Add e2e test for uploading a file to a workspace

Author: philmcmahon

README.md

@@ -162,8 +162,8 @@ sbt 'int:testOnly controllers.api.WorkspacesITest'
 
 ### End-to-end tests
 
-Run `./scripts/test-e2e.sh` to test genesis account creation against disposable
-databases and a separate Giant instance. See [the E2E setup guide](frontend/e2e/README.md)
+Run `./scripts/test-e2e.sh` to test genesis account creation and workspace PDF
+uploads against disposable databases and a separate Giant instance. See [the E2E setup guide](frontend/e2e/README.md)
 for prerequisites, ports and debugging. The GitHub Actions E2E workflow runs the
 same script.
 

backend/conf/e2e.conf

@@ -14,8 +14,9 @@ s3.endpoint = "http://127.0.0.1:13900"
 pekko.remote.artery.canonical.port = 11234
 pekko.cluster.seed-nodes = ["pekko://pfi@localhost:11234"]
 
-# Genesis needs storage and authentication, but no background extraction services.
-worker.enabled = false
+# Process uploads with the real local extractors.
+worker.enabled = true
+worker.interval = 1 second
 worker.useExternalExtractors = false
 ingestion.scratchPath = ${E2E_RUNTIME_DIR}"/ingestion"
 preview.workspace = ${E2E_RUNTIME_DIR}"/preview"

frontend/e2e/README.md

@@ -2,6 +2,15 @@
 
 From the repository root, with Docker Compose, Java, sbt and Node installed:
 
+Install the PDF extraction tools first. On Ubuntu/Debian:
+
+```sh
+sudo apt-get update
+sudo apt-get install -y ocrmypdf ghostscript qpdf poppler-utils tesseract-ocr-eng
+```
+
+On macOS, the project's `Brewfile` includes the extraction tools.
+
 ```sh
 npm ci --prefix frontend
 cd frontend
@@ -17,8 +26,12 @@ Elasticsearch and Garage containers, creates storage buckets, runs PostgreSQL
 migrations, starts the backend and Vite, then runs Playwright. GitHub Actions runs
 the same script. No AWS credentials or pre-existing Giant account are needed.
 
-The single test creates the genesis user through the UI, skips optional 2FA,
-then logs in from a fresh browser context and checks administrator access.
+The genesis project creates the initial user through the UI, skips optional 2FA,
+then logs in from a fresh browser context and checks administrator access. The
+Chromium project depends on genesis and tests creating a workspace, uploading
+`toast_sandwich_en_wiki.pdf` through its upload dialog, and waiting up to three
+minutes for the file's processing icon to become a document icon. An error icon
+fails the test. The fixture uses the real document text and OCR extractors.
 Retries are disabled because genesis requires an empty user store. Rerun the
 script to get a fresh instance; running Playwright alone does not reset anything.
 
@@ -32,6 +45,12 @@ Logs are saved in `frontend/e2e-artifacts/`, the HTML report in
 `frontend/playwright-report/`, and failure traces/screenshots in
 `frontend/test-results/`. CI uploads these even when a run fails.
 
-Background workers are disabled for this genesis-only suite. Before adding
-upload/extraction coverage, enable the required workers in `backend/conf/e2e.conf`
-and install the extraction tools exercised by the fixtures in CI.
+To run just the upload scenario (including its genesis dependency):
+
+```sh
+./scripts/test-e2e.sh workspace-upload.spec.ts
+```
+
+Local extraction workers are enabled in `backend/conf/e2e.conf`; external
+extractors are disabled. The PDF fixture comes from Wikipedia's English
+"Toast sandwich" article: https://en.wikipedia.org/wiki/Toast_sandwich.

frontend/e2e/toast_sandwich_en_wiki.pdf

@@ -0,0 +1,56 @@
+import { expect, test } from "@playwright/test";
+import path from "node:path";
+
+test("uploads a PDF to a new workspace and completes extraction", async ({
+  page,
+}) => {
+  test.setTimeout(240_000);
+  const workspaceName = `Toast sandwich ${Date.now()}`;
+  const filename = "toast_sandwich_en_wiki.pdf";
+
+  await page.goto("/workspaces");
+  await page.getByPlaceholder("Username", { exact: true }).fill("genesis-e2e");
+  await page
+    .getByPlaceholder("Password", { exact: true })
+    .fill("Genesis-e2e-password");
+  await page.getByRole("button", { name: "Login", exact: true }).click();
+
+  await page
+    .getByRole("button", { name: "New Workspace", exact: true })
+    .click();
+  await page.getByPlaceholder("Name", { exact: true }).fill(workspaceName);
+  await page.getByRole("button", { name: "Create", exact: true }).click();
+  await expect(
+    page.getByRole("heading", { name: workspaceName }),
+  ).toBeVisible();
+
+  await page
+    .getByRole("button", { name: "Upload to workspace", exact: true })
+    .click();
+  const fileChooser = page.waitForEvent("filechooser");
+  await page.getByRole("button", { name: "Add Files", exact: true }).click();
+  await (await fileChooser).setFiles(path.join(__dirname, filename));
+  await page.getByRole("button", { name: "Upload", exact: true }).click();
+  // Use the upload form's completion button, not the modal's corner dismiss.
+  await page
+    .locator("form")
+    .getByRole("button", { name: "Close", exact: true })
+    .click();
+
+  const fileRow = page.getByRole("row").filter({
+    has: page.getByText(filename, { exact: true }),
+  });
+  await expect(fileRow).toBeVisible();
+
+  // A vanished spinner alone is insufficient: wait for either terminal icon,
+  // then require the document icon so an extraction error cannot count as success.
+  // The selectors correspond to Workspaces.renderIcon's processed/failed cases.
+  await expect(
+    fileRow.locator(
+      "svg.file-browser__icon, i.exclamation.triangle.file-browser__icon",
+    ),
+  ).toBeVisible({ timeout: 180_000 });
+  await expect(fileRow.locator("svg.file-browser__icon")).toBeVisible();
+  await expect(fileRow.locator(".loader.file-browser__icon")).toHaveCount(0);
+  await expect(fileRow.getByText("processed", { exact: true })).toBeVisible();
+});