Spaces:
Running
Running
File size: 12,427 Bytes
6246bba | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 | """
User Feedback Tracking System
Tracks user feedback on search results for continuous improvement:
- Thumbs up/down on answers
- Relevance ratings on sources
- Intent classification accuracy
- Search strategy effectiveness
Stores feedback in ClickHouse for analysis and model improvement.
"""
import logging
from typing import Dict, List, Any, Optional
from datetime import datetime
from dataclasses import dataclass, asdict
import json
logger = logging.getLogger(__name__)
@dataclass
class FeedbackEvent:
"""User feedback event"""
# Identifiers
session_id: str
query_id: str
user_id: Optional[int]
# Query info
query: str
expanded_query: Optional[str]
# Classification info
intent_classified: str
intent_confidence: float
intent_method: str
# Search info
search_strategy: str
live_results_count: int
db_results_count: int
total_sources: int
# Feedback
feedback_type: str # "thumbs_up", "thumbs_down", "source_rating", "intent_correction"
feedback_value: Any # True/False for thumbs, 1-5 for rating, corrected intent for correction
feedback_comment: Optional[str]
# Metadata
timestamp: str
response_time_ms: float
cache_hit: bool
class FeedbackTracker:
"""
Track and store user feedback for continuous improvement.
Features:
- Multiple feedback types (thumbs, ratings, corrections)
- ClickHouse storage for analytics
- Async logging (non-blocking)
- Aggregation and reporting
"""
def __init__(self, analytics_db=None):
"""
Initialize feedback tracker.
Args:
analytics_db: ClickHouse analytics database adapter
"""
self.analytics_db = analytics_db
self._ensure_table_exists()
def _ensure_table_exists(self):
"""Create feedback table if it doesn't exist"""
if not self.analytics_db:
return
try:
create_table_query = """
CREATE TABLE IF NOT EXISTS user_feedback (
session_id String,
query_id String,
user_id Nullable(Int32),
query String,
expanded_query Nullable(String),
intent_classified String,
intent_confidence Float32,
intent_method String,
search_strategy String,
live_results_count Int32,
db_results_count Int32,
total_sources Int32,
feedback_type String,
feedback_value String,
feedback_comment Nullable(String),
timestamp DateTime,
response_time_ms Float32,
cache_hit UInt8
) ENGINE = MergeTree()
ORDER BY (timestamp, session_id)
"""
self.analytics_db.execute(create_table_query)
logger.info("✅ Feedback table ensured")
except Exception as e:
logger.error(f"Failed to create feedback table: {e}")
def record_feedback(
self,
session_id: str,
query: str,
feedback_type: str,
feedback_value: Any,
query_metadata: Dict[str, Any],
feedback_comment: Optional[str] = None,
user_id: Optional[int] = None
):
"""
Record user feedback.
Args:
session_id: User session ID
query: Original query
feedback_type: Type of feedback (thumbs_up, thumbs_down, etc.)
feedback_value: Feedback value
query_metadata: Metadata about the query and response
feedback_comment: Optional comment from user
user_id: Optional user ID
"""
try:
# Create feedback event
event = FeedbackEvent(
session_id=session_id,
query_id=query_metadata.get("query_id", f"{session_id}_{int(datetime.utcnow().timestamp())}"),
user_id=user_id,
query=query,
expanded_query=query_metadata.get("expanded_query"),
intent_classified=query_metadata.get("intent", "UNKNOWN"),
intent_confidence=query_metadata.get("intent_confidence", 0.0),
intent_method=query_metadata.get("intent_method", "unknown"),
search_strategy=query_metadata.get("search_strategy", "unknown"),
live_results_count=query_metadata.get("live_results_count", 0),
db_results_count=query_metadata.get("db_results_count", 0),
total_sources=query_metadata.get("total_sources", 0),
feedback_type=feedback_type,
feedback_value=str(feedback_value),
feedback_comment=feedback_comment,
timestamp=datetime.utcnow().isoformat(),
response_time_ms=query_metadata.get("response_time_ms", 0.0),
cache_hit=query_metadata.get("cache_hit", False)
)
# Store in ClickHouse
if self.analytics_db:
self._store_feedback(event)
# Log feedback
logger.info(
f"Feedback recorded: {feedback_type}={feedback_value} "
f"for query='{query}' (intent={event.intent_classified})"
)
except Exception as e:
logger.error(f"Failed to record feedback: {e}")
def _store_feedback(self, event: FeedbackEvent):
"""Store feedback event in ClickHouse"""
try:
insert_query = """
INSERT INTO user_feedback (
session_id, query_id, user_id,
query, expanded_query,
intent_classified, intent_confidence, intent_method,
search_strategy, live_results_count, db_results_count, total_sources,
feedback_type, feedback_value, feedback_comment,
timestamp, response_time_ms, cache_hit
) VALUES
"""
values = (
event.session_id,
event.query_id,
event.user_id,
event.query,
event.expanded_query,
event.intent_classified,
event.intent_confidence,
event.intent_method,
event.search_strategy,
event.live_results_count,
event.db_results_count,
event.total_sources,
event.feedback_type,
event.feedback_value,
event.feedback_comment,
event.timestamp,
event.response_time_ms,
1 if event.cache_hit else 0
)
self.analytics_db.execute(insert_query, [values])
except Exception as e:
logger.error(f"Failed to store feedback in ClickHouse: {e}")
def get_feedback_stats(self, days: int = 7) -> Dict[str, Any]:
"""
Get feedback statistics for the last N days.
Args:
days: Number of days to analyze
Returns:
Dictionary with feedback statistics
"""
if not self.analytics_db:
return {}
try:
query = f"""
SELECT
feedback_type,
COUNT(*) as count,
AVG(intent_confidence) as avg_confidence,
AVG(response_time_ms) as avg_response_time,
SUM(cache_hit) / COUNT(*) as cache_hit_rate
FROM user_feedback
WHERE timestamp >= now() - INTERVAL {days} DAY
GROUP BY feedback_type
ORDER BY count DESC
"""
results = self.analytics_db.query(query)
stats = {
"total_feedback": sum(r["count"] for r in results),
"by_type": {
r["feedback_type"]: {
"count": r["count"],
"avg_confidence": r["avg_confidence"],
"avg_response_time": r["avg_response_time"],
"cache_hit_rate": r["cache_hit_rate"]
}
for r in results
},
"period_days": days
}
return stats
except Exception as e:
logger.error(f"Failed to get feedback stats: {e}")
return {}
def get_intent_accuracy(self, days: int = 7) -> Dict[str, Any]:
"""
Get intent classification accuracy based on user corrections.
Args:
days: Number of days to analyze
Returns:
Dictionary with accuracy metrics
"""
if not self.analytics_db:
return {}
try:
query = f"""
SELECT
intent_classified,
COUNT(*) as total,
SUM(CASE WHEN feedback_type = 'intent_correction' THEN 1 ELSE 0 END) as corrections,
AVG(intent_confidence) as avg_confidence
FROM user_feedback
WHERE timestamp >= now() - INTERVAL {days} DAY
GROUP BY intent_classified
ORDER BY total DESC
"""
results = self.analytics_db.query(query)
accuracy = {
"by_intent": {
r["intent_classified"]: {
"total": r["total"],
"corrections": r["corrections"],
"accuracy": 1.0 - (r["corrections"] / r["total"]) if r["total"] > 0 else 0.0,
"avg_confidence": r["avg_confidence"]
}
for r in results
},
"period_days": days
}
return accuracy
except Exception as e:
logger.error(f"Failed to get intent accuracy: {e}")
return {}
def get_low_confidence_queries(self, threshold: float = 0.7, limit: int = 100) -> List[Dict[str, Any]]:
"""
Get queries with low intent classification confidence.
Args:
threshold: Confidence threshold (queries below this)
limit: Maximum number of queries to return
Returns:
List of low-confidence queries
"""
if not self.analytics_db:
return []
try:
query = f"""
SELECT
query,
intent_classified,
intent_confidence,
intent_method,
COUNT(*) as occurrences
FROM user_feedback
WHERE intent_confidence < {threshold}
GROUP BY query, intent_classified, intent_confidence, intent_method
ORDER BY occurrences DESC, intent_confidence ASC
LIMIT {limit}
"""
results = self.analytics_db.query(query)
return results
except Exception as e:
logger.error(f"Failed to get low confidence queries: {e}")
return []
# ═══════════════════════════════════════════════════════════════════════════
# SINGLETON INSTANCE
# ═══════════════════════════════════════════════════════════════════════════
# Will be initialized with dependencies in main.py
feedback_tracker: Optional[FeedbackTracker] = None
def initialize_feedback_tracker(analytics_db=None):
"""Initialize global feedback tracker instance"""
global feedback_tracker
feedback_tracker = FeedbackTracker(analytics_db)
logger.info("Feedback tracker initialized")
|