Skip to main content

Google Cloud Functions

This page shows how to deploy a Google Cloud Function (gen2) from an OCI artefact using the Fractal SDK.

For the artefact contract and prerequisites, see the main guide.

How the agent deploys​

Cloud Functions gen2 build from a source zip (GCS or Cloud Source Repos), so the artefact must be a zip — a container-image artefact is rejected. The agent:

  1. Pulls the zip artefact from your Artifact Registry (as the fractal-cloud-agent service account).
  2. Creates an agent-owned GCS source bucket if needed and uploads the zip.
  3. Deploys the function with that bucket/object as its StorageSource.

The agent owns the source-bucket lifecycle; you only publish the artefact. entryPoint is required.

TypeScript SDK​

A function is split across the two levels like every other component. The blueprint declares an abstract Function and governs the vendor-neutral parameters; the Live System selects the GcpFunction offer and supplies the GCP-specific ones.

1. Declare the Function in the blueprint​

import {createFractal, Function} from '@fractal_cloud/sdk/model';

export const fractal = createFractal({
  id: 'orders',
  version: {major: 1, minor: 0, patch: 0},
  description: 'Orders processing function.',
  boundedContextId: {
    ownerType: 'Personal',
    ownerId: process.env['OWNER_ID']!,
    name: 'reusable-templates',
  },
  blueprint: bp => {
    const orders = bp.add(
      Function({id: 'orders-fn', displayName: 'Orders Function'})
        .withRuntime('java21')          // guardrail
        .withMemory(512)                // guardrail: MB
        .withTimeout(30)                // guardrail: seconds
        .withEnvironment({LOG_LEVEL: 'INFO'}),
    );
    return {orders};
  },
  operations: s => ({
    /** The artefact the application ships. */
    withArtifact: (uri: string) => s.orders.set('sourceArtifact', uri),
  }),
});

withRuntime, withMemory, withTimeout, withConcurrency and withEnvironment are vendor-neutral and identical on every cloud — the same blueprint satisfies an AWS Lambda or an Azure Function without editing.

2. Select the Cloud Functions offer in the Live System​

import {GcpFunction} from '@fractal_cloud/sdk/model';

const liveSystem = fractal
  .specialize()
  .withArtifact('europe-west1-docker.pkg.dev/my-project/functions/orders-fn:1.4.0')
  .toLiveSystem({
    name: 'orders-dev',
    environment,
    select: {
      'orders-fn': GcpFunction({
        region: 'europe-west1',
        entryPoint: 'com.example.Orders',
      }),
    },
  });

entryPoint is a GCP concept and is required by the offer, so it lives there rather than on the component — a blueprint carrying it could never be satisfied by another vendor.

See the Component Reference for every parameter the offer accepts, and Core API for the authoring surface.