Fix LLM e2e runner: precheck server, correct S10 table name

precheckServer() polls /healthz once before iterating scenarios. When
the server is down, the runner now exits 2 with a clear startup hint
instead of spawning Playwright per scenario (each timing out and
filing a duplicate "did not become healthy" task via ask add).

Also corrects S10-progress-batch.md: the YAML db assertion referenced
`progress` but the actual table (per internal/repository/playback_progress.go)
is `playback_progress`, with `media_id` as the joinable column.

Verified: precheck cleanly exits 2 when the server is down (15 s timeout);
full LLM e2e suite passes 13/13 with the server up.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

2 files changed, 68 insertions, 1 deletions

diff --git a/player-server/test/e2e-llm/runner/index.ts b/player-server/test/e2e-llm/runner/index.ts
index fb973c9..4ec03c0 100644
--- a/player-server/test/e2e-llm/runner/index.ts
+++ b/player-server/test/e2e-llm/runner/index.ts
@@ -330,6 +330,67 @@ function runScenario(scenario: Scenario): boolean {
 }
 
 // ---------------------------------------------------------------------------
+// Server precheck
+// ---------------------------------------------------------------------------
+
+// PLAYER_URL is the base URL the harness uses (matches the README + Playwright
+// config). Centralising the constant keeps the precheck and any future direct
+// HTTP calls aligned with the rest of the harness.
+const PLAYER_URL = process.env['PLAYER_URL'] || 'http://localhost:8080';
+
+// How long to wait for /healthz before declaring the server missing. The
+// Playwright suite waits up to 15 s per scenario internally; doing the same
+// up-front means we surface a missing server as a single clear error instead
+// of one Playwright failure (and one bogus ask task) per scenario.
+const PRECHECK_TIMEOUT_MS = 15_000;
+const PRECHECK_POLL_INTERVAL_MS = 200;
+
+/**
+ * waitForServer polls `${PLAYER_URL}/healthz` until it responds 2xx or the
+ * deadline passes. Returns true if the server became healthy, false otherwise.
+ * Using a synchronous loop (spawnSync sleep) keeps the runner's overall
+ * control flow synchronous, matching how runScenario invokes Playwright.
+ */
+function waitForServer(timeoutMs: number): boolean {
+  const deadline = Date.now() + timeoutMs;
+  while (Date.now() < deadline) {
+    // Use curl for the probe so we don't need to depend on a Node fetch
+    // shim — keeps the harness usable on older Node where fetch is missing.
+    const probe = spawnSync(
+      'curl',
+      ['-fsS', '-o', '/dev/null', '-w', '%{http_code}', `${PLAYER_URL}/healthz`],
+      { encoding: 'utf8' },
+    );
+    if (probe.status === 0 && probe.stdout.trim().startsWith('2')) return true;
+    spawnSync('sleep', [String(PRECHECK_POLL_INTERVAL_MS / 1000)]);
+  }
+  return false;
+}
+
+/**
+ * precheckServer fails fast with a clear, actionable message if the server
+ * is not reachable. Without this, each scenario would individually fail with
+ * the same "did not become healthy" error and openAskTask would file one
+ * spurious task per scenario — drowning the queue in duplicates of the same
+ * root cause. Returning a non-zero exit before the scenario loop avoids both.
+ */
+function precheckServer(): void {
+  if (waitForServer(PRECHECK_TIMEOUT_MS)) return;
+
+  console.error(
+    `[runner] ERROR: Player server at ${PLAYER_URL} is not reachable on /healthz ` +
+      `after ${PRECHECK_TIMEOUT_MS / 1000}s.`,
+  );
+  console.error('[runner] Start it from player-server/ before running the e2e suite:');
+  console.error(
+    '[runner]   MEDIA_ROOT=./testmedia SECURE_COOKIES=false ' +
+      'DB_PATH=/tmp/player-e2e-llm.db ./player',
+  );
+  console.error('[runner] Override the target URL with PLAYER_URL.');
+  process.exit(2);
+}
+
+// ---------------------------------------------------------------------------
 // Entry point
 // ---------------------------------------------------------------------------
 
@@ -346,6 +407,12 @@ function main(): void {
     process.exit(0);
   }
 
+  // Fail fast with a clear hint if the server is down. Without this the
+  // runner would spawn Playwright per scenario, time out on each, and file
+  // a duplicate ask task per failure — exactly what produced the legacy
+  // "Server at http://localhost:8080 did not become healthy" task pile.
+  precheckServer();
+
   console.log(`[runner] Running ${scenarios.length} scenario(s)…`);
 
   let failures = 0;
diff --git a/player-server/test/e2e-llm/scenarios/S10-progress-batch.md b/player-server/test/e2e-llm/scenarios/S10-progress-batch.md
index ad6ead9..a5a635c 100644
--- a/player-server/test/e2e-llm/scenarios/S10-progress-batch.md
+++ b/player-server/test/e2e-llm/scenarios/S10-progress-batch.md
@@ -6,7 +6,7 @@ preconditions:
   server_state: running        # server running with an existing admin account and at least two media items
   fixtures: []
 assertions:
-  - db: "SELECT id FROM progress WHERE position_seconds=30"
+  - db: "SELECT media_id FROM playback_progress WHERE position_seconds=30"
   - status_code: "POST /api/v1/progress/batch 200"
 skip: false
 ---