A CLI for CLIs

Use natural language to get help with commands and shell utilities from multiple LLMs at once:

helpme "get out of the git detached head state"

helpme "convert screencap.mov to 5fps gif, one-quarter size"

helpme "prettify xml file sitemap-partners.xml"

helpme "extract geolocation info from IMG_9967.HEIC without exiftool"

# view help
helpme

# basic usage
helpme "some natural language request"

# list current known models
helpme list

# print input/output token costs for a model
helpme cost claude-haiku-4-5-20251001

# use specific model(s)
helpme "some request" --model claude-haiku-4-5-20251001,gpt-5-mini

This project is distributed as a gem which can be installed from source.

# download and build
git clone git@github.com:anandaroop/helpme.git
cd helpme
bundle install

# install locally
bundle exec rake install

You will need API keys in your ENV for all the model providers helpme uses.

See .env.example for the default ENV var names.

(helpme will also fall back to more generically named API keys if they are in your ENV already.)

helpme can be customized as follows:

• Update the system prompt in lib/helpme/response.rb to tailor helpme to your setup.

  • For example I nudge it to use imagemagick for images, ffmpeg for videos; http instead of curl, etc.

• Update the list of models in lib/helpme/models.rb to determine which models will be used by default. (See below for refreshing model list)

• Update the list of providers in lib/helpme/commands/list_command.rb to determine which additional models can be listed or costed. (If you add providers, you'll likely need to account for API key configuration too in lib/helpme/cli.rb)

See: https://rubyllm.com/models/#refreshing-the-registry

> bundle console

# update model list in memory
RubyLLM.models.refresh!

# persist updated list
RubyLLM.models.save_to_json

# view model ids
RubyLLM::models.map(&:id)

# filter ids by family
RubyLLM::models.map(&:id).select{ |id| id =~ /sonnet/i}