Universal LLM CLI

A powerful, universal Command Line Interface for interacting with local LLMs (via Ollama) and OpenAI-compatible APIs directly from your terminal.

It features a "Clean Architecture" design, filesystem access tools (Agentic capabilities), and persistent configuration.

Features

• Universal Compatibility: Works out-of-the-box with Ollama, but can be configured for OpenAI, Claude, Mistral, etc.
• Agentic Capabilities: The model can list, read, and write files in your project (safe mode with scoped permissions).
• Persistent Config: Remembers your settings (URL, model, system prompt) across sessions.
• Slash Commands: Built-in commands like /help, /config, /set, /lang.
• Themes & i18n: Supports English/Polish and multiple color themes.
• Project Context: Open the CLI in a specific project folder using -p <path>.

Installation

npm install -g universal-llm-cli

Or run directly without installing:

npx universal-llm-cli

Quick Start

1. Start the CLI:

   unicli

2. Configure (if not using default Ollama):

   /set url https://api.openai.com/v1
   /set key sk-proj-...
   /set model gpt-4o

3. Chat: Just type your message.
4. Use Tools: Ask the model to do something with files:

   "List files in src/ and create a new README.md with a summary."

Configuration

Configuration is stored in ~/.universal-cli-config.json. You can view it using /config.

Commands

• /help - Show available commands
• /config - Show current configuration
• /set <key> <value> - Update config (url, model, key, system)
• /lang <pl|en> - Change interface language
• /bye - Exit

License

MIT