One Review Man — AI‑generated programming comedy IP

A programming parody of One‑Punch Man, generated chapter‑by‑chapter with AI. Content is authored in English and translated to other languages (currently Russian).

This repository is a monorepo for the project:

• The Eidos gem: a Ruby engine + CLI for building and managing IP worlds ("storyworlds"). Eidos handles world management, canonical lore (Story Bible), canon versioning, content production, translation, and publishing.
• The One Review Man storyworld: the actual content (chapters, characters, comics) produced by Eidos.
• The generated Jekyll site: the public reading surface, built from the storyworld content.

Eidos is reusable — you can install it as a gem (gem install eidos) and use it for any storyworld, not just this one.

Start here

• Pitch — what Eidos is, who it's for, and what it isn't. Read first.
• Usage Guide — every workflow end-to-end, organized by what you want to do.

Project structure

one-review-man/
├── eidos/                     # 💎 Eidos gem (engine + CLI + SDK)
│   ├── lib/eidos/             # Engine + SDK (Eidos:: namespace)
│   ├── lib/eidos/cli/         # Thor-based CLI classes
│   ├── exe/eidos              # Unified `eidos` binary (installed by the gem)
│   ├── bin/{world,bible,canon,produce,translate,publish}
│   │                          # Domain-specific dev binaries (same commands,
│   │                          # split for convenience during development)
│   ├── templates/jekyll/      # Jekyll site template
│   └── spec/                  # RSpec tests
├── worlds/one-review-man/     # 🌍 The storyworld content
│   ├── content/               # Chapters, characters, comics (+ translations)
│   └── data/                  # world_config.yml, story_bible, settings.yml
└── site/                      # 📖 Generated Jekyll site

Quick start (monorepo dev mode)

# Install Eidos dependencies
cd eidos
bundle install

# Generate the next chapter of One Review Man
MOCK_AI=true bin/produce chapter -w ../worlds/one-review-man --auto

# Or use the unified CLI
exe/eidos chapter list -w ../worlds/one-review-man
exe/eidos character show kenji_yamamoto -w ../worlds/one-review-man

# Run the full test suite (544 examples)
MOCK_AI=true bundle exec rspec

Using Eidos as a gem (for your own storyworld)

gem install eidos
eidos --version    # eidos 0.2.0
eidos world new -w /path/to/my-world

From Ruby:

require 'eidos'

Eidos.configure { |c| c.worlds_path = '/path/to/worlds' }

world = Eidos::World.new('my-world')
puts world.status[:title]
puts world.chapters.count
world.bible.characters.each { |c| puts "#{c.id}: #{c.name}" }

See eidos/README.md for the full SDK and CLI reference.

CLI overview

All commands accept -w / --world-dir to point at a storyworld directory.

eidos world   ...   # Create, inspect, manage worlds
eidos bible   ...   # Manage the Story Bible (characters, locations, facts)
eidos canon   ...   # Canon versioning (snapshots, branches)
eidos produce ...   # Generate chapters, comics, illustrations
eidos translate ... # Translate content to other languages
eidos publish ...   # Build the Jekyll site
eidos chapter ...   # SDK-based chapter browsing (list, show)
eidos character ... # SDK-based character browsing (list, show, update)
eidos probe MODEL   # Cheap smoke-test for a provider/model (auth, reachability, latency)
eidos version       # Show the installed Eidos version

The bin/<subcommand> scripts inside eidos/bin/ are equivalent to eidos <subcommand> and exist for monorepo dev convenience.

Content workflow

1. Generate — eidos produce chapter writes worlds/one-review-man/content/chapters/NNN-chapter.md and updates the Story Bible with any new characters/locations.
2. Translate — eidos translate all ru writes .ru.md siblings, using a glossary built from existing character translations.
3. Publish — eidos publish jekyll --dest site assembles content into the Jekyll template. cd site && bundle exec jekyll serve runs it at http://localhost:4000.

Development modes

• MOCK_AI=true — deterministic offline responses from eidos/spec/support/mock_responses.yml. Use for tests and local iteration without API calls.
• DEBUG_AI=1 or --debug — write request/response artifacts to tmp/ai_debug/.

LLM configuration

Per‑world config lives at worlds/<name>/data/settings.yml:

llm:
  provider: openai
  model: gpt-4o-mini
  temperature: 0.7
  timeout: 240
  default_options:
    max_tokens: 12000
  task_options:
    generation:
      max_tokens: 8000
    translation:
      max_tokens: 12000

Provide OPENAI_API_KEY via the environment. gpt-5* / o3* model quirks (max_completion_tokens, temperature) are handled automatically.

Docker

docker compose build
docker compose up -d

License

MIT.

— One Review Man: where AI meets programming parody, one perfect pull request at a time. 🤖📚💻⚡