CLI Reference

Complete reference for all Pulse command-line interface commands.

Installation

# Install globally with npm
npm install -g @anthropic/pulse

# Or with pnpm
pnpm add -g @anthropic/pulse

# Verify installation
pulse --version

Global Options

These options are available for all commands:

FlagDescription
--help, -hShow help for command
--version, -vShow CLI version
--config, -cPath to config file (default: pulse.yaml)
--verboseEnable verbose output
--jsonOutput in JSON format
--quiet, -qSuppress non-error output

Commands

pulse init

Initialize a new Pulse project

Usage

pulse init [project-name] [options]

Options

--templateUse a starter templatedefault: default
--runtimeDefault runtime for modelsdefault: python3.11

pulse validate

Validate configuration files

Usage

pulse validate [options]

Options

--configPath to config filedefault: pulse.yaml
--strictFail on warningsdefault: false

pulse snapshot

Create a data snapshot

Usage

pulse snapshot <datasource> [options]

Options

--forceForce snapshot even if recentdefault: false
--dry-runPreview without creatingdefault: false

pulse train

Train a model

Usage

pulse train <model> [options]

Options

--snapshotUse specific snapshotdefault: latest
--watchStream training logsdefault: true
--gpuRequest GPU resourcesdefault: false

pulse deploy

Deploy a model to inference runtime

Usage

pulse deploy <model> [options]

Options

--versionDeploy specific versiondefault: latest
--canaryCanary deployment percentagedefault: 0
--rollbackRollback to previous versiondefault: false

pulse infer

Run inference on a deployed model

Usage

pulse infer <model> [options]

Options

--inputInput JSON or file pathdefault: -
--outputOutput formatdefault: json

pulse lineage

View lineage for a model or inference

Usage

pulse lineage <id> [options]

Options

--depthLineage depth to showdefault: full
--formatOutput formatdefault: tree

pulse drift

Check for model drift

Usage

pulse drift <model> [options]

Options

--baselineBaseline snapshot for comparisondefault: training
--thresholdDrift detection thresholddefault: 0.1

Examples

Full Workflow

# Initialize project
pulse init fraud-detection

# Validate configuration
pulse validate

# Create a snapshot of training data
pulse snapshot transactions-db

# Train the model
pulse train fraud-detector --watch

# Deploy to production
pulse deploy fraud-detector

# Run inference
echo '{"amount": 1500, "merchant": "electronics"}' | pulse infer fraud-detector

# Check lineage
pulse lineage fraud-detector@latest

# Monitor for drift
pulse drift fraud-detector --threshold 0.05

CI/CD Integration

# In your CI pipeline
pulse validate --strict
pulse train fraud-detector --snapshot snap_abc123
pulse deploy fraud-detector --canary 10

# After validation
pulse deploy fraud-detector --canary 100

Environment Variables

Configure Pulse behavior with environment variables:

VariableDescription
PULSE_API_KEYAPI key for Pulse cloud services
PULSE_CONFIGDefault config file path
PULSE_LOG_LEVELLog level (debug, info, warn, error)
PULSE_CACHE_DIRLocal cache directory