Setup

1. Clone the repository:

git clone <repository-url>
cd llm-cli

2. Install dependencies:

npm install

3. Configure API keys (optional):

   • For DeepSeek: Set DEEPSEEK_API_KEY environment variable
   • For ChatGPT: Set OPENAI_API_KEY environment variable
   • For Ollama: Ensure Ollama is running (default: http://localhost:11434)

4. Start the CLI:

npm start

Global Installation

To install the CLI globally on your system (so you can run llm-cli from anywhere):

1. Install the package globally from the local directory:

npm install -g .

2. Configure API keys (same as above).

3. Run the CLI:

llm-cli

Usage

Basic Commands

Once started, you'll see a banner with LLM configuration. Type your questions or commands:

LLM> What's the weather in Tokyo tomorrow?

Interactive Commands

Use slash commands for system operations:

• /help or /h - Show help
• /exit, /quit, /q, or /x - Exit the program
• /clear or /c - Clear conversation history
• /temp <value> or /t <value> - Set temperature (0.0 to 1.0)
• /debug or /d - Enable debug mode

Available Tools

The LLM has access to the following tools for assisting with development tasks:

File Operations

• list_workspace - List files on the user workspace
• get_file - Read the contents of a single file
• get_files - Read multiple files using glob patterns
• new_file - Create or overwrite a file
• edit_file - Edit an existing file with new content

Development Tools

• npm_install - Install npm packages (specific packages or all dependencies)
  • Parameters:
    • packages (array): Package names to install (optional)
    • dev (boolean): Install as dev dependency (default: false)
    • global (boolean): Install globally (default: false)

Utility Tools

• get_date - Get current date (YYYY-MM-DD)
• get_weather - Get weather information for a location and date

Examples

File Operations

LLM> Show me the contents of src/tools.ts
[Tool Called: get_file({"file_name":"src/tools.ts"})]
... file contents displayed ...

LLM> Create a new file called test.js with console.log("Hello World")
[Tool Called: new_file({"file_name":"test.js","content":"console.log(\"Hello World\")"})]
FILE WRITTEN

LLM> Edit test.js to add a function
[Tool Called: edit_file({"file_name":"test.js","content":"function greet() {\\n  console.log(\"Hello World\");\\n}\\n\\ngreet();"})]
FILE_EDITED_SUCCESSFULLY

Package Management

LLM> Install express and cors as dependencies
[Tool Called: npm_install({"packages":["express","cors"]})]
Successfully installed packages: express, cors

LLM> Install jest as a dev dependency
[Tool Called: npm_install({"packages":["jest"],"dev":true})]
Successfully installed packages: jest

LLM> Install all dependencies from package.json
[Tool Called: npm_install({})]
Successfully installed all dependencies from package.json

Project Analysis

LLM> List all TypeScript files in the project
[Tool Called: get_files({"glob_pattern":"**/*.ts"})]
... files displayed ...

LLM> What's the current date?
[Tool Called: get_date({})]
2024-01-15

Configuration

LLM Selection

The CLI automatically selects the LLM based on available API keys:

1. DeepSeek (if DEEPSEEK_API_KEY is set)
2. ChatGPT (if OPENAI_API_KEY is set)
3. Ollama (default, requires Ollama running locally)

Environment Variables

• DEEPSEEK_API_KEY: Your DeepSeek API key
• OPENAI_API_KEY: Your OpenAI API key
• OLLAMA_HOST: Custom Ollama host URL (default: http://localhost:11434)

Default Settings

• Model: DeepSeek uses deepseek-reasoner, Ollama uses llama3.2
• Temperature: 0.1 (adjustable with /temp command)
• Max Tokens: 65536

Project Structure

llm-cli/
├── src/
│   ├── api.ts              # LLM API communication
│   ├── cli.ts              # Command-line interface
│   ├── index.ts            # Main entry point
│   ├── tools.ts            # Tool registry and interface
│   └── tools/              # Individual tool implementations
│       ├── workspace.ts    # List workspace files
│       ├── getFile.ts      # Read single file
│       ├── getFiles.ts     # Read multiple files
│       ├── newFile.ts      # Create files
│       ├── editFile.ts     # Edit files
│       ├── npmInstall.ts   # Install npm packages
│       ├── getDate.ts      # Get current date
│       └── getWeather.ts   # Get weather info
├── bin/
│   └── llm-cli.js          # Global CLI entry point
├── package.json
└── README.md

Development

Adding New Tools

To add a new tool:

1. Create a new file in src/tools/ implementing the ToolInterface
2. Import and register the tool in src/tools.ts
3. Update the README with tool documentation

Building

The project uses TypeScript with tsx for execution. No separate build step is required for development.

License

[Add appropriate license information here]

Contributing

[Add contribution guidelines here]

Acknowledgments

• Built with TypeScript and Node.js
• Uses Axios for HTTP requests
• Supports DeepSeek, Ollama, and ChatGPT APIs
• Inspired by various CLI tools and AI assistants

This file was generated by Deepseek! Using this tool :)