Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,16 +1,16 @@
|
|
| 1 |
"""
|
| 2 |
-
app.py
|
| 3 |
|
| 4 |
-
|
| 5 |
-
✅
|
| 6 |
-
✅
|
| 7 |
-
✅
|
| 8 |
-
✅
|
|
|
|
| 9 |
✅ Optimisations HF Space gratuit renforcées
|
| 10 |
✅ Variables d'environnement sécurisées
|
| 11 |
"""
|
| 12 |
|
| 13 |
-
|
| 14 |
import os
|
| 15 |
import logging
|
| 16 |
import time
|
|
@@ -30,14 +30,21 @@ logging.basicConfig(
|
|
| 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 de chunking
|
| 38 |
except ImportError as e:
|
| 39 |
logger.error(f"❌ Erreur import modules chunking: {e}")
|
| 40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
|
| 42 |
# ✅ Configuration sécurisée variables d'environnement HF Space
|
| 43 |
os.environ["TOKENIZERS_PARALLELISM"] = "false"
|
|
@@ -46,11 +53,14 @@ os.environ["TRANSFORMERS_VERBOSITY"] = "error"
|
|
| 46 |
os.environ["HF_HOME"] = "/app/cache/huggingface"
|
| 47 |
os.environ["TRANSFORMERS_CACHE"] = "/app/cache/transformers"
|
| 48 |
|
| 49 |
-
#
|
|
|
|
|
|
|
|
|
|
| 50 |
app = FastAPI(
|
| 51 |
-
title="Chunking Sémantique Intelligent API
|
| 52 |
-
description="API de découpage récursif hiérarchique avec parentalité - Powered by
|
| 53 |
-
version="
|
| 54 |
docs_url="/docs",
|
| 55 |
redoc_url="/redoc"
|
| 56 |
)
|
|
@@ -64,7 +74,7 @@ app.add_middleware(
|
|
| 64 |
allow_headers=["*"],
|
| 65 |
)
|
| 66 |
|
| 67 |
-
# ✅ Instance globale pipeline
|
| 68 |
pipeline = None
|
| 69 |
executor = ThreadPoolExecutor(max_workers=1) # HF Space gratuit = 1 worker max
|
| 70 |
|
|
@@ -82,44 +92,59 @@ async def catch_exceptions_middleware(request: Request, call_next):
|
|
| 82 |
content={
|
| 83 |
"error": "Erreur interne du serveur",
|
| 84 |
"detail": str(e),
|
| 85 |
-
"timestamp": time.time()
|
|
|
|
| 86 |
}
|
| 87 |
)
|
| 88 |
|
| 89 |
@app.on_event("startup")
|
| 90 |
async def startup_event():
|
| 91 |
-
"""✅ Initialisation pipeline
|
| 92 |
global pipeline
|
| 93 |
try:
|
| 94 |
-
logger.info("🚀 Initialisation SmartChunkerPipeline
|
| 95 |
|
| 96 |
-
# ✅
|
| 97 |
pipeline = SmartChunkerPipeline()
|
| 98 |
await pipeline.initialize()
|
| 99 |
|
| 100 |
-
# Test santé initial
|
| 101 |
-
|
| 102 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 103 |
|
| 104 |
-
|
| 105 |
-
config_info = await pipeline.get_config_info_v4()
|
| 106 |
-
logger.info(f"⚙️ LLM: {config_info['models']['llm_model']}")
|
| 107 |
-
logger.info(f"🧬 Embedding: {config_info['models']['embedding_model']}")
|
| 108 |
-
logger.info(f"🦛 Chonkie: {config_info['models']['chonkie_available']}")
|
| 109 |
|
| 110 |
except Exception as e:
|
| 111 |
-
logger.error(f"❌ Erreur initialisation pipeline
|
| 112 |
raise
|
| 113 |
|
| 114 |
@app.on_event("shutdown")
|
| 115 |
async def shutdown_event():
|
| 116 |
-
"""✅ Nettoyage à l'arrêt optimisé
|
| 117 |
global pipeline, executor
|
| 118 |
try:
|
| 119 |
logger.info("🛑 Arrêt du service - nettoyage en cours...")
|
| 120 |
|
| 121 |
if pipeline:
|
| 122 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 123 |
|
| 124 |
if executor:
|
| 125 |
executor.shutdown(wait=True)
|
|
@@ -127,20 +152,22 @@ async def shutdown_event():
|
|
| 127 |
# Nettoyage mémoire final
|
| 128 |
gc.collect()
|
| 129 |
|
| 130 |
-
logger.info("✅ Nettoyage terminé")
|
| 131 |
|
| 132 |
except Exception as e:
|
| 133 |
logger.error(f"⚠️ Erreur lors du nettoyage: {e}")
|
| 134 |
|
| 135 |
@app.get("/")
|
| 136 |
async def root():
|
| 137 |
-
"""Point d'entrée racine avec informations service
|
| 138 |
return {
|
| 139 |
"service": "Chunking Sémantique Intelligent API",
|
| 140 |
-
"version": "
|
| 141 |
"status": "running",
|
|
|
|
| 142 |
"features": [
|
| 143 |
"Chunking sémantique avec Chonkie",
|
|
|
|
| 144 |
"Hiérarchie récursive intelligente",
|
| 145 |
"Relations bidirectionnelles",
|
| 146 |
"Export Obsidian [[Titre]], id",
|
|
@@ -152,68 +179,119 @@ async def root():
|
|
| 152 |
"GET /health - Vérification santé",
|
| 153 |
"GET /config - Configuration système",
|
| 154 |
"POST /chunk - Chunking principal",
|
| 155 |
-
"POST /chunk-batch - Chunking par lots"
|
|
|
|
| 156 |
],
|
| 157 |
-
"documentation": "/docs"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 158 |
}
|
| 159 |
|
| 160 |
@app.get("/health")
|
| 161 |
async def health_check():
|
| 162 |
-
"""✅ Vérification santé complète
|
| 163 |
try:
|
| 164 |
if pipeline is None:
|
| 165 |
return {
|
| 166 |
"status": "error",
|
| 167 |
"message": "Pipeline non initialisé",
|
| 168 |
-
"version": "
|
| 169 |
"timestamp": time.time()
|
| 170 |
}
|
| 171 |
|
| 172 |
-
# ✅
|
| 173 |
-
health_result =
|
|
|
|
| 174 |
|
| 175 |
-
#
|
| 176 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 177 |
|
| 178 |
return {
|
| 179 |
**health_result,
|
| 180 |
"memory_info": memory_info,
|
| 181 |
-
"version": "
|
|
|
|
| 182 |
}
|
| 183 |
|
| 184 |
except Exception as e:
|
| 185 |
-
logger.error(f"❌ Erreur health check: {e}")
|
| 186 |
return {
|
| 187 |
"status": "error",
|
| 188 |
"message": f"Erreur health check: {str(e)}",
|
| 189 |
-
"version": "
|
| 190 |
"timestamp": time.time()
|
| 191 |
}
|
| 192 |
|
| 193 |
@app.get("/config")
|
| 194 |
async def get_config():
|
| 195 |
-
"""✅
|
| 196 |
try:
|
| 197 |
if pipeline is None:
|
| 198 |
raise HTTPException(status_code=503, detail="Pipeline non initialisé")
|
| 199 |
|
| 200 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 201 |
return {
|
| 202 |
**config_info,
|
| 203 |
-
"api_version": "
|
| 204 |
-
"timestamp": time.time()
|
|
|
|
| 205 |
}
|
| 206 |
|
| 207 |
except Exception as e:
|
| 208 |
-
logger.error(f"❌ Erreur récupération config: {e}")
|
| 209 |
raise HTTPException(status_code=500, detail=f"Erreur config: {str(e)}")
|
| 210 |
|
| 211 |
@app.post("/chunk", response_model=ChunkResponse)
|
| 212 |
async def chunk_text(request: ChunkRequest):
|
| 213 |
"""
|
| 214 |
-
✅ Point d'entrée principal chunking sémantique intelligent
|
| 215 |
|
| 216 |
-
Traitement récursif hiérarchique avec:
|
|
|
|
| 217 |
- Chonkie SemanticChunker + RecursiveChunker
|
| 218 |
- Relations bidirectionnelles complètes
|
| 219 |
- Export Obsidian format [[Titre]], id
|
|
@@ -225,7 +303,7 @@ async def chunk_text(request: ChunkRequest):
|
|
| 225 |
start_time = time.time()
|
| 226 |
|
| 227 |
try:
|
| 228 |
-
logger.info(f"📝 Début chunking
|
| 229 |
|
| 230 |
# Validation entrées renforcée
|
| 231 |
if not request.text or len(request.text.strip()) < 10:
|
|
@@ -240,11 +318,24 @@ async def chunk_text(request: ChunkRequest):
|
|
| 240 |
detail="Texte trop long (max 500,000 caractères pour Space gratuit)"
|
| 241 |
)
|
| 242 |
|
| 243 |
-
# ✅
|
| 244 |
-
result =
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 245 |
|
| 246 |
processing_time = time.time() - start_time
|
| 247 |
-
logger.info(f"✅ Chunking
|
| 248 |
|
| 249 |
return result
|
| 250 |
|
|
@@ -252,11 +343,14 @@ async def chunk_text(request: ChunkRequest):
|
|
| 252 |
# Re-lever les HTTPException sans modification
|
| 253 |
raise
|
| 254 |
except Exception as e:
|
| 255 |
-
logger.error(f"❌ Erreur chunking
|
| 256 |
|
| 257 |
-
# Nettoyage mémoire en cas d'erreur
|
| 258 |
try:
|
| 259 |
-
|
|
|
|
|
|
|
|
|
|
| 260 |
except:
|
| 261 |
pass
|
| 262 |
|
|
@@ -264,12 +358,12 @@ async def chunk_text(request: ChunkRequest):
|
|
| 264 |
|
| 265 |
raise HTTPException(
|
| 266 |
status_code=500,
|
| 267 |
-
detail=f"Erreur traitement chunking
|
| 268 |
)
|
| 269 |
|
| 270 |
@app.post("/chunk-batch")
|
| 271 |
async def chunk_batch(requests: List[ChunkRequest]):
|
| 272 |
-
"""✅ Traitement par lots optimisé
|
| 273 |
|
| 274 |
# Validation limite batch pour Space gratuit
|
| 275 |
max_batch_size = 3
|
|
@@ -286,12 +380,18 @@ async def chunk_batch(requests: List[ChunkRequest]):
|
|
| 286 |
results = []
|
| 287 |
|
| 288 |
try:
|
| 289 |
-
logger.info(f"📦 Début chunking batch
|
| 290 |
|
| 291 |
for idx, request in enumerate(requests):
|
| 292 |
try:
|
| 293 |
-
# Traitement individuel avec
|
| 294 |
-
result =
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 295 |
results.append({
|
| 296 |
"success": True,
|
| 297 |
"index": idx,
|
|
@@ -313,11 +413,12 @@ async def chunk_batch(requests: List[ChunkRequest]):
|
|
| 313 |
|
| 314 |
# Nettoyage mémoire après batch
|
| 315 |
try:
|
| 316 |
-
|
|
|
|
| 317 |
except:
|
| 318 |
pass
|
| 319 |
|
| 320 |
-
logger.info(f"✅ Batch
|
| 321 |
|
| 322 |
return {
|
| 323 |
"results": results,
|
|
@@ -325,57 +426,65 @@ async def chunk_batch(requests: List[ChunkRequest]):
|
|
| 325 |
"successful": len(successful_results),
|
| 326 |
"failed": len(requests) - len(successful_results),
|
| 327 |
"total_processing_time": total_time,
|
| 328 |
-
"version": "
|
| 329 |
}
|
| 330 |
|
| 331 |
except Exception as e:
|
| 332 |
-
logger.error(f"❌ Erreur chunking batch
|
| 333 |
gc.collect()
|
| 334 |
raise HTTPException(
|
| 335 |
status_code=500,
|
| 336 |
-
detail=f"Erreur traitement batch
|
| 337 |
)
|
| 338 |
|
| 339 |
-
# ✅
|
| 340 |
@app.post("/test")
|
| 341 |
async def test_chunking():
|
| 342 |
-
"""Endpoint de test pour validation déploiement
|
| 343 |
if pipeline is None:
|
| 344 |
raise HTTPException(status_code=503, detail="Pipeline non initialisé")
|
| 345 |
|
| 346 |
try:
|
| 347 |
# Test avec texte simple
|
| 348 |
test_request = ChunkRequest(
|
| 349 |
-
text="Ceci est un test de chunking sémantique intelligent
|
| 350 |
-
"Le système utilise
|
|
|
|
| 351 |
"Il génère des relations hiérarchiques bidirectionnelles. "
|
| 352 |
"L'export Obsidian utilise le format [[Titre]], id. "
|
| 353 |
"Les agents IA reçoivent une base de connaissance structurée.",
|
| 354 |
-
titre="Test Chunking
|
| 355 |
-
source_id="
|
| 356 |
include_metadata=True
|
| 357 |
)
|
| 358 |
|
| 359 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 360 |
|
| 361 |
return {
|
| 362 |
"test_status": "success",
|
| 363 |
"chunks_generated": result.total_chunks,
|
| 364 |
"processing_time": result.processing_time,
|
| 365 |
"features_tested": [
|
|
|
|
| 366 |
"Chunking sémantique Chonkie",
|
| 367 |
"Relations hiérarchiques",
|
| 368 |
"Export Obsidian",
|
| 369 |
"Base connaissance agents"
|
| 370 |
],
|
| 371 |
-
"version": "
|
|
|
|
| 372 |
}
|
| 373 |
|
| 374 |
except Exception as e:
|
| 375 |
-
logger.error(f"❌ Erreur test chunking
|
| 376 |
raise HTTPException(
|
| 377 |
status_code=500,
|
| 378 |
-
detail=f"Test échoué
|
| 379 |
)
|
| 380 |
|
| 381 |
# ✅ Gestion erreur 404 personnalisée
|
|
@@ -388,7 +497,8 @@ async def not_found_handler(request: Request, exc):
|
|
| 388 |
"message": f"L'endpoint {request.url.path} n'existe pas",
|
| 389 |
"available_endpoints": ["/", "/health", "/config", "/chunk", "/chunk-batch", "/test"],
|
| 390 |
"documentation": "/docs",
|
| 391 |
-
"version": "
|
|
|
|
| 392 |
}
|
| 393 |
)
|
| 394 |
|
|
|
|
| 1 |
"""
|
| 2 |
+
app.py v5.0 ADAPTÉ - FastAPI pour Chunking Sémantique Intelligent
|
| 3 |
|
| 4 |
+
ADAPTATIONS MAJEURES v5.0:
|
| 5 |
+
✅ Compatible avec LlamaIndex moderne modulaire (llama-index-core)
|
| 6 |
+
✅ Import SmartChunkerPipeline adapté pour nouvelle architecture
|
| 7 |
+
✅ Méthodes corrigées pour nouvelle structure
|
| 8 |
+
✅ Health check v5.0 avec support modulaire
|
| 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
|
|
|
|
| 30 |
)
|
| 31 |
logger = logging.getLogger(__name__)
|
| 32 |
|
| 33 |
+
# ✅ ADAPTATION: Import SmartChunkerPipeline v5.0 avec fallback
|
| 34 |
try:
|
| 35 |
+
from chunker_pipeline import SmartChunkerPipeline # Version moderne
|
| 36 |
from schemas import ChunkRequest, ChunkResponse, ChunkMetadata
|
| 37 |
+
logger.info("✅ Modules de chunking v5.0 importés avec succès (LlamaIndex moderne)")
|
| 38 |
except ImportError as e:
|
| 39 |
logger.error(f"❌ Erreur import modules chunking: {e}")
|
| 40 |
+
# Fallback pour tests
|
| 41 |
+
try:
|
| 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 |
# ✅ Configuration sécurisée variables d'environnement HF Space
|
| 50 |
os.environ["TOKENIZERS_PARALLELISM"] = "false"
|
|
|
|
| 53 |
os.environ["HF_HOME"] = "/app/cache/huggingface"
|
| 54 |
os.environ["TRANSFORMERS_CACHE"] = "/app/cache/transformers"
|
| 55 |
|
| 56 |
+
# ✅ NOUVEAU: Configuration spécifique LlamaIndex moderne
|
| 57 |
+
os.environ["LLAMA_INDEX_CACHE_DIR"] = "/app/cache/llamaindex"
|
| 58 |
+
|
| 59 |
+
# Initialisation FastAPI avec optimisations v5.0
|
| 60 |
app = FastAPI(
|
| 61 |
+
title="Chunking Sémantique Intelligent API v5.0",
|
| 62 |
+
description="API de découpage récursif hiérarchique avec parentalité - Powered by LlamaIndex Moderne + Chonkie",
|
| 63 |
+
version="5.0.0",
|
| 64 |
docs_url="/docs",
|
| 65 |
redoc_url="/redoc"
|
| 66 |
)
|
|
|
|
| 74 |
allow_headers=["*"],
|
| 75 |
)
|
| 76 |
|
| 77 |
+
# ✅ Instance globale pipeline v5.0 (chargée une seule fois)
|
| 78 |
pipeline = None
|
| 79 |
executor = ThreadPoolExecutor(max_workers=1) # HF Space gratuit = 1 worker max
|
| 80 |
|
|
|
|
| 92 |
content={
|
| 93 |
"error": "Erreur interne du serveur",
|
| 94 |
"detail": str(e),
|
| 95 |
+
"timestamp": time.time(),
|
| 96 |
+
"version": "5.0.0"
|
| 97 |
}
|
| 98 |
)
|
| 99 |
|
| 100 |
@app.on_event("startup")
|
| 101 |
async def startup_event():
|
| 102 |
+
"""✅ Initialisation pipeline v5.0 au démarrage avec LlamaIndex moderne"""
|
| 103 |
global pipeline
|
| 104 |
try:
|
| 105 |
+
logger.info("🚀 Initialisation SmartChunkerPipeline v5.0 (LlamaIndex moderne)...")
|
| 106 |
|
| 107 |
+
# ✅ ADAPTATION: SmartChunkerPipeline avec support modulaire
|
| 108 |
pipeline = SmartChunkerPipeline()
|
| 109 |
await pipeline.initialize()
|
| 110 |
|
| 111 |
+
# Test santé initial avec méthodes adaptatives
|
| 112 |
+
try:
|
| 113 |
+
health = await pipeline.health_check_v4()
|
| 114 |
+
logger.info(f"✅ Pipeline v5.0 initialisé - Status: {health['status']}")
|
| 115 |
+
except AttributeError:
|
| 116 |
+
# Fallback pour anciennes méthodes
|
| 117 |
+
health = await pipeline.get_health_status() if hasattr(pipeline, 'get_health_status') else {"status": "unknown"}
|
| 118 |
+
logger.info(f"✅ Pipeline v5.0 initialisé - Status: {health.get('status', 'initialized')}")
|
| 119 |
+
|
| 120 |
+
# Log informations configuration avec fallback
|
| 121 |
+
try:
|
| 122 |
+
config_info = await pipeline.get_config_info_v4()
|
| 123 |
+
logger.info(f"⚙️ LLM: {config_info.get('models', {}).get('llm_model', 'N/A')}")
|
| 124 |
+
logger.info(f"🧬 Embedding: {config_info.get('models', {}).get('embedding_model', 'N/A')}")
|
| 125 |
+
logger.info(f"🦛 Chonkie: {config_info.get('models', {}).get('chonkie_available', False)}")
|
| 126 |
+
except AttributeError:
|
| 127 |
+
logger.info("⚙️ Configuration détaillée non disponible (mode fallback)")
|
| 128 |
|
| 129 |
+
logger.info("🔧 LlamaIndex moderne modulaire configuré")
|
|
|
|
|
|
|
|
|
|
|
|
|
| 130 |
|
| 131 |
except Exception as e:
|
| 132 |
+
logger.error(f"❌ Erreur initialisation pipeline v5.0: {e}")
|
| 133 |
raise
|
| 134 |
|
| 135 |
@app.on_event("shutdown")
|
| 136 |
async def shutdown_event():
|
| 137 |
+
"""✅ Nettoyage à l'arrêt optimisé v5.0"""
|
| 138 |
global pipeline, executor
|
| 139 |
try:
|
| 140 |
logger.info("🛑 Arrêt du service - nettoyage en cours...")
|
| 141 |
|
| 142 |
if pipeline:
|
| 143 |
+
# Nettoyage adaptatif selon les méthodes disponibles
|
| 144 |
+
if hasattr(pipeline, 'cleanup'):
|
| 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)
|
|
|
|
| 152 |
# Nettoyage mémoire final
|
| 153 |
gc.collect()
|
| 154 |
|
| 155 |
+
logger.info("✅ Nettoyage v5.0 terminé")
|
| 156 |
|
| 157 |
except Exception as e:
|
| 158 |
logger.error(f"⚠️ Erreur lors du nettoyage: {e}")
|
| 159 |
|
| 160 |
@app.get("/")
|
| 161 |
async def root():
|
| 162 |
+
"""Point d'entrée racine avec informations service v5.0"""
|
| 163 |
return {
|
| 164 |
"service": "Chunking Sémantique Intelligent API",
|
| 165 |
+
"version": "5.0.0",
|
| 166 |
"status": "running",
|
| 167 |
+
"architecture": "LlamaIndex Moderne Modulaire",
|
| 168 |
"features": [
|
| 169 |
"Chunking sémantique avec Chonkie",
|
| 170 |
+
"LlamaIndex moderne (llama-index-core)",
|
| 171 |
"Hiérarchie récursive intelligente",
|
| 172 |
"Relations bidirectionnelles",
|
| 173 |
"Export Obsidian [[Titre]], id",
|
|
|
|
| 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 |
+
"documentation": "/docs",
|
| 186 |
+
"compatible_with": [
|
| 187 |
+
"llama-index-core (moderne)",
|
| 188 |
+
"llama-index-embeddings-huggingface",
|
| 189 |
+
"chonkie >= 0.1.0"
|
| 190 |
+
]
|
| 191 |
}
|
| 192 |
|
| 193 |
@app.get("/health")
|
| 194 |
async def health_check():
|
| 195 |
+
"""✅ Vérification santé complète v5.0 avec fallbacks adaptatifs"""
|
| 196 |
try:
|
| 197 |
if pipeline is None:
|
| 198 |
return {
|
| 199 |
"status": "error",
|
| 200 |
"message": "Pipeline non initialisé",
|
| 201 |
+
"version": "5.0.0",
|
| 202 |
"timestamp": time.time()
|
| 203 |
}
|
| 204 |
|
| 205 |
+
# ✅ ADAPTATION: Health check avec fallbacks multiples
|
| 206 |
+
health_result = None
|
| 207 |
+
memory_info = None
|
| 208 |
|
| 209 |
+
# Tentative méthode v4.0
|
| 210 |
+
try:
|
| 211 |
+
health_result = await pipeline.health_check_v4()
|
| 212 |
+
except AttributeError:
|
| 213 |
+
# Fallback méthode alternative
|
| 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 |
+
memory_info = pipeline.get_memory_usage_v4()
|
| 227 |
+
except AttributeError:
|
| 228 |
+
try:
|
| 229 |
+
memory_info = pipeline.get_memory_usage()
|
| 230 |
+
except AttributeError:
|
| 231 |
+
memory_info = {"status": "monitoring_unavailable"}
|
| 232 |
|
| 233 |
return {
|
| 234 |
**health_result,
|
| 235 |
"memory_info": memory_info,
|
| 236 |
+
"version": "5.0.0",
|
| 237 |
+
"architecture": "LlamaIndex Moderne"
|
| 238 |
}
|
| 239 |
|
| 240 |
except Exception as e:
|
| 241 |
+
logger.error(f"❌ Erreur health check v5.0: {e}")
|
| 242 |
return {
|
| 243 |
"status": "error",
|
| 244 |
"message": f"Erreur health check: {str(e)}",
|
| 245 |
+
"version": "5.0.0",
|
| 246 |
"timestamp": time.time()
|
| 247 |
}
|
| 248 |
|
| 249 |
@app.get("/config")
|
| 250 |
async def get_config():
|
| 251 |
+
"""✅ Informations configuration système v5.0 avec adaptation modulaire"""
|
| 252 |
try:
|
| 253 |
if pipeline is None:
|
| 254 |
raise HTTPException(status_code=503, detail="Pipeline non initialisé")
|
| 255 |
|
| 256 |
+
# ✅ ADAPTATION: Configuration avec fallbacks
|
| 257 |
+
config_info = {}
|
| 258 |
+
|
| 259 |
+
try:
|
| 260 |
+
config_info = await pipeline.get_config_info_v4()
|
| 261 |
+
except AttributeError:
|
| 262 |
+
# Fallback configuration basique
|
| 263 |
+
config_info = {
|
| 264 |
+
"version": "5.0.0",
|
| 265 |
+
"architecture": "LlamaIndex Moderne Modulaire",
|
| 266 |
+
"models": {
|
| 267 |
+
"llm_available": hasattr(pipeline, 'llm'),
|
| 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 |
+
"api_version": "5.0.0",
|
| 280 |
+
"timestamp": time.time(),
|
| 281 |
+
"llamaindex_architecture": "moderne_modulaire"
|
| 282 |
}
|
| 283 |
|
| 284 |
except Exception as e:
|
| 285 |
+
logger.error(f"❌ Erreur récupération config v5.0: {e}")
|
| 286 |
raise HTTPException(status_code=500, detail=f"Erreur config: {str(e)}")
|
| 287 |
|
| 288 |
@app.post("/chunk", response_model=ChunkResponse)
|
| 289 |
async def chunk_text(request: ChunkRequest):
|
| 290 |
"""
|
| 291 |
+
✅ Point d'entrée principal chunking sémantique intelligent v5.0
|
| 292 |
|
| 293 |
+
Traitement récursif hiérarchique avec LlamaIndex moderne:
|
| 294 |
+
- llama-index-core + llama-index-embeddings-huggingface
|
| 295 |
- Chonkie SemanticChunker + RecursiveChunker
|
| 296 |
- Relations bidirectionnelles complètes
|
| 297 |
- Export Obsidian format [[Titre]], id
|
|
|
|
| 303 |
start_time = time.time()
|
| 304 |
|
| 305 |
try:
|
| 306 |
+
logger.info(f"📝 Début chunking v5.0: {request.titre or 'Sans titre'}")
|
| 307 |
|
| 308 |
# Validation entrées renforcée
|
| 309 |
if not request.text or len(request.text.strip()) < 10:
|
|
|
|
| 318 |
detail="Texte trop long (max 500,000 caractères pour Space gratuit)"
|
| 319 |
)
|
| 320 |
|
| 321 |
+
# ✅ ADAPTATION: Appel méthode avec fallbacks
|
| 322 |
+
result = None
|
| 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 |
+
logger.info(f"✅ Chunking v5.0 terminé: {result.total_chunks} chunks en {processing_time:.2f}s")
|
| 339 |
|
| 340 |
return result
|
| 341 |
|
|
|
|
| 343 |
# Re-lever les HTTPException sans modification
|
| 344 |
raise
|
| 345 |
except Exception as e:
|
| 346 |
+
logger.error(f"❌ Erreur chunking v5.0: {str(e)}")
|
| 347 |
|
| 348 |
+
# Nettoyage mémoire en cas d'erreur avec fallback
|
| 349 |
try:
|
| 350 |
+
if hasattr(pipeline, '_cleanup_memory_v4'):
|
| 351 |
+
await pipeline._cleanup_memory_v4()
|
| 352 |
+
elif hasattr(pipeline, '_cleanup_memory'):
|
| 353 |
+
await pipeline._cleanup_memory()
|
| 354 |
except:
|
| 355 |
pass
|
| 356 |
|
|
|
|
| 358 |
|
| 359 |
raise HTTPException(
|
| 360 |
status_code=500,
|
| 361 |
+
detail=f"Erreur traitement chunking v5.0: {str(e)}"
|
| 362 |
)
|
| 363 |
|
| 364 |
@app.post("/chunk-batch")
|
| 365 |
async def chunk_batch(requests: List[ChunkRequest]):
|
| 366 |
+
"""✅ Traitement par lots optimisé v5.0 (limité HF Space gratuit)"""
|
| 367 |
|
| 368 |
# Validation limite batch pour Space gratuit
|
| 369 |
max_batch_size = 3
|
|
|
|
| 380 |
results = []
|
| 381 |
|
| 382 |
try:
|
| 383 |
+
logger.info(f"📦 Début chunking batch v5.0: {len(requests)} textes")
|
| 384 |
|
| 385 |
for idx, request in enumerate(requests):
|
| 386 |
try:
|
| 387 |
+
# Traitement individuel avec méthodes adaptatives
|
| 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,
|
|
|
|
| 413 |
|
| 414 |
# Nettoyage mémoire après batch
|
| 415 |
try:
|
| 416 |
+
if hasattr(pipeline, '_cleanup_memory_v4'):
|
| 417 |
+
await pipeline._cleanup_memory_v4()
|
| 418 |
except:
|
| 419 |
pass
|
| 420 |
|
| 421 |
+
logger.info(f"✅ Batch v5.0 terminé: {len(successful_results)}/{len(requests)} succès en {total_time:.2f}s")
|
| 422 |
|
| 423 |
return {
|
| 424 |
"results": results,
|
|
|
|
| 426 |
"successful": len(successful_results),
|
| 427 |
"failed": len(requests) - len(successful_results),
|
| 428 |
"total_processing_time": total_time,
|
| 429 |
+
"version": "5.0.0"
|
| 430 |
}
|
| 431 |
|
| 432 |
except Exception as e:
|
| 433 |
+
logger.error(f"❌ Erreur chunking batch v5.0: {e}")
|
| 434 |
gc.collect()
|
| 435 |
raise HTTPException(
|
| 436 |
status_code=500,
|
| 437 |
+
detail=f"Erreur traitement batch v5.0: {str(e)}"
|
| 438 |
)
|
| 439 |
|
| 440 |
+
# ✅ Endpoint test adapté pour v5.0
|
| 441 |
@app.post("/test")
|
| 442 |
async def test_chunking():
|
| 443 |
+
"""Endpoint de test pour validation déploiement v5.0"""
|
| 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="Ceci est un test de chunking sémantique intelligent v5.0. "
|
| 451 |
+
"Le système utilise LlamaIndex moderne avec llama-index-core. "
|
| 452 |
+
"Il intègre Chonkie pour le découpage sémantique avancé. "
|
| 453 |
"Il génère des relations hiérarchiques bidirectionnelles. "
|
| 454 |
"L'export Obsidian utilise le format [[Titre]], id. "
|
| 455 |
"Les agents IA reçoivent une base de connaissance structurée.",
|
| 456 |
+
titre="Test Chunking v5.0",
|
| 457 |
+
source_id="test_v5",
|
| 458 |
include_metadata=True
|
| 459 |
)
|
| 460 |
|
| 461 |
+
# Test avec méthodes adaptatives
|
| 462 |
+
result = None
|
| 463 |
+
try:
|
| 464 |
+
result = await pipeline.process_text(test_request)
|
| 465 |
+
except AttributeError:
|
| 466 |
+
result = await pipeline.process_text_sync(test_request)
|
| 467 |
|
| 468 |
return {
|
| 469 |
"test_status": "success",
|
| 470 |
"chunks_generated": result.total_chunks,
|
| 471 |
"processing_time": result.processing_time,
|
| 472 |
"features_tested": [
|
| 473 |
+
"LlamaIndex moderne (llama-index-core)",
|
| 474 |
"Chunking sémantique Chonkie",
|
| 475 |
"Relations hiérarchiques",
|
| 476 |
"Export Obsidian",
|
| 477 |
"Base connaissance agents"
|
| 478 |
],
|
| 479 |
+
"version": "5.0.0",
|
| 480 |
+
"architecture": "LlamaIndex Moderne Modulaire"
|
| 481 |
}
|
| 482 |
|
| 483 |
except Exception as e:
|
| 484 |
+
logger.error(f"❌ Erreur test chunking v5.0: {e}")
|
| 485 |
raise HTTPException(
|
| 486 |
status_code=500,
|
| 487 |
+
detail=f"Test échoué v5.0: {str(e)}"
|
| 488 |
)
|
| 489 |
|
| 490 |
# ✅ Gestion erreur 404 personnalisée
|
|
|
|
| 497 |
"message": f"L'endpoint {request.url.path} n'existe pas",
|
| 498 |
"available_endpoints": ["/", "/health", "/config", "/chunk", "/chunk-batch", "/test"],
|
| 499 |
"documentation": "/docs",
|
| 500 |
+
"version": "5.0.0",
|
| 501 |
+
"architecture": "LlamaIndex Moderne"
|
| 502 |
}
|
| 503 |
)
|
| 504 |
|