исправлена работа gradio app зависания, установлен AdminVector для доступа к ChromaDB, исправлена ошибка размера векторов из-за которой лицо не регестрировалось в vector db

docker-compose.yml

@@ -17,7 +17,7 @@ services:
       - face_recognition_net
 
   chromadb:
-    image: chromadb/chroma:latest
+    image: chromadb/chroma:0.4.15
     container_name: face_recognition_chroma
     restart: always
     ports:
@@ -38,7 +38,7 @@ services:
     container_name: face_recognition_app
     restart: always
     ports:
-      - "7860:7860"
+      - "8080:8080"
     volumes:
       - ./data:/app/data
       - ./models:/app/models
@@ -62,9 +62,51 @@ services:
               count: 1
               capabilities: [gpu]
 
+  postgres:
+    image: postgres:15
+    container_name: vectoradmin_postgres
+    restart: unless-stopped
+    environment:
+      POSTGRES_USER: vectoradmin
+      POSTGRES_PASSWORD: "your-db-password"
+      POSTGRES_DB: vdbms
+    volumes:
+      - pgdata:/var/lib/postgresql/data
+    networks:
+      - face_recognition_net
+
+  vector-admin:
+    image: mintplexlabs/vectoradmin:latest
+    container_name: vectoradmin_ui
+    restart: unless-stopped
+    ports:
+      - "3001:3001"
+    depends_on:
+      - postgres
+      - chromadb
+    environment:
+      # Порт, на котором слушает UI
+      SERVER_PORT: 3001
+      # Секреты для JWT и Inngest
+      JWT_SECRET: "your-jwt-secret"
+      INNGEST_EVENT_KEY: "background_workers"
+      INNGEST_SIGNING_KEY: "your-inngest-signing-key"
+      INNGEST_LANDING_PAGE: "true"
+      # База данных для метаданных VectorAdmin
+      DATABASE_CONNECTION_STRING: "postgresql://vectoradmin:your-db-password@postgres:5432/vdbms"
+      # Подключение к ChromaDB (основное хранилище векторов)
+      VECTORS_PROVIDER_URL: "http://chromadb:8000"
+      VECTORS_PROVIDER_TYPE: "chroma"
+      # Автоматическая настройка администратора
+      SYS_EMAIL: "admin@vectoradmin.com"
+      SYS_PASSWORD: "admin123"
+    networks:
+      - face_recognition_net
+
 volumes:
   mongo_data:
   chroma_data:
+  pgdata:
 
 networks:
   face_recognition_net:

main.py

@@ -1,4 +1,9 @@
 from pathlib import Path
+import warnings
+
+# Suppress numpy deprecation warnings from InsightFace
+warnings.filterwarnings("ignore", message=".*rcond parameter will change.*")
+warnings.filterwarnings("ignore", category=FutureWarning, module="insightface")
 import gradio as gr
 import numpy as np
 from typing import Optional, Tuple, List
@@ -50,6 +55,19 @@ async def initialize_app():
         identify_use_case=identify_use_case
     )
 
+async def main():
+    app = await initialize_app()
+    app.launch(
+        server_name="0.0.0.0", 
+        server_port=8080,
+        share=False,
+        debug=True,
+        # enable_queue=False,
+        show_error=True,
+        quiet=False,
+        prevent_thread_lock=False,
+        max_threads=40
+    )
+
 if __name__ == "__main__":
-    app = asyncio.run(initialize_app())
-    app.launch(server_name="0.0.0.0", server_port=7860)
+    asyncio.run(main())

src/infrastructure/config.py

@@ -11,7 +11,7 @@ class Settings(BaseSettings):
     
     QUALITY_THRESHOLD: float = 0.7
     VERIFICATION_THRESHOLD: float = 0.8
-    IDENTIFICATION_THRESHOLD: float = 0.75
+    IDENTIFICATION_THRESHOLD: float = 0.65
     
     MAX_FACES_PER_USER: int = 10
     

src/infrastructure/ml/detectors.py

@@ -11,7 +11,24 @@ class RetinaFaceDetector:
     def __init__(self):
         detector_logger.info("Initializing RetinaFace detector")
         try:
-            self.app = FaceAnalysis(providers=['CUDAExecutionProvider', 'CPUExecutionProvider'])
+            # Import onnxruntime to check available providers
+            import onnxruntime as ort
+            
+            # Check available providers and configure accordingly
+            available_providers = ort.get_available_providers()
+            providers = []
+            
+            if 'CUDAExecutionProvider' in available_providers:
+                providers.append('CUDAExecutionProvider')
+                detector_logger.info("CUDA provider available for face detection")
+            elif 'CPUExecutionProvider' in available_providers:
+                providers.append('CPUExecutionProvider')
+                detector_logger.info("Using CPU provider for face detection")
+            else:
+                providers = available_providers
+                detector_logger.warning(f"Using available providers: {available_providers}")
+            
+            self.app = FaceAnalysis(providers=providers)
             self.app.prepare(ctx_id=0, det_size=(640, 640))
             detector_logger.info("RetinaFace detector initialized successfully", 
                                providers=self.app.models.keys() if hasattr(self.app, 'models') else "unknown")

src/infrastructure/ml/recognizers.py

@@ -13,9 +13,24 @@ class SFaceRecognizer:
         recognizer_logger.info("Initializing SFace recognizer")
         try:
             self.model_path = self._download_model()
+            
+            # Check available providers and configure accordingly
+            available_providers = ort.get_available_providers()
+            providers = []
+            
+            if 'CUDAExecutionProvider' in available_providers:
+                providers.append('CUDAExecutionProvider')
+                recognizer_logger.info("CUDA provider available for inference")
+            elif 'CPUExecutionProvider' in available_providers:
+                providers.append('CPUExecutionProvider')
+                recognizer_logger.info("Using CPU provider for inference")
+            else:
+                providers = available_providers
+                recognizer_logger.warning(f"Using available providers: {available_providers}")
+            
             self.session = ort.InferenceSession(
                 self.model_path,
-                providers=['CUDAExecutionProvider', 'CPUExecutionProvider']
+                providers=providers
             )
             self.input_name = self.session.get_inputs()[0].name
             self.output_name = self.session.get_outputs()[0].name

src/infrastructure/storage/chromadb.py

@@ -23,12 +23,17 @@ class ChromaDBVectorStore(VectorStore):
         max_retries = 30
         for i in range(max_retries):
             try:
-                response = requests.get(f"http://{host}:{port}/api/v1/heartbeat")
-                if response.status_code == 200:
+                # Try v2 API first, then fallback to basic connection test
+                try:
+                    response = requests.get(f"http://{host}:{port}/api/v1/version")
+                except:
+                    response = requests.get(f"http://{host}:{port}")
+                
+                if response.status_code in [200, 404]:  # 404 is OK, means server is responding
                     vector_store_logger.info("ChromaDB is ready!")
                     break
             except Exception as e:
-                vector_store_logger.debug(f"ChromaDB heartbeat check failed (attempt {i+1})", error=e)
+                vector_store_logger.debug(f"ChromaDB connection check failed (attempt {i+1})", error=e)
                 pass
             if i == max_retries - 1:
                 vector_store_logger.warning("ChromaDB not responding, attempting to connect anyway...")
@@ -41,7 +46,10 @@ class ChromaDBVectorStore(VectorStore):
         try:
             vector_store_logger.debug("Trying direct ChromaDB connection...")
             self.client = chromadb.HttpClient(host=host, port=port)
-            self.collection = self.client.get_or_create_collection(name="face_embeddings")
+            self.collection = self.client.get_or_create_collection(
+                name="face_embeddings",
+                metadata={"hnsw:space": "cosine"}
+            )
             vector_store_logger.info("Direct ChromaDB connection successful!")
             connection_successful = True
         except Exception as e:
@@ -53,7 +61,10 @@ class ChromaDBVectorStore(VectorStore):
                 vector_store_logger.debug("Trying ChromaDB connection with settings...")
                 settings = chromadb.config.Settings(allow_reset=True)
                 self.client = chromadb.HttpClient(host=host, port=port, settings=settings)
-                self.collection = self.client.get_or_create_collection(name="face_embeddings")
+                self.collection = self.client.get_or_create_collection(
+                name="face_embeddings",
+                metadata={"hnsw:space": "cosine"}
+            )
                 vector_store_logger.info("ChromaDB connection with settings successful!")
                 connection_successful = True
             except Exception as e:
@@ -65,7 +76,10 @@ class ChromaDBVectorStore(VectorStore):
                 vector_store_logger.debug("Trying persistent client fallback...")
                 import tempfile
                 self.client = chromadb.PersistentClient(path=tempfile.mkdtemp())
-                self.collection = self.client.get_or_create_collection(name="face_embeddings")
+                self.collection = self.client.get_or_create_collection(
+                name="face_embeddings",
+                metadata={"hnsw:space": "cosine"}
+            )
                 vector_store_logger.warning("Using persistent client fallback!")
                 connection_successful = True
             except Exception as e:
@@ -146,6 +160,8 @@ class ChromaDBVectorStore(VectorStore):
                     face_id = results['ids'][0][i]
                     distance = results['distances'][0][i]
                     metadata = results['metadatas'][0][i]
+                    # For cosine distance, similarity = 1 - distance
+                    # Cosine distance ranges from 0 (identical) to 2 (opposite)
                     similarity = 1 - distance
                     similar_faces.append((face_id, similarity, metadata))