deploy-ai-agent

Deploy AI Agent

Deploy AI agent workloads (LangChain, CrewAI, AutoGen, custom frameworks) as managed systemd services with resource monitoring, API key management, and health checks.

Deploy Workflow

1. Understand — Ask what agent framework they're using and what it needs (model provider, API keys, GPU, dependencies)
2. Check resources — Use system_health to verify RAM, disk, and CPU are sufficient. Check for GPU with shell_exec running nvidia-smi or ls /dev/dri
3. Set up environment — Create a Python venv, install Node.js deps, or verify Go binary. Write API keys to the secrets directory.
4. Deploy — Use app_deploy with appropriate resource limits, environment variables pointing to secrets, and a health-check-friendly port
5. Verify — Check app_logs for successful startup. Use system_discover to confirm the agent's port is listening.
6. Monitor — Set up a watcher via watcher_add to auto-restart on failure
7. Remember — Use memory_store to save deployment details for future reference

Common Patterns

FastAPI Agent Server (LangChain / LangServe)

app_deploy({
  name: "my-agent",
  command: "/var/lib/osmoda/apps/my-agent/venv/bin/uvicorn",
  args: ["app:app", "--host", "0.0.0.0", "--port", "8000"],
  working_dir: "/var/lib/osmoda/apps/my-agent",
  env: {
    ANTHROPIC_API_KEY_FILE: "/var/lib/osmoda/secrets/anthropic-key",
    OPENAI_API_KEY_FILE: "/var/lib/osmoda/secrets/openai-key"
  },
  port: 8000,
  memory_max: "1G",
  cpu_quota: "200%"
})

CrewAI Kickoff

app_deploy({
  name: "crew-agent",
  command: "/var/lib/osmoda/apps/crew-agent/venv/bin/python",
  args: ["-m", "crew_agent.main"],
  working_dir: "/var/lib/osmoda/apps/crew-agent",
  env: {
    ANTHROPIC_API_KEY_FILE: "/var/lib/osmoda/secrets/anthropic-key"
  },
  port: 8001,
  memory_max: "2G"
})

Custom Node.js Agent

app_deploy({
  name: "node-agent",
  command: "/usr/bin/node",
  args: ["index.js"],
  working_dir: "/home/user/agent",
  env: {
    NODE_ENV: "production",
    PORT: "3000",
    API_KEY_FILE: "/var/lib/osmoda/secrets/agent-api-key"
  },
  port: 3000,
  memory_max: "512M"
})