feat(REPL): poc for heroku REPL and prompt mode

[justinwilaby]

Heroku CLI REPL and Prompt Mode

This PR introduces two new interaction modes for the Heroku CLI:

1. REPL (Read-Eval-Print Loop) Mode
2. Prompt Mode

Overview

REPL Mode

A new interactive shell experience that allows users to execute Heroku CLI commands continuously without needing to prefix each command with heroku. This mode maintains command history and supports tab completion for commands, flags, and common Heroku resources (apps, teams, spaces, etc.). Execution time for each command improves by up to 500% when using REPL.

Prompt Mode

An interactive prompt mode that guides users through command execution by asking for required and optional inputs in a structured manner. This is particularly helpful for complex commands or when users are unsure about available options.

Testing Instructions

Testing REPL Mode

1. Start the REPL:

   heroku
   # OR
   ./bin/run

2. Try basic commands:

   heroku > apps
   heroku > apps:info

3. Test tab completion:

   • Type app and press TAB to see available app-related commands
   • Use --app followed by TAB to see your available apps
   • Try teams: followed by TAB to see team-related commands

4. Test command history:

   • Use UP/DOWN arrows to navigate through previous commands
   • Type history to see command history

5. Test persistent settings:

   heroku > set app my-app-name
   # Notice prompt changes to: my-app-name >
   heroku > unset app
   # Prompt returns to: heroku >

Testing Prompt Mode

1. Test with app creation:

   heroku apps:create --prompt

   • Follow the prompts for required inputs (name, if specified)
   • Respond to optional inputs (region, team)

2. Test with add-ons:

   heroku addons:create --prompt

   • Follow prompts for addon service selection
   • Input addon plan
   • Specify target app

3. Test with pipeline operations:

   heroku pipelines:create --prompt

   • Follow prompts for pipeline name
   • Select staging app
   • Choose production app

MCP Mode

The CLI supports a special Model Context Protocol (MCP) mode that can be enabled by setting the HEROKU_MCP_MODE=true environment variable. This mode modifies the output format to be more LLM-friendly:

Testing MCP Mode

1. Enable MCP mode:

   export HEROKU_MCP_MODE=true
   heroku

2. Execute commands in the REPL:

   heroku > apps:list

   Output will be wrapped with special markers:

   <<<BEGIN RESULTS>>>
   [command output here]
   <<<END RESULTS>>>
   

3. Error handling:
   When errors occur, they are wrapped in error markers:

   <<<ERROR>>>
   [error message]
   <<<END ERROR>>>
   

MCP Mode Features

• Structured output markers for easy parsing
• Disabled tab completion to prevent noise in output
• Consistent error formatting
• Ideal for LLMs or programmatic CLI interaction

Key Features

• Command history persistence between sessions
• Intelligent tab completion for:
  • Commands and subcommands
  • App names
  • Team names
  • Space names
  • Add-on names
  • Common flag values
• Session state management (via set and unset commands)
• Interactive prompt mode for guided command execution

Notes

• REPL history is stored in ~/.heroku_repl_history
• REPL state is stored in ~/.heroku_repl_state
• Tab completion is context-aware and filters based on user access
• Prompt mode prioritizes required inputs before optional ones

Future Considerations

• Add support for command aliases in REPL mode
• Enhance tab completion with more resource types
• Add support for multi-line command input

[ekozilforce]

This looks really 🔥 . I like how this takes advantage of oclif metadata to deliver value without plugin development overhead.

I'm interested in learning how we can iterate on this to allow flexibility for plugin developers to provide their own prompt behavior. If this project goes down that path, I'd love to be an internal tester using our data plugins. 🤙 when the time comes