import * as GCP from "@alchemy.run/gcp";
import { expect } from "@alchemy.run/test";
import * as Test from "@alchemy.run/test/Alchemy";
import * as ml from "@distilled.cloud/gcp/ml_v1";
import * as Effect from "effect/Effect";
import * as Schedule from "effect/Schedule";
import { logLevel, currentProject, region, runLifecycle, runVersionLifecycle } from "./common.ts";
const { test } = Test.make({ providers: GCP.providers() });
const waitUntilGone = (name: string) =>
ml.getProjectsModelsVersions({ name }).pipe(
Effect.as("found" as const),
Effect.catchTag("NotFound", () => Effect.succeed("gone" as const)),
Effect.repeat({
schedule: Schedule.spaced("2 seconds"),
until: (status) => status === "gone",
times: 10,
}),
);
test.provider(
"getProjectsModelsVersions on a missing version fails with a typed tag",
(stack) =>
Effect.gen(function* () {
const project = yield* currentProject;
yield* stack.destroy();
const error = yield* Effect.flip(
ml.getProjectsModelsVersions({
name: `projects/${project}/models/alchemy-missing-model/versions/v1`,
}),
);
expect(error._tag).toEqual("NotFound");
yield* stack.destroy();
}).pipe(logLevel),
{ tags: ["provider:gcp", "provider:gcp:ml", "live"], timeout: 90_000 },
);
test.provider.skipIf(runLifecycle)(
"createProjectsModelsVersions on a missing model fails with NotFound",
(stack) =>
Effect.gen(function* () {
const project = yield* currentProject;
yield* stack.destroy();
const error = yield* Effect.flip(
ml.createProjectsModelsVersions({
parent: `projects/${project}/models/alchemy-missing-model`,
body: {
name: "v1",
deploymentUri: `gs://${project}-ml-missing/model`,
description: "probe",
},
}),
);
expect(error._tag).toEqual("NotFound");
yield* stack.destroy();
}).pipe(logLevel),
{ tags: ["provider:gcp", "provider:gcp:ml", "live"], timeout: 90_000 },
);
test.provider.skipIf(!runVersionLifecycle)(
"create, update, and delete a model version",
(stack) =>
Effect.gen(function* () {
yield* stack.destroy();
const deploymentUri = process.env.GCP_TEST_ML_DEPLOYMENT_URI ?? "";
const created = yield* stack.deploy(
Effect.gen(function* () {
const model = yield* GCP.ML.Model("Classifier", {
description: "image classifier",
regions: [region],
});
const version = yield* GCP.ML.ModelsVersion("V1", {
model: model.name,
deploymentUri,
runtimeVersion: "2.11",
pythonVersion: "3.7",
framework: "TENSORFLOW",
description: "first version",
labels: { env: "test" },
autoScaling: { minNodes: 0 },
});
return { model, version };
}),
);
expect(created.version.name).toContain("/versions/");
expect(created.version.model).toEqual(created.model.name);
expect(created.version.description).toEqual("first version");
expect(created.version.deploymentUri).toEqual(deploymentUri);
expect(created.version.labels).toMatchObject({ env: "test" });
const fetched = yield* ml.getProjectsModelsVersions({
name: created.version.name,
});
expect(fetched.name).toEqual(created.version.name);
expect(fetched.description).toContain("[alchemy ");
expect(fetched.description).toContain("first version");
const updated = yield* stack.deploy(
Effect.gen(function* () {
const model = yield* GCP.ML.Model("Classifier", {
modelId: created.model.modelId,
description: "image classifier",
regions: [region],
});
const version = yield* GCP.ML.ModelsVersion("V1", {
model: model.name,
versionId: created.version.versionId,
deploymentUri,
runtimeVersion: "2.11",
pythonVersion: "3.7",
framework: "TENSORFLOW",
description: "first version v2",
autoScaling: { minNodes: 0 },
});
return { model, version };
}),
);
expect(updated.version.name).toEqual(created.version.name);
expect(updated.version.description).toEqual("first version v2");
yield* stack.destroy();
const gone = yield* waitUntilGone(created.version.name);
expect(gone).toEqual("gone");
}).pipe(logLevel),
{ tags: ["provider:gcp", "provider:gcp:ml", "live"], timeout: 120_000 },
);