shortcuts

Shortcuts Example

This example demonstrates how to use custom shortcuts in the Inference Gateway CLI.

What are Shortcuts?

Shortcuts are user-defined commands that execute shell commands, scripts, or AI-powered operations within the chat interface. They support:

• Simple commands: Execute any shell command or script
• AI-powered snippets: Execute commands, send output to LLM, use response in templates
• Parameter support: Pass arguments to shortcuts
• JSON output parsing: Structure command output for AI processing

Container Environment: The CLI runs in Alpine Linux with bash and jq pre-installed for scripting and JSON processing.

Quick Start

1. Copy environment file:

   cp .env.example .env

2. Add your API key to .env:

   ANTHROPIC_API_KEY=your-key-here
   # or
   OPENAI_API_KEY=your-key-here

3. Start the services:

   docker compose up -d

4. Run the CLI:

   docker compose run --rm cli

5. Try the shortcuts:

   /hello
   /sysinfo
   /review-comments
   

Custom Shortcuts

This example includes three types of shortcuts in .infer/shortcuts/custom-demo.yaml:

1. Simple Echo Shortcut

- name: hello
  description: "Say hello from the container"
  command: echo
  args:
    - "Hello from the Inference Gateway CLI! 🚀"

Usage: /hello

2. System Information Shortcut

- name: sysinfo
  description: "Display system information"
  command: bash
  args:
    - -c
    - |
      echo "=== System Information ==="
      echo "Hostname: $(hostname)"
      # ... more commands

Usage: /sysinfo

3. AI-Powered Review Comment Reply Generator

- name: review-comments
  description: "Generate suggested replies to code review comments"
  command: bash
  args:
    - -c
    - |
      # Fetch mock GitHub review comments
      jq -n '{repo: "...", prNumber: "123", comments: [...]}'
  snippet:
    prompt: |
      Draft professional replies to review comments...
      Generate a gh CLI command to post all replies.
    template: |
      ! {llm}

Usage: /review-comments

How it works:

1. Fetches mock review comments from a PR (simulates GitHub API)
2. AI analyzes the comments and drafts professional replies
3. AI generates an executable gh CLI command with the replies
4. Command is placed in your input box with ! prefix
5. Review and press Enter to execute and post the replies to GitHub

This demonstrates the snippet + template feature where:

• The command outputs JSON data
• The snippet sends data to AI with instructions
• The template formats the AI response as an executable command
• Perfect for workflows that need human review before execution!

Shortcut Structure

Basic Shortcut

shortcuts:
  - name: shortcut-name
    description: "What this shortcut does"
    command: bash
    args:
      - -c
      - "echo 'Hello World'"

AI-Powered Shortcut (with Snippet)

shortcuts:
  - name: ai-shortcut
    description: "AI-powered operation"
    command: bash
    args:
      - -c
      - |
        # Execute command and output JSON
        jq -n --arg data "value" '{result: $data}'
    snippet:
      prompt: |
        Analyze this data: {result}

        Provide insights based on the data.
      template: |
        ## Analysis Results

        {llm}

How it works:

1. command executes and produces output (preferably JSON)
2. Output fields become variables (e.g., {result})
3. prompt is sent to the LLM with variable substitution
4. LLM response becomes {llm} variable
5. template formats the final output shown to user

Built-in Shortcuts

The CLI includes several built-in shortcuts (available by default):

• /help - Show help information
• /clear - Clear conversation history
• /exit - Exit the CLI
• /switch <conversation-id> - Switch conversations
• /theme <theme-name> - Change UI theme

Adding Your Own Shortcuts

1. Create a new file in .infer/shortcuts/:

   touch .infer/shortcuts/custom-myshortcuts.yaml

2. Define your shortcuts:

   shortcuts:
     - name: my-command
       description: "My custom command"
       command: echo
       args:
         - "My output"

3. Restart the CLI or reload configuration

4. Use your shortcut: /my-command

Tips

• File naming: Use custom-*.yaml for custom shortcuts
• Shell support: Both bash and sh are available in the container
• JSON output: Use jq to format command output for AI processing (pre-installed)
• Error handling: Check command success and provide helpful error messages
• Parameter support: Use $1, $2, etc. in bash scripts to accept arguments
• Testing: Test shortcuts locally before adding to production configs

Volume Mounts

The docker-compose.yml mounts shortcuts as read-only:

volumes:
  - ./.infer/shortcuts:/home/infer/.infer/shortcuts:ro

This ensures:

• Shortcuts are available in the CLI
• Container cannot modify your local shortcuts
• Easy to version control your shortcuts

Next Steps

• Check out other examples in examples/ directory
• Explore built-in shortcuts in the main .infer/shortcuts/ directory
• Create shortcuts for your common workflows

Troubleshooting

Shortcut not found:

• Check file is in .infer/shortcuts/ directory
• Verify YAML syntax is correct
• Restart the CLI container

AI snippet not working:

• Ensure API key is configured
• Check prompt formatting and variable names
• Verify JSON output structure

Cleanup

docker compose down -v

This removes containers and volumes, including persisted conversation data.