TestGenAI

✨ A CLI gem that adds tests for untested methods in your code ✨

TestGenAI is a Ruby CLI gem for development environments that finds untested methods in your codebase and generates tests for them using an AI service. Point it at your project and it scans your code, finds methods with no test coverage, and asks an LLM to write tests for them, matching your existing style along the way.

See this article for the motivation and design behind TestGenAI.

TestGenAI can scan your code two ways. If SimpleCov is available in your project, it runs your test suite with coverage tracking and finds methods where every line has zero hits, which correctly handles files that are only partially tested. Without SimpleCov, it falls back to checking whether a matching spec or test file exists for each source file. That's simpler, but it can't tell a partially tested file from a fully tested one, so a base class that's only exercised through subclass specs will show up as fully untested.

• Ruby 3.4.5
• Thor for the CLI
• ruby_llm to talk to LLM providers like Anthropic and OpenAI
• parser and prism for parsing and scanning Ruby source
• RSpec, StandardRB, Minitest, and SimpleCov for testing and linting the gem itself

Add TestGenAI to your project's Gemfile in the development group, or install the gem directly:

gem install testgenai

TestGenAI needs an LLM provider to generate tests. Configure it via environment variables or CLI flags:

By default TestGenAI uses the file-existence scanner described above. If you want the more accurate SimpleCov-based scanner, add SimpleCov to your project and configure it to start when COVERAGE=true is set:

# Gemfile
gem "simplecov", require: false, group: [:development, :test]

# spec/spec_helper.rb or test/test_helper.rb
if ENV["COVERAGE"]
  require "simplecov"
  SimpleCov.start
end

With this in place, TestGenAI runs your test suite with COVERAGE=true to generate coverage/.resultset.json, then uses that data to find methods where every executable line has zero hits, including methods in files that are only partially tested. If coverage/.resultset.json already exists from a previous run, TestGenAI uses it directly without re-running the test suite.

If SimpleCov isn't set up correctly, TestGenAI falls back to the file-existence scanner and prints one of these messages:

• SimpleCov not found in Gemfile or gemspec. Using file-existence scanner. means SimpleCov isn't declared in your Gemfile or gemspec.
• Coverage generation failed. Using file-existence scanner. means SimpleCov was found, but running the suite with COVERAGE=true didn't produce coverage/.resultset.json. Check that SimpleCov is configured to start when COVERAGE=true is set, and that your suite runs cleanly with COVERAGE=true bundle exec rspec (or the minitest equivalent).

Run everything through the bin/testgenai executable once it's installed:

testgenai scan      Scan for untested methods and report them
testgenai context   Scan and show the LLM context that would be sent (diagnostic)
testgenai generate  Full pipeline: scan → context → generate → validate
testgenai version   Show version
testgenai help      Show help

Finds untested methods and prints them. No API calls are made.

bin/testgenai scan
bin/testgenai scan --test-framework minitest

Like scan, but also prints the context snippet that would be sent to the LLM for each method. Useful for inspecting what the generator will see before running a full generate.

Runs the full pipeline: scan, build context, call the LLM, validate the generated tests, and report results.

bin/testgenai generate --provider anthropic --model claude-opus-4-7
bin/testgenai generate --provider anthropic --model claude-opus-4-7 --conventions

Generated tests are injected into existing spec/test files when they exist, or written to new files mirroring the source path under spec/ (rspec) or test/ (minitest). If validation fails after injection, the original file is restored and the generated tests are saved to a per-method fallback file (e.g. widget_render_spec.rb) for manual review.

The --conventions flag analyzes your existing test suite to detect patterns (naming conventions, shared contexts, matchers, etc.) and includes them in the LLM prompt so generated tests match your project's style. The analysis result is cached in spec/conventions.md (or test/conventions.md) and refreshed automatically when your test files change.

This repo also uses mise to run the gem's own test suite and linter:

• mise run test runs the RSpec suite (bundle exec rspec)
• mise run lint runs the StandardRB style check (bundle exec standardrb)

Design docs and implementation plans for each major feature live under docs/:

• docs/specs/ has the design docs for the initial build, the SimpleCov integration, and the conventions extractor
• docs/plans/ has the implementation plan that goes with each spec

TestGenAI is released under the MIT License. See LICENSE for the full text.