Skip to main content

​
Overview

Fly.io runs Docker containers on edge servers worldwide. Polpo runs as a single machine with a persistent volume.

​
Deploy

  1. Install the Fly CLI 
    curl -L https://fly.io/install.sh | sh
fly auth login
  2. Create fly.toml In your project directory (or an empty directory): 
    app = "polpo"
primary_region = "iad"

[build]
  image = "ghcr.io/lumea-labs/polpo:latest"

[env]
  NODE_ENV = "production"

[http_service]
  internal_port = 3000
  force_https = true

[mounts]
  source = "polpo_data"
  destination = "/workspace"
  3. Create the app and volume 
    fly apps create polpo
fly volumes create polpo_data --region iad --size 1
  4. Set secrets 
    fly secrets set ANTHROPIC_API_KEY=sk-ant-...
# Add other provider keys as needed
  5. Deploy 
    fly deploy

​
Initialize

After the first deploy, SSH in and init the workspace: 
fly ssh console
node /app/dist/cli/index.js init --dir /workspace
exit
Or hit the API directly: 
curl https://polpo.fly.dev/api/v1/health

​
Scaling

Polpo is a single-process orchestrator. Run one machine per app. Agent tasks are child processes within the same machine. 
fly scale count 1
fly scale vm shared-cpu-2x  # 2 vCPU, 512 MB — adjust for workload
For CPU-intensive agent workloads (many concurrent agents), scale up the VM size rather than the machine count.

​
Notes

  • The persistent volume is region-specific. Pick a region close to your LLM provider endpoints.
  • Fly.io auto-stops idle machines by default. Disable this for always-on orchestration: 
    [http_service]
  auto_stop_machines = "off"
  auto_start_machines = true
  min_machines_running = 1
  • Fly.io supports custom domains via fly certs add.