Shelly-AI

A command-line interface for interacting with AI models (GPT-4o, Claude, Deepseek, etc) via OpenRouter.

It's available on npmjs.com at https://www.npmjs.com/package/shelly-ai

Features

• Interactive chat with AI models
• One-time query mode
• Automatic command detection and execution
• Configuration wizard
• Model selection from popular models.
• File context support (analyze code, documents, etc.)
• Conversation history management

Easy install

Firstly you will need API keys from Openrouter

npm install -g shelly-ai

Then run

shelly-ai --setup

or

sa --setup

Paste your API key that you obtained from https://openrouter.ai/settings/keys here, and select a model of your choice. You can choose any models (Free or Paid) from https://openrouter.ai/models.

Installation from code

# Clone the repository
git clone https://github.com/yourusername/shelly-ai.git
cd shelly-ai

# Install dependencies
npm install

# Install globally (optional)
npm install -g .

Setup

First, run the setup wizard to configure your API key and preferred model:

shelly-ai --setup

Or use the shorthand:

sa --setup

Usage

Interactive Chat

# Start an interactive chat
sa

One-time Query

# Ask a question without starting interactive mode
sa "How do I list all Docker containers?"

# Ask a question and continue in interactive mode
sa --continue "How do I list all Docker containers?"

File Context

# Analyze a file
sa -f ./myfile.js "Can you explain this code?"

# Analyze multiple files
sa -f ./myfile.js ./otherfile.txt "What do these files do?"

# Just analyze files without a specific question
sa -f ./myfile.js ./otherfile.txt

# Important: Always use quotes around your query when using file context
sa -f ./myfile.js "What does this file do?"    # CORRECT
sa -f ./myfile.js What does this file do?      # INCORRECT - will parse words as filenames

Command Execution

# Auto-execute any commands suggested by the assistant
sa --execute "How do I create a new directory in the current folder?"

Streaming

By default, responses are streamed in real-time for a more interactive experience:

# Streaming is enabled by default
sa "Tell me about Node.js"

# Explicitly enable streaming
sa --stream "Tell me about Node.js"

# Disable streaming
sa --no-stream "Tell me about Node.js"

Streaming shows the model's response word by word as it's being generated, which provides a more interactive experience.

Other Options

# List available models
sa --list-models

# Specify a model to use
sa --model anthropic/claude-3.5-sonnet

# Set response temperature (0-1)
sa --temperature 0.8

# Clear conversation history
sa --clear

Chat Commands

Within interactive chat mode, you can use these commands:

• .exit - Exit the chat
• .clear - Clear conversation history
• .help - Show help message
• .model [modelId] - Change model (leave blank to list available models)
• .temp [0-1] - Change temperature (0 = deterministic, 1 = creative)
• .save [filename] - Save conversation to file
• .continue - Continue generating from last response
• .history - Show conversation history
• .file [path] - Load a file as context

License

MIT