@tracepact/cli
v0.5.0
Published
CLI for TracePact — interactive setup, model catalog, test runner
Downloads
764
Maintainers
Readme
@tracepact/cli
Command-line interface for TracePact — run tests, scaffold projects, audit skills, manage cache, and more.
Installation
npm install -g @tracepact/cli
# Or use via npx
npx tracepactCommands
tracepact / tracepact run
Run TracePact tests via Vitest.
tracepact # Run tests in mock mode
tracepact --live # Run against real LLMs
tracepact --live --budget 50000 # Enforce token budget
tracepact --record # Record cassettes (implies --live)
tracepact --replay ./cassettes # Replay from cassettes
tracepact --json # Output JSON results
tracepact --provider anthropic # Use specific provider
tracepact -- --grep "security" # Pass args through to Vitest| Flag | Description |
|------|-------------|
| --live | Run against real LLM APIs |
| --full | Include expensive tests (Tier 3-4) |
| --record | Record cassettes from live runs |
| --replay <dir> | Replay from recorded cassettes |
| --no-cache | Skip response cache |
| --budget <tokens> | Abort if token usage exceeds threshold |
| --json | Enable JSON reporter |
| --provider <name> | Select provider (openai, anthropic, etc.) |
| --health-check-strict | Exit if provider health check fails |
tracepact init
Interactive setup wizard for new projects.
tracepact init # Interactive wizard
tracepact init --demo # Self-contained demo (no API keys)
tracepact init --system-prompt # Raw system prompt template
tracepact init --skill SKILL.md # Generate from existing skill
tracepact init --pattern api-client # Use a pattern template
tracepact init --force # Overwrite existing filesCreates tracepact.config.ts, tracepact.vitest.ts, and a test file template. The interactive wizard walks you through provider, model, and judge selection.
tracepact audit
Static analysis of a SKILL.md file (no API key needed).
tracepact audit SKILL.md
tracepact audit SKILL.md --format json
tracepact audit SKILL.md --fail-on high # Exit 1 if high+ severity foundChecks for:
- Risky tool combinations (e.g. bash + network)
- Prompt hygiene issues
- Skill completeness
- Opaque tool definitions
tracepact capture
Auto-generate a test file from a live run or recorded cassette.
tracepact capture --skill SKILL.md --prompt "Review this code"
tracepact capture --skill SKILL.md --prompt "test" --cassette ./cassettes/run.json --dry-run
tracepact capture --skill SKILL.md --prompt "test" --out tests/generated.tracepact.ts
tracepact capture --skill SKILL.md --prompt "test" --with-semantictracepact cache
Manage the response cache.
tracepact cache list # Show cached entries
tracepact cache clear # Delete all cache
tracepact cache clear --stale # Delete only expired entries
tracepact cache verify # Check cache integritytracepact models
Browse the model catalog.
tracepact models # List all providers and models
tracepact models openai # Show only OpenAI models
tracepact models --verbose # Include pricing details
tracepact models --refresh # Force refresh from models.devShows API key status, context window, and pricing per model.
tracepact cost-report
Show token usage from the last test run.
tracepact cost-reportDisplays total tokens, API calls vs cache hits, per-provider and per-test breakdowns. Reads from .tracepact/last-run-tokens.json.
tracepact doctor
Environment and configuration health check.
tracepact doctorChecks:
- Node.js version (>=20)
- Vitest installation
- Config file presence (
tracepact.config.ts) - SKILL.md existence
- API keys for all known providers
- Docker/Podman runtime availability
- Cache directory writability
Environment Variables
| Variable | Description |
|----------|-------------|
| TRACEPACT_LIVE | Enable live LLM calls (1 to enable) |
| TRACEPACT_FULL | Enable expensive tests (1 to enable) |
| TRACEPACT_PROVIDER | Default provider name |
| TRACEPACT_BUDGET | Max token budget |
| TRACEPACT_RECORD | Enable cassette recording |
| TRACEPACT_REPLAY | Path to cassette directory for replay |
| TRACEPACT_NO_CACHE | Disable response caching |
License
MIT
