debatable

AI debate simulator - two LLM instances argue PRO and NEG on a given topic, with web search, streaming TUI, and markdown report generation.

Install

npm install -g debatable

Or run directly:

npx debatable
bunx debatable

Usage

debatable [options] <topic>

Arguments:
  topic                    Debate topic

Options:
  -r, --rounds <n>         Number of debate rounds (default: 3)
  -m, --min-searches <n>   Minimum searches per round (default: 2)
  -k, --api-key <key>      API key (passed directly)
  --api-key-env <var>      Environment variable name containing the API key (default: OPENROUTER_API_KEY)
  --model <model>          LLM model to use (default: openai/gpt-4o)
  --pro-model <model>      Override model for PRO
  --con-model <model>      Override model for NEG
  --judge-model <model>    Override model for judge
  -o, --output <path>      Output path for markdown report
  --headless               Run in headless mode (no TUI)
  --completions <shell>    Generate shell completions (bash, zsh, fish, powershell, elvish, nushell)
  -h, --help               Show help

Shell completions are generated via [usage](https://usage.jdx.dev) (available in the dev shell).
Supported shells: bash, zsh, fish, powershell, nushell.

Examples

debatable "Is AI a threat to humanity?"
debatable -r 5 "Should we colonize Mars?"
debatable --headless -o report.md "Is pineapple on pizza acceptable?"
debatable --api-key-env ANTHROPIC_API_KEY "Will AI replace all jobs?" --headless
debatable --completions bash > /etc/bash_completion.d/debatable

Requirements

• Bun runtime
• usage CLI for shell completion generation (or use devenv shell)
• An API key for a supported AI provider — set via OPENROUTER_API_KEY (default), --api-key-env, or --api-key

Development

devenv.sh
bun install
bun dev

License

The Unlicense