Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,397 +1,518 @@
|
|
| 1 |
"""
|
| 2 |
-
app.py
|
| 3 |
|
| 4 |
-
|
| 5 |
-
✅
|
| 6 |
-
✅
|
| 7 |
-
✅
|
| 8 |
-
✅
|
| 9 |
-
✅ Gestion erreurs améliorée pour compatibilité
|
| 10 |
-
✅ Optimisations HF Space gratuit renforcées
|
| 11 |
✅ Variables d'environnement sécurisées
|
|
|
|
|
|
|
| 12 |
"""
|
| 13 |
|
| 14 |
import os
|
| 15 |
import logging
|
| 16 |
import time
|
|
|
|
|
|
|
|
|
|
| 17 |
from fastapi import FastAPI, HTTPException, Request
|
| 18 |
from fastapi.middleware.cors import CORSMiddleware
|
| 19 |
from fastapi.responses import JSONResponse
|
| 20 |
from pydantic import BaseModel, Field
|
| 21 |
from typing import List, Dict, Any, Optional
|
| 22 |
-
import asyncio
|
| 23 |
from concurrent.futures import ThreadPoolExecutor
|
| 24 |
-
import gc
|
| 25 |
|
| 26 |
# Configuration logging optimisée
|
| 27 |
logging.basicConfig(
|
| 28 |
level=logging.INFO,
|
| 29 |
-
format="%(asctime)s - %(name)s - %(levelname)s - %(message)s"
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
)
|
| 31 |
logger = logging.getLogger(__name__)
|
| 32 |
|
| 33 |
-
# ✅
|
| 34 |
try:
|
| 35 |
-
from chunker_pipeline import SmartChunkerPipeline
|
| 36 |
from schemas import ChunkRequest, ChunkResponse, ChunkMetadata
|
| 37 |
-
logger.info("✅ Modules
|
| 38 |
except ImportError as e:
|
| 39 |
-
logger.error(f"❌
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
from pipeline import ChunkingPipeline as SmartChunkerPipeline
|
| 43 |
-
from schemas import ChunkRequest, ChunkResponse, ChunkMetadata
|
| 44 |
-
logger.warning("⚠️ Utilisation pipeline fallback - fonctionnalités limitées")
|
| 45 |
-
except ImportError as e2:
|
| 46 |
-
logger.error(f"❌ Erreur import fallback: {e2}")
|
| 47 |
-
raise
|
| 48 |
|
| 49 |
-
# ✅
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
|
| 56 |
-
#
|
| 57 |
-
|
| 58 |
|
| 59 |
-
#
|
| 60 |
app = FastAPI(
|
| 61 |
-
title="Chunking Sémantique Intelligent API
|
| 62 |
-
description="
|
| 63 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
docs_url="/docs",
|
| 65 |
-
redoc_url="/redoc"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 66 |
)
|
| 67 |
|
| 68 |
-
#
|
| 69 |
app.add_middleware(
|
| 70 |
CORSMiddleware,
|
| 71 |
-
allow_origins=["*"], #
|
| 72 |
allow_credentials=True,
|
| 73 |
-
allow_methods=["GET", "POST", "OPTIONS"],
|
| 74 |
allow_headers=["*"],
|
|
|
|
| 75 |
)
|
| 76 |
|
| 77 |
-
# ✅
|
| 78 |
pipeline = None
|
| 79 |
executor = ThreadPoolExecutor(max_workers=1) # HF Space gratuit = 1 worker max
|
|
|
|
|
|
|
| 80 |
|
| 81 |
-
# ✅
|
| 82 |
@app.middleware("http")
|
| 83 |
-
async def
|
| 84 |
-
"""Middleware pour
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
try:
|
| 86 |
response = await call_next(request)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 87 |
return response
|
|
|
|
| 88 |
except Exception as e:
|
| 89 |
-
logger.error(f"❌ Erreur
|
|
|
|
|
|
|
| 90 |
return JSONResponse(
|
| 91 |
status_code=500,
|
| 92 |
content={
|
| 93 |
"error": "Erreur interne du serveur",
|
| 94 |
"detail": str(e),
|
|
|
|
| 95 |
"timestamp": time.time(),
|
| 96 |
-
"
|
|
|
|
| 97 |
}
|
| 98 |
)
|
| 99 |
|
|
|
|
| 100 |
@app.on_event("startup")
|
| 101 |
async def startup_event():
|
| 102 |
-
"""
|
| 103 |
global pipeline
|
|
|
|
| 104 |
try:
|
| 105 |
-
logger.info("🚀
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 106 |
|
| 107 |
-
#
|
|
|
|
| 108 |
pipeline = SmartChunkerPipeline()
|
| 109 |
await pipeline.initialize()
|
| 110 |
|
| 111 |
-
#
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
|
| 129 |
-
logger.info("
|
| 130 |
|
| 131 |
except Exception as e:
|
| 132 |
-
logger.error(f"❌
|
|
|
|
| 133 |
raise
|
| 134 |
|
| 135 |
@app.on_event("shutdown")
|
| 136 |
async def shutdown_event():
|
| 137 |
-
"""
|
| 138 |
global pipeline, executor
|
|
|
|
| 139 |
try:
|
| 140 |
-
logger.info("🛑 Arrêt du service
|
| 141 |
|
|
|
|
| 142 |
if pipeline:
|
| 143 |
-
|
| 144 |
-
|
| 145 |
-
await pipeline.cleanup()
|
| 146 |
-
elif hasattr(pipeline, '_cleanup_memory_v4'):
|
| 147 |
-
await pipeline._cleanup_memory_v4()
|
| 148 |
|
|
|
|
| 149 |
if executor:
|
| 150 |
-
executor.shutdown(wait=True)
|
|
|
|
| 151 |
|
| 152 |
# Nettoyage mémoire final
|
| 153 |
gc.collect()
|
| 154 |
|
| 155 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 156 |
|
| 157 |
except Exception as e:
|
| 158 |
-
logger.error(f"⚠️ Erreur lors
|
|
|
|
|
|
|
| 159 |
|
| 160 |
-
@app.get("/")
|
| 161 |
async def root():
|
| 162 |
-
"""
|
|
|
|
|
|
|
| 163 |
return {
|
| 164 |
-
"service": "Chunking Sémantique Intelligent API",
|
| 165 |
-
"version": "
|
| 166 |
-
"status": "
|
| 167 |
-
"
|
|
|
|
|
|
|
| 168 |
"features": [
|
| 169 |
-
"Chunking sémantique avec Chonkie",
|
| 170 |
-
"
|
| 171 |
-
"
|
| 172 |
-
"
|
| 173 |
-
"
|
| 174 |
-
"
|
| 175 |
-
"
|
| 176 |
-
],
|
| 177 |
-
"endpoints": [
|
| 178 |
-
"GET / - Informations service",
|
| 179 |
-
"GET /health - Vérification santé",
|
| 180 |
-
"GET /config - Configuration système",
|
| 181 |
-
"POST /chunk - Chunking principal",
|
| 182 |
-
"POST /chunk-batch - Chunking par lots",
|
| 183 |
-
"POST /test - Test de fonctionnement"
|
| 184 |
],
|
| 185 |
-
|
| 186 |
-
"
|
| 187 |
-
"
|
| 188 |
-
|
| 189 |
-
|
| 190 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 191 |
}
|
| 192 |
|
| 193 |
-
@app.get("/health")
|
| 194 |
async def health_check():
|
| 195 |
-
"""
|
| 196 |
try:
|
| 197 |
if pipeline is None:
|
| 198 |
return {
|
| 199 |
-
"status": "error",
|
| 200 |
"message": "Pipeline non initialisé",
|
| 201 |
-
"version": "
|
| 202 |
-
"timestamp": time.time()
|
|
|
|
|
|
|
| 203 |
}
|
| 204 |
|
| 205 |
-
#
|
| 206 |
-
health_result =
|
| 207 |
-
memory_info = None
|
| 208 |
|
| 209 |
-
#
|
| 210 |
-
|
| 211 |
-
|
| 212 |
-
|
| 213 |
-
|
| 214 |
-
try:
|
| 215 |
-
health_result = await pipeline.get_health_status()
|
| 216 |
-
except AttributeError:
|
| 217 |
-
# Fallback basique
|
| 218 |
-
health_result = {
|
| 219 |
-
"status": "running",
|
| 220 |
-
"checks": {"initialization": True},
|
| 221 |
-
"version": "5.0.0"
|
| 222 |
-
}
|
| 223 |
-
|
| 224 |
-
# Tentative récupération mémoire
|
| 225 |
try:
|
| 226 |
-
|
| 227 |
-
|
| 228 |
-
|
| 229 |
-
|
| 230 |
-
|
| 231 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 232 |
|
| 233 |
return {
|
| 234 |
**health_result,
|
|
|
|
| 235 |
"memory_info": memory_info,
|
| 236 |
-
"
|
| 237 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 238 |
}
|
| 239 |
|
| 240 |
except Exception as e:
|
| 241 |
-
logger.error(f"❌ Erreur health check
|
| 242 |
return {
|
| 243 |
-
"status": "error",
|
| 244 |
"message": f"Erreur health check: {str(e)}",
|
| 245 |
-
"version": "
|
| 246 |
-
"timestamp": time.time()
|
|
|
|
| 247 |
}
|
| 248 |
|
| 249 |
-
@app.get("/config")
|
| 250 |
async def get_config():
|
| 251 |
-
"""
|
| 252 |
try:
|
| 253 |
if pipeline is None:
|
| 254 |
raise HTTPException(status_code=503, detail="Pipeline non initialisé")
|
| 255 |
|
| 256 |
-
|
| 257 |
-
config_info = {}
|
| 258 |
|
| 259 |
-
|
| 260 |
-
|
| 261 |
-
|
| 262 |
-
|
| 263 |
-
|
| 264 |
-
|
| 265 |
-
|
| 266 |
-
|
| 267 |
-
|
| 268 |
-
"embedding_available": hasattr(pipeline, 'embed_model'),
|
| 269 |
-
"chonkie_available": hasattr(pipeline, 'chonkie_semantic')
|
| 270 |
-
},
|
| 271 |
-
"chunking_config": {
|
| 272 |
-
"pipeline_type": "SmartChunkerPipeline",
|
| 273 |
-
"initialized": pipeline._is_initialized if hasattr(pipeline, '_is_initialized') else True
|
| 274 |
-
}
|
| 275 |
-
}
|
| 276 |
|
| 277 |
return {
|
| 278 |
**config_info,
|
| 279 |
-
"
|
| 280 |
-
"
|
| 281 |
-
"
|
| 282 |
}
|
| 283 |
|
| 284 |
except Exception as e:
|
| 285 |
-
logger.error(f"❌ Erreur récupération config
|
| 286 |
raise HTTPException(status_code=500, detail=f"Erreur config: {str(e)}")
|
| 287 |
|
| 288 |
-
@app.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 289 |
async def chunk_text(request: ChunkRequest):
|
| 290 |
"""
|
| 291 |
-
|
| 292 |
|
| 293 |
-
|
| 294 |
-
-
|
| 295 |
-
-
|
| 296 |
-
- Relations bidirectionnelles complètes
|
| 297 |
- Export Obsidian format [[Titre]], id
|
| 298 |
-
-
|
|
|
|
|
|
|
| 299 |
"""
|
| 300 |
if pipeline is None:
|
| 301 |
-
raise HTTPException(
|
|
|
|
|
|
|
|
|
|
| 302 |
|
| 303 |
start_time = time.time()
|
| 304 |
|
| 305 |
try:
|
| 306 |
-
logger.info(f"📝 Début chunking
|
| 307 |
|
| 308 |
# Validation entrées renforcée
|
| 309 |
if not request.text or len(request.text.strip()) < 10:
|
| 310 |
raise HTTPException(
|
| 311 |
status_code=400,
|
| 312 |
-
detail="Le texte doit contenir au moins 10 caractères"
|
| 313 |
)
|
| 314 |
|
| 315 |
-
|
|
|
|
|
|
|
| 316 |
raise HTTPException(
|
| 317 |
status_code=400,
|
| 318 |
-
detail="Texte trop long (
|
| 319 |
)
|
| 320 |
|
| 321 |
-
#
|
| 322 |
-
result =
|
| 323 |
-
|
| 324 |
-
# Tentative méthode v4.0
|
| 325 |
-
try:
|
| 326 |
-
result = await pipeline.process_text(request)
|
| 327 |
-
except AttributeError:
|
| 328 |
-
# Fallback méthode alternative
|
| 329 |
-
try:
|
| 330 |
-
result = await pipeline.process_text_sync(request)
|
| 331 |
-
except AttributeError:
|
| 332 |
-
raise HTTPException(
|
| 333 |
-
status_code=500,
|
| 334 |
-
detail="Méthode de traitement non disponible dans cette version du pipeline"
|
| 335 |
-
)
|
| 336 |
|
| 337 |
processing_time = time.time() - start_time
|
| 338 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 339 |
|
| 340 |
return result
|
| 341 |
|
| 342 |
except HTTPException:
|
| 343 |
-
# Re-lever les HTTPException sans modification
|
| 344 |
raise
|
| 345 |
except Exception as e:
|
| 346 |
-
logger.error(f"❌ Erreur chunking
|
| 347 |
|
| 348 |
-
# Nettoyage mémoire
|
| 349 |
try:
|
| 350 |
-
|
| 351 |
-
|
| 352 |
-
elif hasattr(pipeline, '_cleanup_memory'):
|
| 353 |
-
await pipeline._cleanup_memory()
|
| 354 |
except:
|
| 355 |
pass
|
| 356 |
|
| 357 |
-
gc.collect()
|
| 358 |
-
|
| 359 |
raise HTTPException(
|
| 360 |
status_code=500,
|
| 361 |
-
detail=f"Erreur traitement
|
| 362 |
)
|
| 363 |
|
| 364 |
-
@app.post("/chunk-batch")
|
| 365 |
async def chunk_batch(requests: List[ChunkRequest]):
|
| 366 |
-
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 367 |
|
| 368 |
# Validation limite batch pour Space gratuit
|
| 369 |
max_batch_size = 3
|
| 370 |
if len(requests) > max_batch_size:
|
| 371 |
raise HTTPException(
|
| 372 |
status_code=400,
|
| 373 |
-
detail=f"Maximum {max_batch_size} textes par lot sur HF Space gratuit"
|
| 374 |
)
|
| 375 |
|
| 376 |
if pipeline is None:
|
| 377 |
-
raise HTTPException(status_code=503, detail="Pipeline non initialisé")
|
| 378 |
|
| 379 |
start_time = time.time()
|
| 380 |
results = []
|
| 381 |
|
| 382 |
try:
|
| 383 |
-
logger.info(f"📦 Début
|
| 384 |
|
| 385 |
for idx, request in enumerate(requests):
|
| 386 |
try:
|
| 387 |
-
|
| 388 |
-
result = None
|
| 389 |
-
|
| 390 |
-
try:
|
| 391 |
-
result = await pipeline.process_text(request)
|
| 392 |
-
except AttributeError:
|
| 393 |
-
result = await pipeline.process_text_sync(request)
|
| 394 |
|
|
|
|
| 395 |
results.append({
|
| 396 |
"success": True,
|
| 397 |
"index": idx,
|
|
@@ -399,122 +520,187 @@ async def chunk_batch(requests: List[ChunkRequest]):
|
|
| 399 |
"result": result
|
| 400 |
})
|
| 401 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 402 |
except Exception as e:
|
| 403 |
-
logger.error(f"❌ Erreur
|
| 404 |
results.append({
|
| 405 |
"success": False,
|
| 406 |
"index": idx,
|
| 407 |
-
"source_id": request.source_id,
|
| 408 |
"error": str(e)
|
| 409 |
})
|
| 410 |
|
| 411 |
total_time = time.time() - start_time
|
| 412 |
successful_results = [r for r in results if r["success"]]
|
| 413 |
|
| 414 |
-
# Nettoyage
|
| 415 |
try:
|
| 416 |
-
|
| 417 |
-
await pipeline._cleanup_memory_v4()
|
| 418 |
except:
|
| 419 |
pass
|
| 420 |
|
| 421 |
-
logger.info(
|
|
|
|
|
|
|
|
|
|
| 422 |
|
| 423 |
return {
|
| 424 |
"results": results,
|
| 425 |
-
"
|
| 426 |
-
|
| 427 |
-
|
| 428 |
-
|
| 429 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 430 |
}
|
| 431 |
|
| 432 |
except Exception as e:
|
| 433 |
-
logger.error(f"❌ Erreur
|
| 434 |
gc.collect()
|
| 435 |
raise HTTPException(
|
| 436 |
status_code=500,
|
| 437 |
-
detail=f"Erreur traitement batch
|
| 438 |
)
|
| 439 |
|
| 440 |
-
|
| 441 |
-
@app.post("/test")
|
| 442 |
async def test_chunking():
|
| 443 |
-
"""
|
| 444 |
if pipeline is None:
|
| 445 |
-
raise HTTPException(status_code=503, detail="Pipeline non initialisé")
|
| 446 |
|
| 447 |
try:
|
| 448 |
-
# Test avec texte simple
|
| 449 |
test_request = ChunkRequest(
|
| 450 |
-
text="
|
| 451 |
-
|
| 452 |
-
|
| 453 |
-
|
| 454 |
-
|
| 455 |
-
|
| 456 |
-
|
| 457 |
-
|
| 458 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 459 |
)
|
| 460 |
|
| 461 |
-
|
| 462 |
-
result =
|
| 463 |
-
|
| 464 |
-
|
| 465 |
-
|
| 466 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 467 |
|
| 468 |
return {
|
| 469 |
-
"test_status": "
|
| 470 |
-
"
|
| 471 |
-
"
|
| 472 |
-
|
| 473 |
-
"
|
| 474 |
-
"
|
| 475 |
-
"
|
| 476 |
-
|
| 477 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 478 |
],
|
| 479 |
-
"version": "
|
| 480 |
-
"
|
| 481 |
}
|
| 482 |
|
| 483 |
except Exception as e:
|
| 484 |
-
logger.error(f"❌
|
| 485 |
raise HTTPException(
|
| 486 |
status_code=500,
|
| 487 |
-
detail=f"Test échoué
|
| 488 |
)
|
| 489 |
|
| 490 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 491 |
@app.exception_handler(404)
|
| 492 |
async def not_found_handler(request: Request, exc):
|
|
|
|
| 493 |
return JSONResponse(
|
| 494 |
status_code=404,
|
| 495 |
content={
|
| 496 |
-
"error": "Endpoint non trouvé",
|
| 497 |
"message": f"L'endpoint {request.url.path} n'existe pas",
|
| 498 |
-
"available_endpoints":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 499 |
"documentation": "/docs",
|
| 500 |
-
"version": "
|
| 501 |
-
"architecture": "LlamaIndex Moderne"
|
| 502 |
}
|
| 503 |
)
|
| 504 |
|
| 505 |
-
#
|
| 506 |
if __name__ == "__main__":
|
| 507 |
import uvicorn
|
| 508 |
|
|
|
|
|
|
|
| 509 |
# Configuration optimisée pour HF Space gratuit
|
| 510 |
uvicorn.run(
|
| 511 |
"app:app",
|
| 512 |
host="0.0.0.0",
|
| 513 |
-
port=7860,
|
| 514 |
-
reload=False, #
|
| 515 |
access_log=False, # Économie ressources
|
| 516 |
log_level="info",
|
| 517 |
workers=1, # HF Space gratuit = 1 worker
|
| 518 |
timeout_keep_alive=30,
|
| 519 |
-
limit_concurrency=10 # Limite connexions simultanées
|
|
|
|
| 520 |
)
|
|
|
|
| 1 |
"""
|
| 2 |
+
app.py v4.0 FINAL - FastAPI pour Chunking Sémantique Intelligent
|
| 3 |
|
| 4 |
+
CORRECTIONS ET AMÉLIORATIONS:
|
| 5 |
+
✅ Import SmartChunkerPipeline (correct)
|
| 6 |
+
✅ Méthodes synchronisées avec chunker_pipeline.py
|
| 7 |
+
✅ Gestion d'erreurs robuste
|
| 8 |
+
✅ Endpoints optimisés pour n8n
|
|
|
|
|
|
|
| 9 |
✅ Variables d'environnement sécurisées
|
| 10 |
+
✅ Monitoring et health checks complets
|
| 11 |
+
✅ Configuration HF Space gratuit optimisée
|
| 12 |
"""
|
| 13 |
|
| 14 |
import os
|
| 15 |
import logging
|
| 16 |
import time
|
| 17 |
+
import asyncio
|
| 18 |
+
import gc
|
| 19 |
+
from pathlib import Path
|
| 20 |
from fastapi import FastAPI, HTTPException, Request
|
| 21 |
from fastapi.middleware.cors import CORSMiddleware
|
| 22 |
from fastapi.responses import JSONResponse
|
| 23 |
from pydantic import BaseModel, Field
|
| 24 |
from typing import List, Dict, Any, Optional
|
|
|
|
| 25 |
from concurrent.futures import ThreadPoolExecutor
|
|
|
|
| 26 |
|
| 27 |
# Configuration logging optimisée
|
| 28 |
logging.basicConfig(
|
| 29 |
level=logging.INFO,
|
| 30 |
+
format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
|
| 31 |
+
handlers=[
|
| 32 |
+
logging.StreamHandler(),
|
| 33 |
+
logging.FileHandler("/app/logs/app.log", mode="a") if os.path.exists("/app/logs") else logging.StreamHandler()
|
| 34 |
+
]
|
| 35 |
)
|
| 36 |
logger = logging.getLogger(__name__)
|
| 37 |
|
| 38 |
+
# ✅ IMPORTS PRINCIPAUX - Vérification de compatibilité
|
| 39 |
try:
|
| 40 |
+
from chunker_pipeline import SmartChunkerPipeline
|
| 41 |
from schemas import ChunkRequest, ChunkResponse, ChunkMetadata
|
| 42 |
+
logger.info("✅ Modules chunking v4.0 importés avec succès")
|
| 43 |
except ImportError as e:
|
| 44 |
+
logger.error(f"❌ ERREUR CRITIQUE - Import modules chunking: {e}")
|
| 45 |
+
logger.error("Vérifiez que les fichiers chunker_pipeline.py et schemas.py existent")
|
| 46 |
+
raise
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
|
| 48 |
+
# ✅ CONFIGURATION ENVIRONNEMENT HF SPACE SÉCURISÉE
|
| 49 |
+
def setup_environment():
|
| 50 |
+
"""Configuration optimisée pour Hugging Face Space gratuit"""
|
| 51 |
+
|
| 52 |
+
# Cache HuggingFace optimisé
|
| 53 |
+
cache_base = "/app/cache"
|
| 54 |
+
os.environ["HF_HOME"] = f"{cache_base}/huggingface"
|
| 55 |
+
os.environ["TRANSFORMERS_CACHE"] = f"{cache_base}/transformers"
|
| 56 |
+
os.environ["HF_HUB_CACHE"] = f"{cache_base}/hub"
|
| 57 |
+
|
| 58 |
+
# Optimisations performance
|
| 59 |
+
os.environ["TOKENIZERS_PARALLELISM"] = "false"
|
| 60 |
+
os.environ["HF_HUB_DISABLE_PROGRESS_BARS"] = "1"
|
| 61 |
+
os.environ["TRANSFORMERS_VERBOSITY"] = "error"
|
| 62 |
+
os.environ["PYTHONUNBUFFERED"] = "1"
|
| 63 |
+
|
| 64 |
+
# Création dossiers cache sécurisés
|
| 65 |
+
cache_dirs = [
|
| 66 |
+
f"{cache_base}/huggingface",
|
| 67 |
+
f"{cache_base}/transformers",
|
| 68 |
+
f"{cache_base}/hub",
|
| 69 |
+
f"{cache_base}/llm",
|
| 70 |
+
f"{cache_base}/embeddings",
|
| 71 |
+
"/app/logs"
|
| 72 |
+
]
|
| 73 |
+
|
| 74 |
+
for cache_dir in cache_dirs:
|
| 75 |
+
try:
|
| 76 |
+
os.makedirs(cache_dir, exist_ok=True)
|
| 77 |
+
os.chmod(cache_dir, 0o755)
|
| 78 |
+
except Exception as e:
|
| 79 |
+
logger.warning(f"⚠️ Impossible de créer {cache_dir}: {e}")
|
| 80 |
+
|
| 81 |
+
logger.info("✅ Environnement HF Space configuré")
|
| 82 |
|
| 83 |
+
# Configuration environnement
|
| 84 |
+
setup_environment()
|
| 85 |
|
| 86 |
+
# ✅ INITIALISATION FASTAPI OPTIMISÉE
|
| 87 |
app = FastAPI(
|
| 88 |
+
title="🧠 Chunking Sémantique Intelligent API",
|
| 89 |
+
description="""
|
| 90 |
+
**API de découpage récursif hiérarchique avec parentalité**
|
| 91 |
+
|
| 92 |
+
🚀 **Fonctionnalités:**
|
| 93 |
+
- Chunking sémantique avec Chonkie + LlamaIndex
|
| 94 |
+
- Relations bidirectionnelles parent/enfant
|
| 95 |
+
- Export Obsidian format [[Titre]], id
|
| 96 |
+
- Base connaissance pour agents IA spécialisés
|
| 97 |
+
- 100% gratuit sur HuggingFace Space
|
| 98 |
+
|
| 99 |
+
🔧 **Optimisé pour n8n et automation**
|
| 100 |
+
""",
|
| 101 |
+
version="4.0.0",
|
| 102 |
docs_url="/docs",
|
| 103 |
+
redoc_url="/redoc",
|
| 104 |
+
openapi_tags=[
|
| 105 |
+
{"name": "chunking", "description": "Endpoints de chunking principal"},
|
| 106 |
+
{"name": "monitoring", "description": "Santé et configuration"},
|
| 107 |
+
{"name": "test", "description": "Tests et validation"}
|
| 108 |
+
]
|
| 109 |
)
|
| 110 |
|
| 111 |
+
# ✅ CORS ÉTENDU POUR N8N ET INTÉGRATIONS
|
| 112 |
app.add_middleware(
|
| 113 |
CORSMiddleware,
|
| 114 |
+
allow_origins=["*"], # Nécessaire pour n8n
|
| 115 |
allow_credentials=True,
|
| 116 |
+
allow_methods=["GET", "POST", "PUT", "DELETE", "OPTIONS"],
|
| 117 |
allow_headers=["*"],
|
| 118 |
+
expose_headers=["*"]
|
| 119 |
)
|
| 120 |
|
| 121 |
+
# ✅ VARIABLES GLOBALES
|
| 122 |
pipeline = None
|
| 123 |
executor = ThreadPoolExecutor(max_workers=1) # HF Space gratuit = 1 worker max
|
| 124 |
+
startup_time = time.time()
|
| 125 |
+
request_count = 0
|
| 126 |
|
| 127 |
+
# ✅ MIDDLEWARE MONITORING ET SÉCURITÉ
|
| 128 |
@app.middleware("http")
|
| 129 |
+
async def monitoring_middleware(request: Request, call_next):
|
| 130 |
+
"""Middleware pour monitoring et gestion erreurs globales"""
|
| 131 |
+
global request_count
|
| 132 |
+
start_time = time.time()
|
| 133 |
+
request_count += 1
|
| 134 |
+
|
| 135 |
+
# Headers sécurité
|
| 136 |
+
response = None
|
| 137 |
try:
|
| 138 |
response = await call_next(request)
|
| 139 |
+
response.headers["X-API-Version"] = "4.0.0"
|
| 140 |
+
response.headers["X-Powered-By"] = "Chunking-Semantic-AI"
|
| 141 |
+
|
| 142 |
+
# Log performance
|
| 143 |
+
process_time = time.time() - start_time
|
| 144 |
+
if process_time > 5.0: # Log requêtes lentes
|
| 145 |
+
logger.warning(f"⚠️ Requête lente: {request.url.path} - {process_time:.2f}s")
|
| 146 |
+
|
| 147 |
return response
|
| 148 |
+
|
| 149 |
except Exception as e:
|
| 150 |
+
logger.error(f"❌ Erreur middleware {request.url.path}: {str(e)}")
|
| 151 |
+
|
| 152 |
+
# Réponse d'erreur structurée
|
| 153 |
return JSONResponse(
|
| 154 |
status_code=500,
|
| 155 |
content={
|
| 156 |
"error": "Erreur interne du serveur",
|
| 157 |
"detail": str(e),
|
| 158 |
+
"path": str(request.url.path),
|
| 159 |
"timestamp": time.time(),
|
| 160 |
+
"request_id": request_count,
|
| 161 |
+
"version": "4.0.0"
|
| 162 |
}
|
| 163 |
)
|
| 164 |
|
| 165 |
+
# ✅ ÉVÉNEMENTS LIFECYCLE
|
| 166 |
@app.on_event("startup")
|
| 167 |
async def startup_event():
|
| 168 |
+
"""Initialisation complète au démarrage"""
|
| 169 |
global pipeline
|
| 170 |
+
|
| 171 |
try:
|
| 172 |
+
logger.info("🚀 === DÉMARRAGE API CHUNKING SÉMANTIQUE v4.0 ===")
|
| 173 |
+
|
| 174 |
+
# Vérification espace disque
|
| 175 |
+
import shutil
|
| 176 |
+
total, used, free = shutil.disk_usage("/app")
|
| 177 |
+
free_gb = free / (1024**3)
|
| 178 |
+
logger.info(f"💾 Espace libre: {free_gb:.1f}GB")
|
| 179 |
+
|
| 180 |
+
if free_gb < 1.0:
|
| 181 |
+
logger.warning("⚠️ Espace disque faible (<1GB)")
|
| 182 |
|
| 183 |
+
# Initialisation pipeline principal
|
| 184 |
+
logger.info("🔧 Initialisation SmartChunkerPipeline...")
|
| 185 |
pipeline = SmartChunkerPipeline()
|
| 186 |
await pipeline.initialize()
|
| 187 |
|
| 188 |
+
# Vérification santé
|
| 189 |
+
health = await pipeline.health_check_v4()
|
| 190 |
+
logger.info(f"🏥 Status santé: {health['status']}")
|
| 191 |
+
|
| 192 |
+
if health['status'] != 'healthy':
|
| 193 |
+
logger.warning(f"⚠️ Pipeline en mode dégradé: {health['status']}")
|
| 194 |
+
|
| 195 |
+
# Configuration système
|
| 196 |
+
config_info = await pipeline.get_config_info_v4()
|
| 197 |
+
logger.info(f"🧠 LLM: {config_info['models']['llm_model']}")
|
| 198 |
+
logger.info(f"🔤 Embedding: {config_info['models']['embedding_model']}")
|
| 199 |
+
logger.info(f"🦛 Chonkie: {'✅' if config_info['models']['chonkie_available'] else '❌'}")
|
| 200 |
+
|
| 201 |
+
# Test rapide de fonctionnement
|
| 202 |
+
test_request = ChunkRequest(
|
| 203 |
+
text="Test d'initialisation du système de chunking.",
|
| 204 |
+
titre="Test Init",
|
| 205 |
+
source_id="init_test"
|
| 206 |
+
)
|
| 207 |
+
|
| 208 |
+
test_result = await pipeline.process_text(test_request)
|
| 209 |
+
logger.info(f"✅ Test init: {test_result.total_chunks} chunks générés")
|
| 210 |
|
| 211 |
+
logger.info("🎉 API Chunking Sémantique v4.0 prête !")
|
| 212 |
|
| 213 |
except Exception as e:
|
| 214 |
+
logger.error(f"❌ ERREUR CRITIQUE lors du démarrage: {e}")
|
| 215 |
+
logger.error("Le service ne pourra pas fonctionner correctement")
|
| 216 |
raise
|
| 217 |
|
| 218 |
@app.on_event("shutdown")
|
| 219 |
async def shutdown_event():
|
| 220 |
+
"""Nettoyage propre à l'arrêt"""
|
| 221 |
global pipeline, executor
|
| 222 |
+
|
| 223 |
try:
|
| 224 |
+
logger.info("🛑 Arrêt du service en cours...")
|
| 225 |
|
| 226 |
+
# Nettoyage pipeline
|
| 227 |
if pipeline:
|
| 228 |
+
await pipeline.cleanup()
|
| 229 |
+
logger.info("✅ Pipeline nettoyé")
|
|
|
|
|
|
|
|
|
|
| 230 |
|
| 231 |
+
# Nettoyage executor
|
| 232 |
if executor:
|
| 233 |
+
executor.shutdown(wait=True, timeout=10)
|
| 234 |
+
logger.info("✅ Executor fermé")
|
| 235 |
|
| 236 |
# Nettoyage mémoire final
|
| 237 |
gc.collect()
|
| 238 |
|
| 239 |
+
# Statistiques finales
|
| 240 |
+
uptime = time.time() - startup_time
|
| 241 |
+
logger.info(f"📊 Statistiques finales:")
|
| 242 |
+
logger.info(f" - Temps de fonctionnement: {uptime:.1f}s")
|
| 243 |
+
logger.info(f" - Requêtes traitées: {request_count}")
|
| 244 |
+
logger.info(f" - Moyenne: {request_count/uptime:.2f} req/s")
|
| 245 |
+
|
| 246 |
+
logger.info("✅ Arrêt propre terminé")
|
| 247 |
|
| 248 |
except Exception as e:
|
| 249 |
+
logger.error(f"⚠️ Erreur lors de l'arrêt: {e}")
|
| 250 |
+
|
| 251 |
+
# ✅ ENDPOINTS PRINCIPAUX
|
| 252 |
|
| 253 |
+
@app.get("/", tags=["monitoring"])
|
| 254 |
async def root():
|
| 255 |
+
"""Page d'accueil avec informations complètes du service"""
|
| 256 |
+
uptime = time.time() - startup_time
|
| 257 |
+
|
| 258 |
return {
|
| 259 |
+
"service": "🧠 Chunking Sémantique Intelligent API",
|
| 260 |
+
"version": "4.0.0",
|
| 261 |
+
"status": "🟢 Opérationnel" if pipeline else "🔴 Non initialisé",
|
| 262 |
+
"uptime_seconds": round(uptime, 1),
|
| 263 |
+
"requests_processed": request_count,
|
| 264 |
+
|
| 265 |
"features": [
|
| 266 |
+
"🧩 Chunking sémantique avec Chonkie",
|
| 267 |
+
"🏗️ Hiérarchie récursive intelligente",
|
| 268 |
+
"🔗 Relations bidirectionnelles parent/enfant",
|
| 269 |
+
"📝 Export Obsidian format [[Titre]], id",
|
| 270 |
+
"🤖 Base connaissance pour agents IA spécialisés",
|
| 271 |
+
"💰 100% gratuit sur HuggingFace Space",
|
| 272 |
+
"🔄 Optimisé pour n8n et automation"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 273 |
],
|
| 274 |
+
|
| 275 |
+
"endpoints": {
|
| 276 |
+
"chunking": [
|
| 277 |
+
"POST /chunk - Chunking principal",
|
| 278 |
+
"POST /chunk-batch - Traitement par lots"
|
| 279 |
+
],
|
| 280 |
+
"monitoring": [
|
| 281 |
+
"GET /health - Vérification santé détaillée",
|
| 282 |
+
"GET /config - Configuration système",
|
| 283 |
+
"GET /stats - Statistiques d'usage"
|
| 284 |
+
],
|
| 285 |
+
"test": [
|
| 286 |
+
"POST /test - Test de validation",
|
| 287 |
+
"GET /ping - Test connectivité simple"
|
| 288 |
+
]
|
| 289 |
+
},
|
| 290 |
+
|
| 291 |
+
"documentation": {
|
| 292 |
+
"interactive": "/docs",
|
| 293 |
+
"redoc": "/redoc"
|
| 294 |
+
},
|
| 295 |
+
|
| 296 |
+
"support": {
|
| 297 |
+
"n8n_compatible": True,
|
| 298 |
+
"max_text_length": "500,000 caractères",
|
| 299 |
+
"max_batch_size": 3,
|
| 300 |
+
"response_format": "JSON structuré"
|
| 301 |
+
}
|
| 302 |
}
|
| 303 |
|
| 304 |
+
@app.get("/health", tags=["monitoring"])
|
| 305 |
async def health_check():
|
| 306 |
+
"""Vérification santé complète et détaillée"""
|
| 307 |
try:
|
| 308 |
if pipeline is None:
|
| 309 |
return {
|
| 310 |
+
"status": "🔴 error",
|
| 311 |
"message": "Pipeline non initialisé",
|
| 312 |
+
"version": "4.0.0",
|
| 313 |
+
"timestamp": time.time(),
|
| 314 |
+
"uptime": time.time() - startup_time,
|
| 315 |
+
"critical": True
|
| 316 |
}
|
| 317 |
|
| 318 |
+
# Health check pipeline
|
| 319 |
+
health_result = await pipeline.health_check_v4()
|
|
|
|
| 320 |
|
| 321 |
+
# Informations mémoire
|
| 322 |
+
memory_info = pipeline.get_memory_usage_v4()
|
| 323 |
+
|
| 324 |
+
# Statistiques système
|
| 325 |
+
import psutil
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 326 |
try:
|
| 327 |
+
cpu_percent = psutil.cpu_percent(interval=1)
|
| 328 |
+
memory_percent = psutil.virtual_memory().percent
|
| 329 |
+
except:
|
| 330 |
+
cpu_percent = 0
|
| 331 |
+
memory_percent = 0
|
| 332 |
+
|
| 333 |
+
# Status coloré
|
| 334 |
+
status_map = {
|
| 335 |
+
"healthy": "🟢 healthy",
|
| 336 |
+
"degraded": "🟡 degraded",
|
| 337 |
+
"unhealthy": "🔴 unhealthy",
|
| 338 |
+
"error": "🔴 error"
|
| 339 |
+
}
|
| 340 |
|
| 341 |
return {
|
| 342 |
**health_result,
|
| 343 |
+
"status": status_map.get(health_result['status'], health_result['status']),
|
| 344 |
"memory_info": memory_info,
|
| 345 |
+
"system_info": {
|
| 346 |
+
"cpu_percent": cpu_percent,
|
| 347 |
+
"memory_percent": memory_percent,
|
| 348 |
+
"uptime": time.time() - startup_time,
|
| 349 |
+
"requests_processed": request_count
|
| 350 |
+
},
|
| 351 |
+
"version": "4.0.0"
|
| 352 |
}
|
| 353 |
|
| 354 |
except Exception as e:
|
| 355 |
+
logger.error(f"❌ Erreur health check: {e}")
|
| 356 |
return {
|
| 357 |
+
"status": "🔴 error",
|
| 358 |
"message": f"Erreur health check: {str(e)}",
|
| 359 |
+
"version": "4.0.0",
|
| 360 |
+
"timestamp": time.time(),
|
| 361 |
+
"critical": True
|
| 362 |
}
|
| 363 |
|
| 364 |
+
@app.get("/config", tags=["monitoring"])
|
| 365 |
async def get_config():
|
| 366 |
+
"""Configuration système détaillée"""
|
| 367 |
try:
|
| 368 |
if pipeline is None:
|
| 369 |
raise HTTPException(status_code=503, detail="Pipeline non initialisé")
|
| 370 |
|
| 371 |
+
config_info = await pipeline.get_config_info_v4()
|
|
|
|
| 372 |
|
| 373 |
+
# Ajout informations runtime
|
| 374 |
+
runtime_info = {
|
| 375 |
+
"python_version": f"{os.sys.version_info.major}.{os.sys.version_info.minor}.{os.sys.version_info.micro}",
|
| 376 |
+
"platform": os.name,
|
| 377 |
+
"workers": 1,
|
| 378 |
+
"max_request_size": "500KB",
|
| 379 |
+
"cache_enabled": True,
|
| 380 |
+
"environment": "HuggingFace Space"
|
| 381 |
+
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 382 |
|
| 383 |
return {
|
| 384 |
**config_info,
|
| 385 |
+
"runtime_info": runtime_info,
|
| 386 |
+
"api_version": "4.0.0",
|
| 387 |
+
"timestamp": time.time()
|
| 388 |
}
|
| 389 |
|
| 390 |
except Exception as e:
|
| 391 |
+
logger.error(f"❌ Erreur récupération config: {e}")
|
| 392 |
raise HTTPException(status_code=500, detail=f"Erreur config: {str(e)}")
|
| 393 |
|
| 394 |
+
@app.get("/stats", tags=["monitoring"])
|
| 395 |
+
async def get_stats():
|
| 396 |
+
"""Statistiques d'usage détaillées"""
|
| 397 |
+
uptime = time.time() - startup_time
|
| 398 |
+
avg_requests_per_minute = (request_count / uptime) * 60 if uptime > 0 else 0
|
| 399 |
+
|
| 400 |
+
return {
|
| 401 |
+
"service_stats": {
|
| 402 |
+
"uptime_seconds": round(uptime, 1),
|
| 403 |
+
"uptime_formatted": f"{int(uptime//3600)}h {int((uptime%3600)//60)}m {int(uptime%60)}s",
|
| 404 |
+
"total_requests": request_count,
|
| 405 |
+
"avg_requests_per_minute": round(avg_requests_per_minute, 2)
|
| 406 |
+
},
|
| 407 |
+
"system_health": {
|
| 408 |
+
"pipeline_initialized": pipeline is not None,
|
| 409 |
+
"memory_usage": pipeline.get_memory_usage_v4() if pipeline else "N/A"
|
| 410 |
+
},
|
| 411 |
+
"version": "4.0.0",
|
| 412 |
+
"timestamp": time.time()
|
| 413 |
+
}
|
| 414 |
+
|
| 415 |
+
@app.post("/chunk", response_model=ChunkResponse, tags=["chunking"])
|
| 416 |
async def chunk_text(request: ChunkRequest):
|
| 417 |
"""
|
| 418 |
+
🧠 ENDPOINT PRINCIPAL - Chunking sémantique intelligent
|
| 419 |
|
| 420 |
+
**Fonctionnalités:**
|
| 421 |
+
- Chunking sémantique avec Chonkie + LlamaIndex
|
| 422 |
+
- Relations hiérarchiques bidirectionnelles
|
|
|
|
| 423 |
- Export Obsidian format [[Titre]], id
|
| 424 |
+
- Base connaissance pour agents IA
|
| 425 |
+
|
| 426 |
+
**Optimisé pour n8n et automation**
|
| 427 |
"""
|
| 428 |
if pipeline is None:
|
| 429 |
+
raise HTTPException(
|
| 430 |
+
status_code=503,
|
| 431 |
+
detail="❌ Pipeline non initialisé - Redémarrez le service"
|
| 432 |
+
)
|
| 433 |
|
| 434 |
start_time = time.time()
|
| 435 |
|
| 436 |
try:
|
| 437 |
+
logger.info(f"📝 Début chunking: {request.titre or 'Sans titre'} ({len(request.text)} chars)")
|
| 438 |
|
| 439 |
# Validation entrées renforcée
|
| 440 |
if not request.text or len(request.text.strip()) < 10:
|
| 441 |
raise HTTPException(
|
| 442 |
status_code=400,
|
| 443 |
+
detail="❌ Le texte doit contenir au moins 10 caractères"
|
| 444 |
)
|
| 445 |
|
| 446 |
+
# Limite HF Space gratuit
|
| 447 |
+
max_length = 500000
|
| 448 |
+
if len(request.text) > max_length:
|
| 449 |
raise HTTPException(
|
| 450 |
status_code=400,
|
| 451 |
+
detail=f"❌ Texte trop long ({len(request.text)} chars). Maximum: {max_length:,} caractères"
|
| 452 |
)
|
| 453 |
|
| 454 |
+
# Traitement principal
|
| 455 |
+
result = await pipeline.process_text(request)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 456 |
|
| 457 |
processing_time = time.time() - start_time
|
| 458 |
+
|
| 459 |
+
# Log succès
|
| 460 |
+
logger.info(
|
| 461 |
+
f"✅ Chunking terminé: {result.total_chunks} chunks, "
|
| 462 |
+
f"{result.total_tokens} tokens en {processing_time:.2f}s"
|
| 463 |
+
)
|
| 464 |
|
| 465 |
return result
|
| 466 |
|
| 467 |
except HTTPException:
|
|
|
|
| 468 |
raise
|
| 469 |
except Exception as e:
|
| 470 |
+
logger.error(f"❌ Erreur chunking: {str(e)}")
|
| 471 |
|
| 472 |
+
# Nettoyage mémoire d'urgence
|
| 473 |
try:
|
| 474 |
+
await pipeline._cleanup_memory_v4()
|
| 475 |
+
gc.collect()
|
|
|
|
|
|
|
| 476 |
except:
|
| 477 |
pass
|
| 478 |
|
|
|
|
|
|
|
| 479 |
raise HTTPException(
|
| 480 |
status_code=500,
|
| 481 |
+
detail=f"❌ Erreur traitement: {str(e)}"
|
| 482 |
)
|
| 483 |
|
| 484 |
+
@app.post("/chunk-batch", tags=["chunking"])
|
| 485 |
async def chunk_batch(requests: List[ChunkRequest]):
|
| 486 |
+
"""
|
| 487 |
+
📦 Traitement par lots optimisé pour HF Space gratuit
|
| 488 |
+
|
| 489 |
+
**Limites:**
|
| 490 |
+
- Maximum 3 textes par lot
|
| 491 |
+
- Traitement séquentiel pour économiser la mémoire
|
| 492 |
+
"""
|
| 493 |
|
| 494 |
# Validation limite batch pour Space gratuit
|
| 495 |
max_batch_size = 3
|
| 496 |
if len(requests) > max_batch_size:
|
| 497 |
raise HTTPException(
|
| 498 |
status_code=400,
|
| 499 |
+
detail=f"❌ Maximum {max_batch_size} textes par lot sur HF Space gratuit"
|
| 500 |
)
|
| 501 |
|
| 502 |
if pipeline is None:
|
| 503 |
+
raise HTTPException(status_code=503, detail="❌ Pipeline non initialisé")
|
| 504 |
|
| 505 |
start_time = time.time()
|
| 506 |
results = []
|
| 507 |
|
| 508 |
try:
|
| 509 |
+
logger.info(f"📦 Début batch: {len(requests)} textes")
|
| 510 |
|
| 511 |
for idx, request in enumerate(requests):
|
| 512 |
try:
|
| 513 |
+
logger.info(f" 📝 Traitement {idx+1}/{len(requests)}: {request.titre or 'Sans titre'}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 514 |
|
| 515 |
+
result = await pipeline.process_text(request)
|
| 516 |
results.append({
|
| 517 |
"success": True,
|
| 518 |
"index": idx,
|
|
|
|
| 520 |
"result": result
|
| 521 |
})
|
| 522 |
|
| 523 |
+
# Nettoyage entre chaque traitement
|
| 524 |
+
if idx < len(requests) - 1: # Pas pour le dernier
|
| 525 |
+
await pipeline._cleanup_memory_v4()
|
| 526 |
+
|
| 527 |
except Exception as e:
|
| 528 |
+
logger.error(f"❌ Erreur batch item {idx}: {e}")
|
| 529 |
results.append({
|
| 530 |
"success": False,
|
| 531 |
"index": idx,
|
| 532 |
+
"source_id": request.source_id or f"item_{idx}",
|
| 533 |
"error": str(e)
|
| 534 |
})
|
| 535 |
|
| 536 |
total_time = time.time() - start_time
|
| 537 |
successful_results = [r for r in results if r["success"]]
|
| 538 |
|
| 539 |
+
# Nettoyage final
|
| 540 |
try:
|
| 541 |
+
await pipeline._cleanup_memory_v4()
|
|
|
|
| 542 |
except:
|
| 543 |
pass
|
| 544 |
|
| 545 |
+
logger.info(
|
| 546 |
+
f"✅ Batch terminé: {len(successful_results)}/{len(requests)} succès "
|
| 547 |
+
f"en {total_time:.2f}s"
|
| 548 |
+
)
|
| 549 |
|
| 550 |
return {
|
| 551 |
"results": results,
|
| 552 |
+
"summary": {
|
| 553 |
+
"total_processed": len(requests),
|
| 554 |
+
"successful": len(successful_results),
|
| 555 |
+
"failed": len(requests) - len(successful_results),
|
| 556 |
+
"success_rate": f"{(len(successful_results)/len(requests)*100):.1f}%",
|
| 557 |
+
"total_processing_time": round(total_time, 2),
|
| 558 |
+
"avg_time_per_item": round(total_time / len(requests), 2)
|
| 559 |
+
},
|
| 560 |
+
"version": "4.0.0",
|
| 561 |
+
"timestamp": time.time()
|
| 562 |
}
|
| 563 |
|
| 564 |
except Exception as e:
|
| 565 |
+
logger.error(f"❌ Erreur batch global: {e}")
|
| 566 |
gc.collect()
|
| 567 |
raise HTTPException(
|
| 568 |
status_code=500,
|
| 569 |
+
detail=f"❌ Erreur traitement batch: {str(e)}"
|
| 570 |
)
|
| 571 |
|
| 572 |
+
@app.post("/test", tags=["test"])
|
|
|
|
| 573 |
async def test_chunking():
|
| 574 |
+
"""🧪 Test de validation du déploiement"""
|
| 575 |
if pipeline is None:
|
| 576 |
+
raise HTTPException(status_code=503, detail="❌ Pipeline non initialisé")
|
| 577 |
|
| 578 |
try:
|
|
|
|
| 579 |
test_request = ChunkRequest(
|
| 580 |
+
text="""
|
| 581 |
+
Ceci est un test complet de chunking sémantique intelligent v4.0.
|
| 582 |
+
|
| 583 |
+
Le système utilise Chonkie pour le découpage sémantique avancé.
|
| 584 |
+
Il génère des relations hiérarchiques bidirectionnelles entre les chunks.
|
| 585 |
+
|
| 586 |
+
L'export Obsidian utilise le format [[Titre]], id pour les liens.
|
| 587 |
+
Les agents IA reçoivent une base de connaissance parfaitement structurée.
|
| 588 |
+
|
| 589 |
+
Ce test valide toutes les fonctionnalités principales du système.
|
| 590 |
+
""",
|
| 591 |
+
titre="Test Validation v4.0",
|
| 592 |
+
source_id="validation_test_v4",
|
| 593 |
+
include_metadata=True,
|
| 594 |
+
export_obsidian=True,
|
| 595 |
+
export_agents=True
|
| 596 |
)
|
| 597 |
|
| 598 |
+
start_time = time.time()
|
| 599 |
+
result = await pipeline.process_text(test_request)
|
| 600 |
+
test_time = time.time() - start_time
|
| 601 |
+
|
| 602 |
+
# Vérifications détaillées
|
| 603 |
+
checks = {
|
| 604 |
+
"chunking_functional": result.total_chunks > 0,
|
| 605 |
+
"metadata_extracted": len(result.chunks[0].metadata.keywords) > 0 if result.chunks else False,
|
| 606 |
+
"hierarchy_built": len(result.hierarchy) > 0,
|
| 607 |
+
"obsidian_export": result.obsidian_export is not None,
|
| 608 |
+
"agent_knowledge": result.agent_knowledge is not None,
|
| 609 |
+
"processing_time_ok": test_time < 30 # Moins de 30s
|
| 610 |
+
}
|
| 611 |
+
|
| 612 |
+
success_rate = sum(checks.values()) / len(checks) * 100
|
| 613 |
|
| 614 |
return {
|
| 615 |
+
"test_status": "✅ SUCCESS" if success_rate == 100 else "⚠️ PARTIAL",
|
| 616 |
+
"success_rate": f"{success_rate:.1f}%",
|
| 617 |
+
"results": {
|
| 618 |
+
"chunks_generated": result.total_chunks,
|
| 619 |
+
"tokens_processed": result.total_tokens,
|
| 620 |
+
"processing_time": round(test_time, 2),
|
| 621 |
+
"hierarchy_levels": len(result.hierarchy)
|
| 622 |
+
},
|
| 623 |
+
"checks": checks,
|
| 624 |
+
"features_validated": [
|
| 625 |
+
"✅ Chunking sémantique Chonkie" if checks["chunking_functional"] else "❌ Chunking failed",
|
| 626 |
+
"✅ Extraction métadonnées" if checks["metadata_extracted"] else "❌ Metadata failed",
|
| 627 |
+
"✅ Relations hiérarchiques" if checks["hierarchy_built"] else "❌ Hierarchy failed",
|
| 628 |
+
"✅ Export Obsidian" if checks["obsidian_export"] else "❌ Obsidian failed",
|
| 629 |
+
"✅ Base agents IA" if checks["agent_knowledge"] else "❌ Agents failed"
|
| 630 |
],
|
| 631 |
+
"version": "4.0.0",
|
| 632 |
+
"timestamp": time.time()
|
| 633 |
}
|
| 634 |
|
| 635 |
except Exception as e:
|
| 636 |
+
logger.error(f"❌ Test validation échoué: {e}")
|
| 637 |
raise HTTPException(
|
| 638 |
status_code=500,
|
| 639 |
+
detail=f"❌ Test échoué: {str(e)}"
|
| 640 |
)
|
| 641 |
|
| 642 |
+
@app.get("/ping", tags=["test"])
|
| 643 |
+
async def ping():
|
| 644 |
+
"""🏓 Test de connectivité simple"""
|
| 645 |
+
return {
|
| 646 |
+
"ping": "pong",
|
| 647 |
+
"timestamp": time.time(),
|
| 648 |
+
"version": "4.0.0",
|
| 649 |
+
"status": "🟢 Opérationnel" if pipeline else "🔴 Non initialisé"
|
| 650 |
+
}
|
| 651 |
+
|
| 652 |
+
# ✅ GESTION D'ERREURS PERSONNALISÉE
|
| 653 |
+
|
| 654 |
@app.exception_handler(404)
|
| 655 |
async def not_found_handler(request: Request, exc):
|
| 656 |
+
"""Gestionnaire 404 personnalisé"""
|
| 657 |
return JSONResponse(
|
| 658 |
status_code=404,
|
| 659 |
content={
|
| 660 |
+
"error": "❌ Endpoint non trouvé",
|
| 661 |
"message": f"L'endpoint {request.url.path} n'existe pas",
|
| 662 |
+
"available_endpoints": {
|
| 663 |
+
"chunking": ["/chunk", "/chunk-batch"],
|
| 664 |
+
"monitoring": ["/health", "/config", "/stats"],
|
| 665 |
+
"test": ["/test", "/ping"],
|
| 666 |
+
"docs": ["/docs", "/redoc"]
|
| 667 |
+
},
|
| 668 |
+
"suggestion": "Consultez /docs pour la documentation complète",
|
| 669 |
+
"version": "4.0.0"
|
| 670 |
+
}
|
| 671 |
+
)
|
| 672 |
+
|
| 673 |
+
@app.exception_handler(422)
|
| 674 |
+
async def validation_exception_handler(request: Request, exc):
|
| 675 |
+
"""Gestionnaire erreurs de validation Pydantic"""
|
| 676 |
+
return JSONResponse(
|
| 677 |
+
status_code=422,
|
| 678 |
+
content={
|
| 679 |
+
"error": "❌ Erreur de validation",
|
| 680 |
+
"message": "Les données envoyées ne respectent pas le format attendu",
|
| 681 |
+
"detail": str(exc),
|
| 682 |
+
"hint": "Vérifiez la structure de votre requête JSON",
|
| 683 |
"documentation": "/docs",
|
| 684 |
+
"version": "4.0.0"
|
|
|
|
| 685 |
}
|
| 686 |
)
|
| 687 |
|
| 688 |
+
# ✅ POINT D'ENTRÉE PRINCIPAL
|
| 689 |
if __name__ == "__main__":
|
| 690 |
import uvicorn
|
| 691 |
|
| 692 |
+
logger.info("🚀 Démarrage direct du serveur...")
|
| 693 |
+
|
| 694 |
# Configuration optimisée pour HF Space gratuit
|
| 695 |
uvicorn.run(
|
| 696 |
"app:app",
|
| 697 |
host="0.0.0.0",
|
| 698 |
+
port=7860, # Port standard HF Space
|
| 699 |
+
reload=False, # Mode production
|
| 700 |
access_log=False, # Économie ressources
|
| 701 |
log_level="info",
|
| 702 |
workers=1, # HF Space gratuit = 1 worker
|
| 703 |
timeout_keep_alive=30,
|
| 704 |
+
limit_concurrency=10, # Limite connexions simultanées
|
| 705 |
+
timeout_graceful_shutdown=30
|
| 706 |
)
|