Fix frontend E2E coverage and gate seeded tests

.github/workflows/frontend_tests.yml

@@ -101,10 +101,12 @@ jobs:
       - name: Install Playwright browsers
         run: npx playwright install --with-deps chromium
 
-      - name: Run E2E tests (mock mode — no backend/secrets needed)
-        run: npx playwright test --project=mock
+      - name: Run E2E tests (mock and seeded — no credentials needed)
+        run: npx playwright test --fail-on-flaky-tests --project=mock --project=seeded
         env:
           CI: true
+          E2E_SEEDED_MODE: true
+          FAIL_ON_SKIPPED_TESTS: true
 
       - name: Upload Playwright report
         uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
@@ -150,3 +152,6 @@ jobs:
 
       - name: Run TypeScript type check
         run: npx tsc --noEmit
+
+      - name: Build production frontend
+        run: npm run build

frontend/README.md

@@ -91,20 +91,20 @@ E2E flow tests run in two modes controlled by Playwright projects and an environ
 
 - **Seeded** (`--project seeded`, default for CI): Messages are stored directly in the database with `send: false` using dummy credentials. No real API keys needed. Tests cover the full UI flow (display, branching, conversation switching, promoting) without calling any external service.
 
-- **Live** (`--project live`, requires `E2E_LIVE_MODE=true`): Messages are sent to real OpenAI endpoints with `send: true`. Each target variant requires its own set of environment variables (e.g., `OPENAI_CHAT_ENDPOINT`, `OPENAI_CHAT_KEY`, `OPENAI_CHAT_MODEL`). Variants whose env vars are missing are automatically skipped. Tests verify that real target responses render correctly.
+- **Live** (`--project live`, requires `E2E_LIVE_MODE=true`): Messages are sent to real OpenAI endpoints with `send: true`. Each target variant requires endpoint and model environment variables plus either an API key or an Azure endpoint accessible through the current Entra identity. Variants without a usable configuration are automatically skipped. Tests verify that real target responses render correctly.
 
 ```bash
-# CI (seeded only — no credentials needed)
+# Seeded integration (no credentials needed)
 npx playwright test --project seeded
 
-# Live integration (requires real API keys)
+# Live integration (uses API keys when present, otherwise Entra authentication)
 E2E_LIVE_MODE=true npx playwright test --project live
 
 # Run both
 E2E_LIVE_MODE=true npx playwright test
 ```
 
-The seeded project runs in the **GitHub Actions** workflow. The live project is intended for an **Azure DevOps pipeline** that has the required secret API keys.
+The mock and seeded projects run in the **GitHub Actions** pull-request workflow. The live project is intended for a protected pipeline with an Entra identity or API keys.
 
 E2E tests use `dev.py` to automatically start both frontend and backend servers. If servers are already running, they will be reused.
 

frontend/e2e/api.spec.ts

@@ -1,6 +1,6 @@
 import { test, expect } from "@playwright/test";
 
-// API tests go through the Vite dev server proxy (/api -> backend:8000)
+// API tests go through the Vite dev server proxy (/api -> configured backend)
 // rather than hitting the backend directly, so they work as soon as
 // Playwright's webServer (port 3000) is ready.
 
@@ -68,26 +68,26 @@ test.describe("Targets API", () => {
   });
 
   test("should create and retrieve a target @seeded", async ({ request }) => {
+    test.setTimeout(90_000);
     const createPayload = {
       type: "OpenAIChatTarget",
+      auth_mode: "api_key",
       params: {
         endpoint: "https://e2e-test.openai.azure.com",
         model_name: "gpt-4o-e2e-test",
         api_key: "e2e-test-key",
       },
     };
 
-    const createResp = await request.post("/api/targets", { data: createPayload });
-    // The endpoint may require credentials or env setup that isn't available
-    // in CI.  Skip gracefully rather than masking real regressions.
-    if (!createResp.ok()) {
-      test.skip(true, `POST /api/targets returned ${createResp.status()} — skipping`);
-      return;
-    }
+    const createResp = await request.post("/api/targets", {
+      data: createPayload,
+      timeout: 60_000,
+    });
+    expect(createResp.ok()).toBe(true);
 
     const created = await createResp.json();
     expect(created).toHaveProperty("target_registry_name");
-    expect(created.target_type).toBe("OpenAIChatTarget");
+    expect(created.identifier.class_name).toBe("OpenAIChatTarget");
 
     // Retrieve via list and check it's there
     const listResp = await request.get("/api/targets?limit=200");
@@ -120,12 +120,6 @@ test.describe("Attacks API", () => {
 
   test("should list attacks @seeded", async ({ request }) => {
     const response = await request.get("/api/attacks");
-    // Backend may return 500 due to stale DB schema or 404 if not implemented.
-    // Only assert when the endpoint is actually healthy.
-    if (!response.ok()) {
-      test.skip(true, `GET /api/attacks returned ${response.status()} — skipping`);
-      return;
-    }
     expect(response.ok()).toBe(true);
   });
 });

frontend/e2e/chat.spec.ts

@@ -500,8 +500,7 @@ test.describe("Multi-modal: Video response", () => {
     },
   ]);
 
-  // Marking skipped for now because this should not use the getByRole query which is too general
-  test.skip("should display video player for video response", async ({ page }) => {
+  test("should display video player for video response", async ({ page }) => {
     await setupVideoMock(page);
     await page.goto("/");
     await activateMockTarget(page);