fix(metrics): add gauge method and use it for rerank avg_score

- Added gauge() method to Metrics class for point-in-time values
- Use gauge instead of timing for avg_score (semantic correctness)

Author: DevanshuNEU

backend/services/observability.py

@@ -324,6 +324,7 @@ class Metrics:
     def __init__(self):
         self._counters: Dict[str, int] = {}
         self._timings: Dict[str, list] = {}
+        self._gauges: Dict[str, float] = {}
 
     def increment(self, name: str, value: int = 1, **tags):
         """Increment a counter"""
@@ -339,10 +340,15 @@ def timing(self, name: str, value_ms: float):
         if len(self._timings[name]) > 1000:
             self._timings[name] = self._timings[name][-1000:]
 
+    def gauge(self, name: str, value: float):
+        """Record a point-in-time value (like avg score, current queue size)"""
+        self._gauges[name] = value
+    
     def get_stats(self) -> Dict:
         """Get all metrics with basic stats"""
         stats = {
             "counters": self._counters.copy(),
+            "gauges": self._gauges.copy(),
             "timings": {}
         }
 
@@ -361,6 +367,7 @@ def reset(self):
         """Reset all metrics"""
         self._counters = {}
         self._timings = {}
+        self._gauges = {}
 
 
 # Global metrics instance

backend/services/search_v3/search_engine.py

@@ -327,7 +327,7 @@ async def _rerank_results(
 
             # metrics for observability
             avg_score = sum(r['rerank_score'] for r in reranked) / len(reranked) if reranked else 0
-            metrics.timing("search.rerank.avg_score", avg_score * 100)  # scale to percentage
+            metrics.gauge("search.rerank.avg_score", avg_score * 100)  # scale to percentage
             metrics.increment("search.rerank.success")
 
             logger.info("Cohere rerank complete",