Commands

`launchgate run <suite>`

Run an eval suite and display results.

Arguments

Argument	Description
`suite`	The eval suite slug (e.g., `rag-faithfulness`)

Options

Flag	Description
`--output <text>`	The AI-generated output to evaluate
`--output-file <path>`	Read output from a file instead
`--input <json>`	Input payload as inline JSON
`--input-file <path>`	Read input from a JSON file
`--api-key <key>`	LaunchGate API key (overrides env var)
`--api-url <url>`	LaunchGate API URL (overrides env var)
`--async`	Return immediately with run ID (don’t wait for results)

Examples

Inline output and input


launchgate run my-suite \
  --output "Paris is the capital of France." \
  --input '{"query": "What is the capital of France?"}'

From files


launchgate run my-suite \
  --output-file ./generated-output.txt \
  --input-file ./test-input.json

Async mode


launchgate run large-suite \
  --output "..." \
  --async
# Prints run ID immediately, exits 0

With explicit API key


launchgate run my-suite \
  --api-key lg_live_abc123 \
  --output "..."

Output format

On success (cleared):


✓ Cleared for launch
  Pass rate: 100% (5/5)
  Duration:  1234ms
  Run ID:    run_abc123

  ✓ Case name 1    1.0 / 1.0
  ✓ Case name 2    0.9 / 0.7
  ✓ Case name 3    1.0 / 0.5

On failure (aborted):


✗ Launch aborted
  Pass rate: 40% (2/5)
  Duration:  2341ms
  Run ID:    run_def456

  ✓ Case name 1    1.0 / 1.0
  ✗ Case name 2    0.3 / 0.7  — Output does not match expected pattern
  ✓ Case name 3    0.9 / 0.5
  ✗ Case name 4    0.1 / 0.8  — Hallucinated information detected
  ✗ Case name 5    0.0 / 1.0  — Expected value not found in output

On skip (API unreachable):


⚠ Eval skipped (API unreachable)
  Run ID: run_ghi789