Local Distributed Stack

Use this when your local stack mixes APIs, workers, a broker, and data services and you need a single deterministic boot from one config.

FlowLayer treats every process as a peer with explicit dependencies, so Postgres, Redis, and Kafka come up before the API and workers that need them — with one source of truth for state and logs.

Server config for a local distributed stack

{
  "session": {
    "bind": "127.0.0.1:6999",
    "token": "dev-token-local-stack"
  },
  "services": {
    "postgres": {
      "cmd": ["postgres", "-D", "./var/postgres"],
      "port": 5432,
      "ready": {
        "type": "tcp"
      }
    },
    "redis": {
      "cmd": ["redis-server", "./infra/redis.conf"],
      "port": 6379,
      "ready": {
        "type": "tcp"
      }
    },
    "kafka": {
      "cmd": ["kafka-server-start.sh", "./infra/kafka/server.properties"],
      "port": 9092,
      "ready": {
        "type": "tcp"
      }
    },
    "api": {
      "cmd": ["pnpm", "--dir", "services/api", "dev"],
      "dependsOn": ["postgres", "redis", "kafka"],
      "port": 3000,
      "ready": {
        "type": "http",
        "url": "http://127.0.0.1:3000/health"
      }
    },
    "worker": {
      "cmd": ["pnpm", "--dir", "services/worker", "dev"],
      "dependsOn": ["api", "redis", "kafka"]
    },
    "orders-producer": {
      "cmd": ["pnpm", "--dir", "services/orders-producer", "dev"],
      "dependsOn": ["api", "kafka"]
    },
    "orders-consumer": {
      "cmd": ["pnpm", "--dir", "services/orders-consumer", "dev"],
      "dependsOn": ["kafka", "postgres", "redis"]
    }
  }
}

Start the server runtime

flowlayer-server -c ./flowlayer.jsonc

Connect the official TUI

Use explicit session flags:

flowlayer-client-tui -addr 127.0.0.1:6999 -token dev-token-local-stack

Or keep a separate client config file:

{
  "session": {
    "addr": "127.0.0.1:6999",
    "token": "dev-token-local-stack"
  }
}

flowlayer-client-tui -config ./flowlayer.tui.jsonc

Server and client config stay separate:

• server listens with session.bind
• client connects with session.addr
• do not reuse server config as TUI config unless you authored it for that purpose

Practical operations

In one session, you can:

• restart only orders-consumer after schema or topic changes
• stop and start worker without touching api
• inspect Kafka-facing services and API logs side by side from the TUI

Command contracts are described in Actions and Logs.

Boundaries

• FlowLayer is not a production orchestrator.
• FlowLayer does not replace Kubernetes, Docker Compose, PM2, or systemd.
• Use this pattern to simulate distributed development behavior and runtime interactions.