Usage
Options
| Flag | Description | Default |
|---|---|---|
-m, --model <model> | (required) Model to test (e.g. gemini-2.5-flash, claude-sonnet-4-20250514) | — |
--api-key <key> | API key for the model provider (overrides env vars) | from env |
--scenario <id> | Skip interactive picker, use this scenario by name/id | interactive |
-n, --runs <count> | Runs per harness | 1 |
-t, --timeout <seconds> | Timeout per run in seconds | 120 |
-q, --quiet | Suppress non-error output | false |
-v, --verbose | Enable debug logging | false |
--json | Output results as JSON | false |
How it works
- Authenticates with Archal (cloud twins require login)
- Detects the provider from your model name (Gemini, Anthropic, OpenAI)
- Resolves the API key from
--api-key,ARCHAL_ENGINE_API_KEY, or provider-specific env vars (GEMINI_API_KEY,OPENAI_API_KEY,ANTHROPIC_API_KEY) - Presents an interactive scenario picker (or uses
--scenarioto skip) - Runs the scenario through each bundled harness sequentially
- Prints a comparison table with satisfaction scores
The interactive scenario picker requires a TTY. In non-interactive environments (CI, piped input), pass
--scenario <name> to select directly.Example
API key resolution
The demo command resolves the engine API key in this order:--api-keyflagARCHAL_ENGINE_API_KEYenvironment variable- Provider-specific env vars:
GEMINI_API_KEY,OPENAI_API_KEY,ANTHROPIC_API_KEY
Related
- archal harness — list available harnesses
- archal run — run a single scenario with a specific harness
- Harness configuration — build your own harness