@mastra/deployer-cloud

A cloud-optimized deployer for Mastra applications with built-in telemetry, logging, and storage integration.

Features

• Cloud-Native Integration: Automatic setup for LibSQL storage and vector databases
• Advanced Logging: Built-in PinoLogger with HTTP transport for cloud logging endpoints
• Telemetry & Monitoring: OpenTelemetry instrumentation with readiness logging
• Evaluation Hooks: Automatic agent evaluation tracking and storage
• Multi-Logger Support: Combines cloud logging with existing application loggers
• Environment-Based Configuration: Smart configuration based on deployment environment

Installation

pnpm add @mastra/deployer-cloud

Usage

The cloud deployer is used as part of the Mastra build process:

import { CloudDeployer } from '@mastra/deployer-cloud';

const deployer = new CloudDeployer();

// Bundle your Mastra application
await deployer.bundle(mastraDir, outputDirectory);

// The deployer automatically:
// - Adds cloud dependencies
// - Sets up instrumentation
// - Configures logging and storage

What It Does

1. Dependency Management

Automatically adds cloud-specific dependencies to your package.json:

• @mastra/loggers - Cloud-optimized logging
• @mastra/libsql - Serverless SQL storage
• @mastra/cloud - Cloud platform utilities

2. Server Entry Generation

Creates a production-ready server entry point with:

• Cloud storage initialization (LibSQL)
• Vector database setup
• Multi-transport logging
• Telemetry and monitoring
• Evaluation hooks for agent metrics

3. Instrumentation

Provides OpenTelemetry instrumentation for:

• Distributed tracing
• Performance monitoring
• Custom telemetry configuration

Environment Variables

The deployer configures your application to use these environment variables:

# Storage Configuration
MASTRA_STORAGE_URL=your-libsql-url
MASTRA_STORAGE_AUTH_TOKEN=your-auth-token

# Logging Configuration
BUSINESS_API_RUNNER_LOGS_ENDPOINT=your-logs-endpoint
BUSINESS_JWT_TOKEN=your-jwt-token

# Studio Configuration
PLAYGROUND_JWT_TOKEN=your-playground-jwt-token

# Runtime Configuration
RUNNER_START_TIME=deployment-start-time
CI=true|false

# Deployment Metadata
TEAM_ID=your-team-id
PROJECT_ID=your-project-id
BUILD_ID=your-build-id

Generated Server Code

The deployer generates a server entry that:

1. Initializes Logging:

   • Sets up PinoLogger with cloud transports
   • Combines with existing application loggers
   • Provides structured JSON logging

2. Configures Storage:

   • Initializes LibSQL store when credentials are provided
   • Sets up vector database for semantic search
   • Integrates with Mastra's memory system

3. Registers Hooks:

   • ON_GENERATION - Tracks agent generation metrics
   • ON_EVALUATION - Stores evaluation results

4. Starts Server:

   • Creates Node.js server with Mastra configuration
   • Disables Studio and Swagger UI for production
   • Exports tools for API access

Project Structure

After deployment, your project will have:

output/
├── package.json          # With cloud dependencies
├── index.mjs            # Main server entry
├── mastra.mjs          # Your Mastra configuration
└── tools/              # Exported tools

Readiness Logging

The deployer includes structured readiness logs for monitoring:

{
  "message": "Server starting|Server started|Runner Initialized",
  "type": "READINESS",
  "startTime": 1234567890,
  "durationMs": 123,
  "metadata": {
    "teamId": "your-team-id",
    "projectId": "your-project-id",
    "buildId": "your-build-id"
  }
}

Testing

The cloud deployer includes comprehensive tests covering:

• Build pipeline functionality
• Server runtime generation
• Dependency management
• Error handling
• Integration scenarios

Run tests with:

pnpm test

Development

For local development:

# Build the deployer
pnpm build

# Run tests
pnpm test

# Lint code
pnpm lint