Spaces:
Running
Running
File size: 7,987 Bytes
72bff80 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 |
"""
Application metrics collection and tracking.
"""
import time
import threading
from typing import Dict, List, Optional
from collections import defaultdict, deque
from datetime import datetime, timedelta
from config import config
class MetricsCollector:
"""Thread-safe metrics collector."""
def __init__(self):
self._lock = threading.Lock()
# Request metrics
self.request_count = 0
self.request_latencies = deque(maxlen=1000) # Keep last 1000
self.request_errors = 0
# Intent distribution
self.intent_counts = defaultdict(int)
# LLM metrics
self.llm_call_count = 0
self.llm_cache_hits = 0
self.llm_cache_misses = 0
self.llm_latencies = deque(maxlen=1000)
self.llm_errors = 0
# Retrieval metrics
self.retrieval_count = 0
self.retrieval_latencies = deque(maxlen=1000)
self.retrieval_empty_results = 0
# Cache metrics
self.cache_hits = 0
self.cache_misses = 0
# Circuit breaker metrics
self.circuit_breaker_opens = 0
self.circuit_breaker_failures = 0
# Active requests
self.active_requests = 0
# Start time
self.start_time = datetime.now()
def record_request(self, latency_ms: float, intent: Optional[str] = None, error: bool = False):
"""Record a request."""
with self._lock:
self.request_count += 1
self.request_latencies.append(latency_ms)
if error:
self.request_errors += 1
if intent:
self.intent_counts[intent] += 1
def record_llm_call(self, latency_ms: float, cache_hit: bool = False, error: bool = False):
"""Record an LLM call."""
with self._lock:
self.llm_call_count += 1
self.llm_latencies.append(latency_ms)
if cache_hit:
self.llm_cache_hits += 1
else:
self.llm_cache_misses += 1
if error:
self.llm_errors += 1
def record_retrieval(self, latency_ms: float, result_count: int):
"""Record a retrieval operation."""
with self._lock:
self.retrieval_count += 1
self.retrieval_latencies.append(latency_ms)
if result_count == 0:
self.retrieval_empty_results += 1
def record_cache_access(self, hit: bool):
"""Record cache access."""
with self._lock:
if hit:
self.cache_hits += 1
else:
self.cache_misses += 1
def record_circuit_breaker_event(self, opened: bool = False, failure: bool = False):
"""Record circuit breaker event."""
with self._lock:
if opened:
self.circuit_breaker_opens += 1
if failure:
self.circuit_breaker_failures += 1
def increment_active_requests(self):
"""Increment active request count."""
with self._lock:
self.active_requests += 1
def decrement_active_requests(self):
"""Decrement active request count."""
with self._lock:
self.active_requests = max(0, self.active_requests - 1)
def get_metrics(self) -> Dict:
"""Get all metrics as a dictionary."""
with self._lock:
uptime = datetime.now() - self.start_time
# Calculate percentiles for latencies
req_latencies_sorted = sorted(self.request_latencies) if self.request_latencies else [0]
llm_latencies_sorted = sorted(self.llm_latencies) if self.llm_latencies else [0]
ret_latencies_sorted = sorted(self.retrieval_latencies) if self.retrieval_latencies else [0]
def percentile(data, p):
if not data:
return 0
k = (len(data) - 1) * p
f = int(k)
c = k - f
if f + 1 < len(data):
return data[f] * (1 - c) + data[f + 1] * c
return data[f]
return {
"uptime_seconds": uptime.total_seconds(),
"timestamp": datetime.now().isoformat(),
# Request metrics
"requests": {
"total": self.request_count,
"active": self.active_requests,
"errors": self.request_errors,
"error_rate": self.request_errors / max(1, self.request_count),
"latency_ms": {
"min": min(req_latencies_sorted),
"max": max(req_latencies_sorted),
"p50": percentile(req_latencies_sorted, 0.50),
"p95": percentile(req_latencies_sorted, 0.95),
"p99": percentile(req_latencies_sorted, 0.99),
}
},
# Intent distribution
"intents": dict(self.intent_counts),
# LLM metrics
"llm": {
"total_calls": self.llm_call_count,
"cache_hits": self.llm_cache_hits,
"cache_misses": self.llm_cache_misses,
"cache_hit_rate": self.llm_cache_hits / max(1, self.llm_call_count),
"errors": self.llm_errors,
"latency_ms": {
"min": min(llm_latencies_sorted),
"max": max(llm_latencies_sorted),
"p50": percentile(llm_latencies_sorted, 0.50),
"p95": percentile(llm_latencies_sorted, 0.95),
}
},
# Retrieval metrics
"retrieval": {
"total_searches": self.retrieval_count,
"empty_results": self.retrieval_empty_results,
"empty_result_rate": self.retrieval_empty_results / max(1, self.retrieval_count),
"latency_ms": {
"min": min(ret_latencies_sorted),
"max": max(ret_latencies_sorted),
"p50": percentile(ret_latencies_sorted, 0.50),
"p95": percentile(ret_latencies_sorted, 0.95),
}
},
# Cache metrics
"cache": {
"hits": self.cache_hits,
"misses": self.cache_misses,
"hit_rate": self.cache_hits / max(1, self.cache_hits + self.cache_misses),
},
# Circuit breaker metrics
"circuit_breaker": {
"opens": self.circuit_breaker_opens,
"failures": self.circuit_breaker_failures,
}
}
def reset_metrics(self):
"""Reset all metrics (use with caution)."""
with self._lock:
self.request_count = 0
self.request_latencies.clear()
self.request_errors = 0
self.intent_counts.clear()
self.llm_call_count = 0
self.llm_cache_hits = 0
self.llm_cache_misses = 0
self.llm_latencies.clear()
self.llm_errors = 0
self.retrieval_count = 0
self.retrieval_latencies.clear()
self.retrieval_empty_results = 0
self.cache_hits = 0
self.cache_misses = 0
self.circuit_breaker_opens = 0
self.circuit_breaker_failures = 0
self.start_time = datetime.now()
# Global metrics instance
metrics = MetricsCollector()
|