feat(quality): A4 — L2 Pro premium judge (8 dims, gpt-4o, Pro-gated, opt-in)

middleware/metrics.py

@@ -70,6 +70,28 @@ quality_l1_judge_cost_usd = Histogram(
     buckets=(0.0001, 0.0005, 0.001, 0.005, 0.01, 0.05, 0.1),
 )
 
+# ---- L2 Pro premium judge (Track A4) ----
+
+quality_l2_judge_total = Counter(
+    "quality_l2_judge_total",
+    "Total L2 (Pro premium judge, 8-dim) verdicts",
+    ["verdict", "model", "tier"],  # verdict: pass | fail | skip | error
+)
+
+quality_l2_judge_duration_seconds = Histogram(
+    "quality_l2_judge_duration_seconds",
+    "L2 Pro judge call duration in seconds",
+    ["model"],
+    buckets=(0.5, 1, 2, 5, 10, 20, 30, 60),
+)
+
+quality_l2_judge_cost_usd = Histogram(
+    "quality_l2_judge_cost_usd",
+    "L2 Pro judge estimated cost in USD",
+    ["model"],
+    buckets=(0.001, 0.005, 0.01, 0.05, 0.1, 0.5, 1.0),
+)
+
 # ---- Retry metrics ----
 
 translation_retry_total = Counter(
@@ -138,6 +160,29 @@ def record_l1_verdict(
         quality_l1_judge_cost_usd.labels(model=model).observe(cost_usd)
 
 
+def record_l2_verdict(
+    verdict: str,
+    model: str = "unknown",
+    tier: str = "pro",
+    duration_seconds: float = None,
+    cost_usd: float = None,
+):
+    """Record an L2 Pro judge verdict.
+
+    Args:
+        verdict: pass | fail | skip | error
+        model: model name (e.g. gpt-4o)
+        tier: pro | business | enterprise
+        duration_seconds: optional, observed in histogram
+        cost_usd: optional, observed in histogram
+    """
+    quality_l2_judge_total.labels(verdict=verdict, model=model, tier=tier).inc()
+    if duration_seconds is not None:
+        quality_l2_judge_duration_seconds.labels(model=model).observe(duration_seconds)
+    if cost_usd is not None:
+        quality_l2_judge_cost_usd.labels(model=model).observe(cost_usd)
+
+
 def record_translation_retry(reason: str, tier: str = "free"):
     """Record a translation retry triggered by a quality issue.