LaunchDarkly Go SDK with Fastly Compute

⚠️ Warning: This is an example implementation meant to serve as a starting point for integrating LaunchDarkly with Fastly Compute. The code is incomplete and should be used at your own risk. Make sure to:

• Add proper error handling
• Implement security best practices
• Add appropriate logging
• Test thoroughly in your environment
• Follow LaunchDarkly and Fastly's latest best practices

This project demonstrates how to integrate the LaunchDarkly Go SDK with Fastly's Compute Runtime. It showcases a practical implementation of feature flagging in a serverless environment, specifically using Fastly's KV Store for persistent data storage.

Overview

This demo application shows how to:

• Initialize the LaunchDarkly Go SDK in Fastly's Compute Runtime
• Use Fastly's KV Store as a persistent data store for LaunchDarkly
• Handle feature flag evaluations in a serverless context
• Maintain feature flag state across requests

Key Components

KV Store Implementation

The kvdatastore package provides a custom implementation of LaunchDarkly's data store interface using Fastly's KV Store. This allows feature flag data to persist between requests, improving performance and reducing API calls to LaunchDarkly's servers.

Main Application

The main application (main.go) demonstrates:

• Initialization of the LaunchDarkly client with custom configuration
• Context creation with Fastly-specific attributes
• Feature flag evaluation
• JSON response formatting

Prerequisites

• Go 1.21 or later
• Fastly CLI
• LaunchDarkly account and SDK key
• Fastly account with Compute@Edge enabled

Configuration

1. Set up your LaunchDarkly SDK key in the environment or update the constant in main.go:

const LD_SDK_KEY = "your-sdk-key"

2. Configure your Fastly service using the fastly.toml file.

Local Development

1. Install dependencies:

go mod download

2. Run the application locally:

fastly compute serve

Deployment

Deploy to Fastly using:

fastly compute publish

How It Works

1. When a request comes in, the application initializes the LaunchDarkly client with custom configuration using Fastly's KV Store.
2. The client creates a context with Fastly-specific attributes (POP, region, service version, etc.).
3. Feature flags are evaluated using this context.
4. Results are returned as JSON responses.

Response Format

The application returns JSON responses in the following format:

{
  "animal": "feature-flag-value",
  "context": {
    "kind": "multi",
    "contexts": [
      {
        "kind": "user",
        "key": "user-123"
      },
      {
        "kind": "fastly-request",
        "key": "request-id",
        "fastly_service_version": "version",
        "fastly_pop": "pop",
        "fastly_region": "region",
        "fastly_service_id": "service-id"
      }
    ]
  },
  "reason": {
    "kind": "RULE_MATCH",
    "ruleIndex": 0,
    "ruleId": "rule-id"
  },
  "serviceVersion": "version"
}

License

This project is licensed under the MIT License - see the LICENSE file for details.