File size: 7,857 Bytes
2fb680d |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 |
import asyncio
import hashlib
import time
import psutil
import gc
from typing import Dict, List, Optional, Any
from collections import OrderedDict
from dataclasses import dataclass
from datetime import datetime, timedelta
import redis
import json
@dataclass
class CacheEntry:
response: str
timestamp: datetime
hit_count: int = 0
class PerformanceOptimizer:
def __init__(self,
cache_size: int = 100,
cache_ttl_hours: int = 24,
use_redis: bool = False):
"""Initialize performance optimizer with caching"""
self.cache_size = cache_size
self.cache_ttl = timedelta(hours=cache_ttl_hours)
# Use Redis if available, fallback to in-memory
self.use_redis = use_redis
if use_redis:
try:
self.redis_client = redis.Redis(
host='localhost',
port=6379,
decode_responses=True
)
self.redis_client.ping()
except:
print("Redis not available, using in-memory cache")
self.use_redis = False
self.cache = OrderedDict()
else:
self.cache = OrderedDict()
# Metrics
self.metrics = {
"cache_hits": 0,
"cache_misses": 0,
"total_requests": 0,
"average_response_time": 0,
"memory_usage_mb": 0
}
def _hash_prompt(self, prompt: str) -> str:
"""Create hash for caching"""
normalized = prompt.lower().strip()
return hashlib.md5(normalized.encode()).hexdigest()
def get_cached_response(self, prompt: str) -> Optional[str]:
"""Get response from cache if available"""
self.metrics["total_requests"] += 1
prompt_hash = self._hash_prompt(prompt)
if self.use_redis:
cached = self.redis_client.get(f"chat:{prompt_hash}")
if cached:
self.metrics["cache_hits"] += 1
# Update hit count
self.redis_client.hincrby(f"chat:stats:{prompt_hash}", "hits", 1)
return json.loads(cached)["response"]
else:
if prompt_hash in self.cache:
entry = self.cache[prompt_hash]
# Check TTL
if datetime.now() - entry.timestamp < self.cache_ttl:
self.metrics["cache_hits"] += 1
entry.hit_count += 1
# Move to end (LRU)
self.cache.move_to_end(prompt_hash)
return entry.response
else:
# Expired
del self.cache[prompt_hash]
self.metrics["cache_misses"] += 1
return None
def cache_response(self, prompt: str, response: str):
"""Cache a response"""
prompt_hash = self._hash_prompt(prompt)
if self.use_redis:
cache_data = {
"response": response,
"timestamp": datetime.now().isoformat()
}
self.redis_client.setex(
f"chat:{prompt_hash}",
int(self.cache_ttl.total_seconds()),
json.dumps(cache_data)
)
self.redis_client.hset(
f"chat:stats:{prompt_hash}",
mapping={"hits": 0, "created": datetime.now().isoformat()}
)
else:
# LRU cache management
if len(self.cache) >= self.cache_size:
# Remove least recently used
self.cache.popitem(last=False)
self.cache[prompt_hash] = CacheEntry(
response=response,
timestamp=datetime.now()
)
def get_metrics(self) -> Dict[str, Any]:
"""Get performance metrics"""
# Update memory usage
process = psutil.Process()
self.metrics["memory_usage_mb"] = process.memory_info().rss / 1024 / 1024
# Calculate cache hit rate
if self.metrics["total_requests"] > 0:
self.metrics["cache_hit_rate"] = (
self.metrics["cache_hits"] / self.metrics["total_requests"]
)
return self.metrics
def clear_cache(self):
"""Clear all cached responses"""
if self.use_redis:
for key in self.redis_client.scan_iter("chat:*"):
self.redis_client.delete(key)
else:
self.cache.clear()
gc.collect()
class MemoryManager:
def __init__(self, max_memory_mb: int = 8192):
"""Initialize memory manager"""
self.max_memory_mb = max_memory_mb
self.warning_threshold = 0.8 # Warn at 80% usage
self.critical_threshold = 0.9 # Critical at 90% usage
def check_memory(self) -> Dict[str, Any]:
"""Check current memory usage"""
process = psutil.Process()
memory_info = process.memory_info()
current_mb = memory_info.rss / 1024 / 1024
percentage = current_mb / self.max_memory_mb
status = "normal"
if percentage > self.critical_threshold:
status = "critical"
elif percentage > self.warning_threshold:
status = "warning"
return {
"current_mb": round(current_mb, 2),
"max_mb": self.max_memory_mb,
"percentage": round(percentage * 100, 2),
"status": status,
"available_mb": round(self.max_memory_mb - current_mb, 2)
}
def optimize_if_needed(self) -> bool:
"""Run optimization if memory usage is high"""
memory_status = self.check_memory()
if memory_status["status"] in ["warning", "critical"]:
# Force garbage collection
gc.collect()
# Clear unused objects
if memory_status["status"] == "critical":
# More aggressive cleanup
gc.collect(2)
return True
return False
class RequestBatcher:
def __init__(self, batch_size: int = 5, timeout_ms: int = 100):
"""Initialize request batcher for efficiency"""
self.batch_size = batch_size
self.timeout_ms = timeout_ms
self.pending_requests = []
self.results = {}
async def add_request(self, request_id: str, prompt: str) -> str:
"""Add request to batch"""
self.pending_requests.append({
"id": request_id,
"prompt": prompt,
"timestamp": time.time()
})
# Process if batch is full
if len(self.pending_requests) >= self.batch_size:
await self._process_batch()
else:
# Wait for timeout
await asyncio.sleep(self.timeout_ms / 1000)
if request_id not in self.results:
await self._process_batch()
return self.results.get(request_id, "Error processing request")
async def _process_batch(self):
"""Process pending requests as batch"""
if not self.pending_requests:
return
batch = self.pending_requests[:self.batch_size]
self.pending_requests = self.pending_requests[self.batch_size:]
# Process batch (simulate concurrent processing)
tasks = []
for request in batch:
# In production, this would call the LLM
tasks.append(self._process_single(request))
results = await asyncio.gather(*tasks)
for request, result in zip(batch, results):
self.results[request["id"]] = result
async def _process_single(self, request: Dict[str, Any]) -> str:
"""Process single request (placeholder)"""
# Simulate processing
await asyncio.sleep(0.1)
return f"Response to: {request['prompt']}"
|