Spaces:
Sleeping
Sleeping
File size: 14,203 Bytes
2b22a59 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 |
"""
utils/database.py - Database Schema Management (Phase 4A)
========================================================
Extends the evaluation_results database with quality scoring tables:
- evaluation_scores: Multi-dimensional quality scores from LLM judge
- error_analysis: Categorized failure patterns
"""
import sqlite3
from pathlib import Path
from typing import Optional
import json
from datetime import datetime
class EvaluationDatabase:
"""
Manages SQLite database schema for RAG evaluation results
Phase 3 Tables:
- evaluation_results: Basic evaluation metrics (accuracy, time, cost)
Phase 4A Tables (NEW):
- evaluation_scores: Quality scores from LLM judge
- error_analysis: Error categorization and patterns
"""
def __init__(self, db_path: str = "data/evaluation_results.db"):
"""
Initialize database connection
Args:
db_path: Path to SQLite database file
"""
self.db_path = Path(db_path)
self.db_path.parent.mkdir(parents=True, exist_ok=True)
self.conn = sqlite3.connect(str(self.db_path))
self.conn.row_factory = sqlite3.Row # Access columns by name
def create_phase4_tables(self):
"""
Create Phase 4A tables for quality evaluation
These tables extend evaluation_results with judge scores and error analysis.
"""
cursor = self.conn.cursor()
# ===================================================================
# Table 1: evaluation_scores
# ===================================================================
cursor.execute("""
CREATE TABLE IF NOT EXISTS evaluation_scores (
id INTEGER PRIMARY KEY AUTOINCREMENT,
evaluation_result_id INTEGER NOT NULL,
-- Multi-dimensional scores (0-10)
correctness_score REAL NOT NULL,
relevance_score REAL NOT NULL,
completeness_score REAL NOT NULL,
clarity_score REAL NOT NULL,
conciseness_score REAL NOT NULL,
overall_score REAL NOT NULL,
-- Judge metadata
confidence REAL NOT NULL,
explanation TEXT NOT NULL,
issues TEXT NOT NULL, -- JSON array of issue types
-- Evaluation metadata
evaluator_model TEXT NOT NULL,
evaluation_cost_usd REAL NOT NULL,
evaluation_time_ms REAL NOT NULL,
timestamp TEXT NOT NULL,
-- Foreign key to evaluation_results
FOREIGN KEY (evaluation_result_id) REFERENCES evaluation_results(id)
ON DELETE CASCADE
)
""")
# Index for fast lookups by evaluation_result_id
cursor.execute("""
CREATE INDEX IF NOT EXISTS idx_scores_result_id
ON evaluation_scores(evaluation_result_id)
""")
# Index for filtering by overall score
cursor.execute("""
CREATE INDEX IF NOT EXISTS idx_scores_overall
ON evaluation_scores(overall_score)
""")
# ===================================================================
# Table 2: error_analysis
# ===================================================================
cursor.execute("""
CREATE TABLE IF NOT EXISTS error_analysis (
id INTEGER PRIMARY KEY AUTOINCREMENT,
evaluation_result_id INTEGER NOT NULL,
-- Error classification
error_type TEXT NOT NULL, -- 'retrieval_failure', 'generation_error', 'hallucination', etc.
error_description TEXT NOT NULL,
severity TEXT NOT NULL, -- 'low', 'medium', 'high', 'critical'
-- Diagnostics
suggested_fix TEXT,
affected_component TEXT, -- 'retriever', 'generator', 'embedder', 'reranker'
-- Metadata
timestamp TEXT NOT NULL,
-- Foreign key to evaluation_results
FOREIGN KEY (evaluation_result_id) REFERENCES evaluation_results(id)
ON DELETE CASCADE
)
""")
# Index for error type analysis
cursor.execute("""
CREATE INDEX IF NOT EXISTS idx_error_type
ON error_analysis(error_type)
""")
# Index for severity filtering
cursor.execute("""
CREATE INDEX IF NOT EXISTS idx_error_severity
ON error_analysis(severity)
""")
self.conn.commit()
print(f"✅ Phase 4A tables created in: {self.db_path}")
def insert_evaluation_score(
self,
evaluation_result_id: int,
correctness_score: float,
relevance_score: float,
completeness_score: float,
clarity_score: float,
conciseness_score: float,
overall_score: float,
confidence: float,
explanation: str,
issues: list,
evaluator_model: str,
evaluation_cost_usd: float,
evaluation_time_ms: float
) -> int:
"""
Insert quality scores from LLM judge
Args:
evaluation_result_id: FK to evaluation_results table
correctness_score: 0-10
relevance_score: 0-10
completeness_score: 0-10
clarity_score: 0-10
conciseness_score: 0-10
overall_score: 0-10 weighted average
confidence: 0-1
explanation: Judge's reasoning
issues: List of issue types
evaluator_model: Judge model name
evaluation_cost_usd: Cost of evaluation
evaluation_time_ms: Evaluation latency
Returns:
ID of inserted record
"""
cursor = self.conn.cursor()
cursor.execute("""
INSERT INTO evaluation_scores (
evaluation_result_id,
correctness_score, relevance_score, completeness_score,
clarity_score, conciseness_score, overall_score,
confidence, explanation, issues,
evaluator_model, evaluation_cost_usd, evaluation_time_ms,
timestamp
) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
""", (
evaluation_result_id,
correctness_score, relevance_score, completeness_score,
clarity_score, conciseness_score, overall_score,
confidence, explanation, json.dumps(issues),
evaluator_model, evaluation_cost_usd, evaluation_time_ms,
datetime.now().isoformat()
))
self.conn.commit()
return cursor.lastrowid
def insert_error_analysis(
self,
evaluation_result_id: int,
error_type: str,
error_description: str,
severity: str = "medium",
suggested_fix: Optional[str] = None,
affected_component: Optional[str] = None
) -> int:
"""
Insert error analysis record
Args:
evaluation_result_id: FK to evaluation_results table
error_type: 'retrieval_failure', 'generation_error', 'hallucination', etc.
error_description: Human-readable description
severity: 'low', 'medium', 'high', 'critical'
suggested_fix: Recommended solution
affected_component: 'retriever', 'generator', 'embedder', 'reranker'
Returns:
ID of inserted record
"""
cursor = self.conn.cursor()
cursor.execute("""
INSERT INTO error_analysis (
evaluation_result_id,
error_type, error_description, severity,
suggested_fix, affected_component,
timestamp
) VALUES (?, ?, ?, ?, ?, ?, ?)
""", (
evaluation_result_id,
error_type, error_description, severity,
suggested_fix, affected_component,
datetime.now().isoformat()
))
self.conn.commit()
return cursor.lastrowid
def get_evaluation_with_scores(self, evaluation_result_id: int) -> Optional[dict]:
"""
Get evaluation result with quality scores
Args:
evaluation_result_id: ID from evaluation_results table
Returns:
Dict with evaluation data + scores, or None if not found
"""
cursor = self.conn.cursor()
# Join evaluation_results with evaluation_scores
cursor.execute("""
SELECT
er.*,
es.correctness_score,
es.relevance_score,
es.completeness_score,
es.clarity_score,
es.conciseness_score,
es.overall_score,
es.confidence,
es.explanation,
es.issues,
es.evaluator_model
FROM evaluation_results er
LEFT JOIN evaluation_scores es ON er.id = es.evaluation_result_id
WHERE er.id = ?
""", (evaluation_result_id,))
row = cursor.fetchone()
if row:
return dict(row)
return None
def get_quality_summary_by_pipeline(self, run_id: str) -> list:
"""
Get quality score summary for each pipeline in a run
Args:
run_id: Evaluation run ID
Returns:
List of dicts with pipeline quality metrics
"""
cursor = self.conn.cursor()
cursor.execute("""
SELECT
er.pipeline_name,
COUNT(es.id) as evaluated_count,
ROUND(AVG(es.correctness_score), 2) as avg_correctness,
ROUND(AVG(es.relevance_score), 2) as avg_relevance,
ROUND(AVG(es.completeness_score), 2) as avg_completeness,
ROUND(AVG(es.clarity_score), 2) as avg_clarity,
ROUND(AVG(es.conciseness_score), 2) as avg_conciseness,
ROUND(AVG(es.overall_score), 2) as avg_overall,
ROUND(AVG(es.confidence), 3) as avg_confidence,
SUM(es.evaluation_cost_usd) as total_eval_cost
FROM evaluation_results er
INNER JOIN evaluation_scores es ON er.id = es.evaluation_result_id
WHERE er.run_id = ?
GROUP BY er.pipeline_name
ORDER BY avg_overall DESC
""", (run_id,))
return [dict(row) for row in cursor.fetchall()]
def get_error_summary(self, run_id: Optional[str] = None) -> list:
"""
Get error analysis summary
Args:
run_id: Optional run ID filter
Returns:
List of error type counts and severity distribution
"""
cursor = self.conn.cursor()
if run_id:
cursor.execute("""
SELECT
ea.error_type,
ea.severity,
COUNT(*) as count,
GROUP_CONCAT(DISTINCT ea.affected_component) as components
FROM error_analysis ea
INNER JOIN evaluation_results er ON ea.evaluation_result_id = er.id
WHERE er.run_id = ?
GROUP BY ea.error_type, ea.severity
ORDER BY count DESC
""", (run_id,))
else:
cursor.execute("""
SELECT
error_type,
severity,
COUNT(*) as count,
GROUP_CONCAT(DISTINCT affected_component) as components
FROM error_analysis
GROUP BY error_type, severity
ORDER BY count DESC
""")
return [dict(row) for row in cursor.fetchall()]
def close(self):
"""Close database connection"""
if self.conn:
self.conn.close()
# ============================================================================
# MIGRATION SCRIPT - Run this to update existing database
# ============================================================================
def migrate_database(db_path: str = "data/evaluation_results.db"):
"""
Migrate existing Phase 3 database to Phase 4A schema
Args:
db_path: Path to evaluation_results.db
"""
print("🔄 Migrating database to Phase 4A schema...")
print("=" * 80)
db = EvaluationDatabase(db_path)
# Create new tables
db.create_phase4_tables()
# Check existing data
cursor = db.conn.cursor()
cursor.execute("SELECT COUNT(*) FROM evaluation_results")
result_count = cursor.fetchone()[0]
cursor.execute("""
SELECT COUNT(*) FROM sqlite_master
WHERE type='table' AND name='evaluation_scores'
""")
scores_table_exists = cursor.fetchone()[0] > 0
print(f"\n📊 Database Status:")
print(f" - Evaluation results: {result_count} records")
print(f" - Quality scores table: {'✅ Created' if scores_table_exists else '❌ Missing'}")
db.close()
print("\n" + "=" * 80)
print("✅ Migration complete!")
print("\n🚀 Next: Run scripts/evaluate_with_judge.py to populate scores")
if __name__ == "__main__":
# Run migration
migrate_database()
|