Update TTS component

docker-compose.yml

@@ -20,15 +20,28 @@ services:
     networks:
       - nointernet
 
-  openedai-speech:
+  ollama_provision:
     build:
-      dockerfile: src/tts/Dockerfile
+      dockerfile: src/ollama_provision/Dockerfile
     environment:
-      - TTS_HOME=voices
+      - MODELS=nomic-embed-text,llama3.1:8b,qwen2.5:7b,gemma3:12b-it-qat,gemma3:27b-it-qat,gpt-oss:20b
     volumes:
-      - voices:/app/voices
-      - speech-config:/app/config
+      - ollama:/root/.ollama
+    restart: no
+    depends_on:
+      ollama:
+        condition: service_healthy
+        restart: true
+    links:
+      - ollama
+    networks:
+      - internet
+
+  speaches-ai:
+    image: ghcr.io/speaches-ai/speaches:latest-cuda
     restart: unless-stopped
+    volumes:
+      - hf-hub-cache:/home/ubuntu/.cache/huggingface/hub
     deploy:
       resources:
         reservations:
@@ -45,20 +58,19 @@ services:
     networks:
       - nointernet
 
-  ollama_provision:
+  speaches-ai_provision:
     build:
-      dockerfile: src/ollama_provision/Dockerfile
-    environment:
-      - MODELS=qwen2.5:latest,qwen2.5-coder:32b,nomic-embed-text:latest,gemma2:latest,mistral:latest,deepseek-r1:7b,deepcoder:14b
+      dockerfile: src/speaches-ai_provision/Dockerfile
     volumes:
-      - ollama:/root/.ollama
+      - hf-hub-cache:/home/ubuntu/.cache/huggingface/hub
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: all
+              capabilities: [gpu]
     restart: no
-    depends_on:
-      ollama:
-        condition: service_healthy
-        restart: true
-    links:
-      - ollama
     networks:
       - internet
 
@@ -89,17 +101,23 @@ services:
       - NGINX_ENVSUBST_OUTPUT_DIR=/etc/nginx
       - API_KEY=${LLM_API_KEY}
     depends_on:
-      - openedai-speech
       - ollama
+      - speaches-ai
       - aichat
     links:
       - ollama
-      - openedai-speech
+      - speaches-ai
       - aichat
     ports:
-      - "127.0.0.1:11434:11434"
+      - "127.0.0.1:10434:10434"
       - "127.0.0.1:8000:8000"
       - "127.0.0.1:8001:8001"
+    healthcheck:
+      test: curl --fail http://localhost:11434 || exit 1
+      interval: 10s
+      retries: 5
+      start_period: 10s
+      timeout: 10s
     restart: unless-stopped
     networks:
       - internet
@@ -107,8 +125,7 @@ services:
 
 volumes:
   ollama:
-  voices:
-  speech-config:
+  hf-hub-cache:
 
 networks:
   internet: