Skip to content

clouds.md

Latest commit

 

History

History
272 lines (198 loc) · 7.04 KB

clouds.md

File metadata and controls

272 lines (198 loc) · 7.04 KB

Cloud Deployment Guide

This guide covers deploying agent-yes to various cloud platforms.

Google Cloud Run

Google Cloud Run allows you to deploy containerized applications serverlessly. You can deploy the agent-yes Docker image directly from GitHub Container Registry.

Prerequisites

  1. Install the Google Cloud SDK
  2. Authenticate with Google Cloud: 
    gcloud auth login
gcloud config set project YOUR_PROJECT_ID

Method 1: Direct Deployment (Simple)

Deploy directly from GitHub Container Registry:

# Deploy the latest version
gcloud run deploy agent-yes \
  --image ghcr.io/snomiao/agent-yes:latest \
  --region us-central1 \
  --platform managed \
  --allow-unauthenticated

# Deploy with environment variables for API keys
gcloud run deploy agent-yes \
  --image ghcr.io/snomiao/agent-yes:latest \
  --region us-central1 \
  --platform managed \
  --set-env-vars ANTHROPIC_API_KEY=your-api-key-here \
  --allow-unauthenticated

Method 2: Using Artifact Registry Remote Repository (Recommended)

For production deployments, use Artifact Registry as a remote repository proxy. This provides better reliability and avoids the 9.9 GB layer size limitation.

Step 1: Create an Artifact Registry remote repository

# Create a remote repository for GitHub Container Registry
gcloud artifacts repositories create ghcr-remote \
  --repository-format=docker \
  --location=us-central1 \
  --mode=remote-repository \
  --remote-repo-config-desc="GitHub Container Registry" \
  --remote-docker-repo=DOCKER-HUB \
  --remote-username=_json_key \
  --remote-password-secret-version=projects/YOUR_PROJECT_ID/secrets/github-pat/versions/latest

# For public repositories (no authentication needed)
gcloud artifacts repositories create ghcr-remote \
  --repository-format=docker \
  --location=us-central1 \
  --mode=remote-repository \
  --remote-repo-config-desc="GitHub Container Registry Public"

Step 2: Deploy using the remote repository

gcloud run deploy agent-yes \
  --image us-central1-docker.pkg.dev/YOUR_PROJECT_ID/ghcr-remote/snomiao/agent-yes:latest \
  --region us-central1 \
  --platform managed

Using Secrets for API Keys

For better security, store API keys in Google Secret Manager:

Step 1: Create secrets

# Create secret for Anthropic API key
echo -n "your-anthropic-api-key" | gcloud secrets create anthropic-api-key \
  --data-file=-

# Create secret for OpenAI API key
echo -n "your-openai-api-key" | gcloud secrets create openai-api-key \
  --data-file=-

Step 2: Deploy with secrets

gcloud run deploy agent-yes \
  --image ghcr.io/snomiao/agent-yes:latest \
  --region us-central1 \
  --set-secrets ANTHROPIC_API_KEY=anthropic-api-key:latest,OPENAI_API_KEY=openai-api-key:latest \
  --platform managed

Configuring Resources

Adjust memory and CPU based on your workload:

gcloud run deploy agent-yes \
  --image ghcr.io/snomiao/agent-yes:latest \
  --region us-central1 \
  --memory 2Gi \
  --cpu 2 \
  --timeout 3600 \
  --max-instances 10 \
  --platform managed

Running Jobs (Non-HTTP)

For batch processing or scheduled tasks, use Cloud Run Jobs:

# Create a job
gcloud run jobs create agent-yes-job \
  --image ghcr.io/snomiao/agent-yes:latest \
  --region us-central1 \
  --set-env-vars ANTHROPIC_API_KEY=your-api-key \
  --task-timeout 3600 \
  --max-retries 3

# Execute the job
gcloud run jobs execute agent-yes-job --region us-central1

# Execute with arguments
gcloud run jobs update agent-yes-job \
  --region us-central1 \
  --args="--","run all tests and commit"

gcloud run jobs execute agent-yes-job --region us-central1

Mounting Configuration Files

To persist credentials across runs, use Cloud Storage FUSE:

# Deploy with volume mount (requires Cloud Storage bucket)
gcloud run deploy agent-yes \
  --image ghcr.io/snomiao/agent-yes:latest \
  --region us-central1 \
  --execution-environment gen2 \
  --add-volume name=config,type=cloud-storage,bucket=your-config-bucket \
  --add-volume-mount volume=config,mount-path=/root/.config

Example: Scheduled Automation

Deploy a scheduled job using Cloud Scheduler:

# Create a Cloud Run job
gcloud run jobs create nightly-tests \
  --image ghcr.io/snomiao/agent-yes:latest \
  --region us-central1 \
  --set-secrets ANTHROPIC_API_KEY=anthropic-api-key:latest \
  --args="--exit-on-idle=600","--","run all tests and commit if passing"

# Create a Cloud Scheduler job to run it nightly at 2 AM
gcloud scheduler jobs create http nightly-tests-scheduler \
  --location us-central1 \
  --schedule "0 2 * * *" \
  --uri "https://us-central1-run.googleapis.com/apis/run.googleapis.com/v1/namespaces/YOUR_PROJECT_ID/jobs/nightly-tests:run" \
  --http-method POST \
  --oauth-service-account-email YOUR_SERVICE_ACCOUNT@YOUR_PROJECT_ID.iam.gserviceaccount.com

Viewing Logs

# View service logs
gcloud run services logs read agent-yes \
  --region us-central1 \
  --limit 50

# Stream logs in real-time
gcloud run services logs tail agent-yes \
  --region us-central1

# View job execution logs
gcloud run jobs executions logs read EXECUTION_NAME \
  --region us-central1

Pricing Considerations

Cloud Run pricing is based on:

  • CPU and Memory: Charged per 100ms of usage
  • Requests: $0.40 per million requests
  • Networking: Standard egress rates apply

For AI CLI tools, consider:

  • Use --timeout to limit execution time
  • Set --max-instances to control costs
  • Use Jobs for non-HTTP workloads (cheaper than services)
  • Store large datasets in Cloud Storage instead of container

Multi-Platform Support

The agent-yes image supports both architectures available on Cloud Run:

  • linux/amd64 (x86_64) - Default on Cloud Run
  • linux/arm64 (aarch64) - Not yet available on Cloud Run

Troubleshooting

Issue: Container exits immediately

Cloud Run services require listening on PORT environment variable. For CLI tools like agent-yes, use Cloud Run Jobs instead:

gcloud run jobs create agent-yes-job --image ghcr.io/snomiao/agent-yes:latest ...

Issue: Timeout errors

Increase timeout for long-running tasks:

gcloud run deploy agent-yes --timeout 3600  # 1 hour max

Issue: Permission denied errors

Ensure the service account has necessary permissions:

gcloud projects add-iam-policy-binding YOUR_PROJECT_ID \
  --member serviceAccount:YOUR_SERVICE_ACCOUNT \
  --role roles/secretmanager.secretAccessor

Other Cloud Platforms

AWS (Coming Soon)

  • AWS Fargate
  • Amazon ECS
  • AWS Lambda (with container support)

Azure (Coming Soon)

  • Azure Container Instances
  • Azure Container Apps
  • Azure Kubernetes Service

DigitalOcean (Coming Soon)

  • App Platform
  • Kubernetes

Fly.io (Coming Soon)

  • Fly Machines
  • Fly Apps

Contributing

Have experience deploying to other cloud platforms? Please contribute by adding documentation for:

  • AWS deployment guides
  • Azure deployment guides
  • Other cloud providers

Open a pull request at: https://github.com/snomiao/agent-yes