update

Author: tom-sapletta-com

containers/llm-orchestrator-min/docker-compose.yml

@@ -0,0 +1,59 @@
+version: '3.8'
+
+services:
+  # API Gateway (Traefik)
+  api-gateway:
+    build: ./microservices/api-gateway
+    container_name: llm-api-gateway
+    ports:
+      - "80:80"      # API
+      - "8080:8080"  # Dashboard
+    volumes:
+      - /var/run/docker.sock:/var/run/docker.sock:ro
+    networks:
+      - llm-network
+    depends_on:
+      - model-service
+    restart: unless-stopped
+
+  # Model Service
+  model-service:
+    build: ./microservices/model-service
+    container_name: llm-model-service
+    environment:
+      - MODEL_PATH=/app/models/tinyllama
+      - USE_INT8=true
+      - MODEL_SERVICE_PORT=5000
+    volumes:
+      - model-data:/app/models
+    networks:
+      - llm-network
+    restart: unless-stopped
+    deploy:
+      resources:
+        limits:
+          memory: 4G
+
+  # Cache Service (można dodać w przyszłości)
+  # cache-service:
+  #   build: ./microservices/cache-service
+  #   container_name: llm-cache-service
+  #   networks:
+  #     - llm-network
+  #   restart: unless-stopped
+
+  # Monitoring Service (można dodać w przyszłości)
+  # monitoring-service:
+  #   build: ./microservices/monitoring-service
+  #   container_name: llm-monitoring-service
+  #   networks:
+  #     - llm-network
+  #   restart: unless-stopped
+
+networks:
+  llm-network:
+    driver: bridge
+
+volumes:
+  model-data:
+    driver: local

containers/llm-orchestrator-min/microservices/api-gateway/Dockerfile

@@ -0,0 +1,16 @@
+FROM traefik:v2.9
+
+# Kopiowanie konfiguracji
+COPY traefik.yml /etc/traefik/traefik.yml
+COPY dynamic_conf.yml /etc/traefik/dynamic_conf.yml
+
+# Ekspozycja portów
+EXPOSE 80
+EXPOSE 8080
+
+# Healthcheck
+HEALTHCHECK --interval=10s --timeout=3s --start-period=5s --retries=3 \
+  CMD wget --no-verbose --tries=1 --spider http://localhost:8080/ping || exit 1
+
+# Uruchomienie Traefik
+ENTRYPOINT ["traefik"]

containers/llm-orchestrator-min/microservices/api-gateway/dynamic_conf.yml

@@ -0,0 +1,84 @@
+## Traefik Dynamic Configuration
+
+# Konfiguracja HTTP
+http:
+  # Middlewares
+  middlewares:
+    # Rate limiting
+    rate-limit:
+      rateLimit:
+        average: 100
+        burst: 50
+    
+    # Retry
+    retry-middleware:
+      retry:
+        attempts: 3
+        initialInterval: "500ms"
+    
+    # Compression
+    compress-middleware:
+      compress: {}
+    
+    # CORS
+    cors-middleware:
+      cors:
+        allowOriginList:
+          - "*"
+        allowHeaders:
+          - "*"
+        allowMethods:
+          - GET
+          - POST
+          - PUT
+          - DELETE
+          - OPTIONS
+  
+  # Routery
+  routers:
+    # Router dla modelu
+    model-router:
+      rule: "PathPrefix(`/api/model`)"
+      service: model-service
+      middlewares:
+        - rate-limit
+        - retry-middleware
+        - compress-middleware
+        - cors-middleware
+    
+    # Router dla cache
+    cache-router:
+      rule: "PathPrefix(`/api/cache`)"
+      service: cache-service
+      middlewares:
+        - rate-limit
+        - retry-middleware
+        - compress-middleware
+    
+    # Router dla monitoringu
+    monitoring-router:
+      rule: "PathPrefix(`/api/monitoring`)"
+      service: monitoring-service
+  
+  # Usługi
+  services:
+    # Usługa modelu
+    model-service:
+      loadBalancer:
+        servers:
+          - url: "http://model-service:5000"
+        passHostHeader: true
+    
+    # Usługa cache
+    cache-service:
+      loadBalancer:
+        servers:
+          - url: "http://cache-service:5000"
+        passHostHeader: true
+    
+    # Usługa monitoringu
+    monitoring-service:
+      loadBalancer:
+        servers:
+          - url: "http://monitoring-service:5000"
+        passHostHeader: true

containers/llm-orchestrator-min/microservices/api-gateway/traefik.yml

@@ -0,0 +1,32 @@
+## Traefik Static Configuration
+
+# Konfiguracja globalna
+global:
+  checkNewVersion: true
+  sendAnonymousUsage: false
+
+# Konfiguracja logowania
+log:
+  level: INFO
+
+# Konfiguracja API
+api:
+  insecure: true
+  dashboard: true
+
+# Konfiguracja dostawców
+providers:
+  docker:
+    endpoint: "unix:///var/run/docker.sock"
+    exposedByDefault: false
+    watch: true
+  file:
+    filename: /etc/traefik/dynamic_conf.yml
+    watch: true
+
+# Konfiguracja punktów wejścia
+entryPoints:
+  web:
+    address: ":80"
+  dashboard:
+    address: ":8080"

containers/llm-orchestrator-min/microservices/model-service/download_model.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+# Skrypt do pobierania modelu TinyLlama
+
+MODEL_DIR="/app/models/tinyllama"
+MODEL_URL_BASE="https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0/resolve/main"
+MODEL_FILES=(
+  "tokenizer.model"
+  "tokenizer_config.json"
+  "config.json"
+  "pytorch_model.bin"
+)
+
+# Tworzenie katalogu dla modelu
+mkdir -p $MODEL_DIR
+
+# Sprawdzenie, czy model już istnieje
+if [ ! -f "$MODEL_DIR/pytorch_model.bin" ]; then
+  echo "Pobieranie modelu TinyLlama..."
+  
+  # Pobieranie plików modelu
+  for file in "${MODEL_FILES[@]}"; do
+    echo "Pobieranie $file..."
+    wget -q "$MODEL_URL_BASE/$file" -O "$MODEL_DIR/$file"
+    
+    # Sprawdzenie, czy pobieranie się powiodło
+    if [ $? -ne 0 ]; then
+      echo "Błąd podczas pobierania $file!"
+      exit 1
+    fi
+  done
+  
+  echo "Model pobrany pomyślnie."
+else
+  echo "Model TinyLlama już istnieje, pomijanie pobierania."
+fi

containers/llm-orchestrator-min/microservices/model-service/requirements.txt

@@ -0,0 +1,5 @@
+torch==2.0.1
+transformers==4.30.2
+flask==2.3.2
+numpy==1.24.3
+gunicorn==21.2.0