Checkpoint: Phase 10: LLM Provider Configuration — Ollama Cloud как дефолт, локальный Ollama закомментирован (GPU only), docker-compose/stack обновлены, .env.example с 4 провайдерами

docker/docker-stack.yml

@@ -1,7 +1,23 @@
 ##############################################################################
 # GoClaw Control Center — Docker Stack (Docker Swarm Production)
 #
+# LLM Provider:
+#   By default the gateway uses Ollama Cloud (https://ollama.com/v1).
+#   Set LLM_BASE_URL and LLM_API_KEY via Docker secrets or environment.
+#
+#   To use a local Ollama instance on a GPU-equipped Swarm node:
+#     1. Uncomment the "ollama" service below.
+#     2. Add the label  gpu=true  to the GPU node:
+#          docker node update --label-add gpu=true <node-id>
+#     3. Change LLM_BASE_URL in the gateway service to: http://ollama:11434
+#
 # Deploy:
+#   # Create required secrets first:
+#   echo "rootpass" | docker secret create mysql-root-password -
+#   echo "pass"     | docker secret create mysql-password -
+#   echo "jwtsecret"| docker secret create jwt-secret -
+#   echo "ollamakey"| docker secret create llm-api-key -
+#
 #   docker stack deploy -c docker/docker-stack.yml goclaw
 #
 # Remove:
@@ -19,14 +35,10 @@ networks:
     attachable: true
 
 volumes:
-  ollama-data:
-    driver: local
   mysql-data:
     driver: local
-
-configs:
-  gateway-env:
-    external: true
+  # ollama-data:       # Uncomment when using local Ollama service below
+  #   driver: local
 
 secrets:
   mysql-root-password:
@@ -35,6 +47,8 @@ secrets:
     external: true
   jwt-secret:
     external: true
+  llm-api-key:
+    external: true
 
 services:
 
@@ -73,45 +87,59 @@ services:
       retries: 5
       start_period: 30s
 
-  # ── Ollama LLM Server ─────────────────────────────────────────────────────
-  ollama:
-    image: ollama/ollama:latest
-    volumes:
-      - ollama-data:/root/.ollama
-    networks:
-      - goclaw-net
-    deploy:
-      replicas: 1
-      placement:
-        constraints:
-          # Pin to GPU node if available
-          - node.labels.gpu == true
-      restart_policy:
-        condition: on-failure
-        delay: 15s
-      resources:
-        limits:
-          memory: 16G
-        reservations:
-          memory: 4G
-    # GPU support via nvidia-container-runtime
-    # Uncomment on GPU-enabled nodes:
-    # runtime: nvidia
-    # environment:
-    #   - NVIDIA_VISIBLE_DEVICES=all
+  # ── Local Ollama LLM Server (GPU ONLY — disabled by default) ─────────────
+  # Uncomment this entire block only on Swarm nodes with a compatible GPU.
+  # After uncommenting, also change gateway LLM_BASE_URL to http://ollama:11434
+  # and remove the llm-api-key secret from gateway (not needed for local Ollama).
+  #
+  # ollama:
+  #   image: ollama/ollama:latest
+  #   volumes:
+  #     - ollama-data:/root/.ollama
+  #   networks:
+  #     - goclaw-net
+  #   deploy:
+  #     replicas: 1
+  #     placement:
+  #       constraints:
+  #         # Pin to GPU-labelled node:  docker node update --label-add gpu=true <id>
+  #         - node.labels.gpu == true
+  #     restart_policy:
+  #       condition: on-failure
+  #       delay: 15s
+  #     resources:
+  #       limits:
+  #         memory: 16G
+  #       reservations:
+  #         memory: 4G
+  #   # NVIDIA GPU support — uncomment on GPU-enabled nodes:
+  #   # runtime: nvidia
+  #   # environment:
+  #   #   - NVIDIA_VISIBLE_DEVICES=all
 
   # ── Go Gateway (Orchestrator + Tool Executor) ─────────────────────────────
   gateway:
     image: git.softuniq.eu/uniqai/goclaw/gateway:latest
     environment:
       PORT: "18789"
-      OLLAMA_BASE_URL: "http://ollama:11434"
+      # ── LLM Provider ─────────────────────────────────────────────────────
+      # Default: Ollama Cloud (requires llm-api-key secret below)
+      LLM_BASE_URL: "${LLM_BASE_URL:-https://ollama.com/v1}"
+      DEFAULT_MODEL: "${DEFAULT_MODEL:-qwen2.5:7b}"
+      # LLM_API_KEY is injected via /run/secrets/llm-api-key (see below)
+      # ── To switch to local GPU Ollama, set: ──────────────────────────────
+      # LLM_BASE_URL: "http://ollama:11434"
+      # (and uncomment the ollama service above)
+      # ─────────────────────────────────────────────────────────────────────
       DATABASE_URL: "goclaw:{{MYSQL_PASSWORD}}@tcp(db:3306)/goclaw?parseTime=true"
       PROJECT_ROOT: "/app"
-      REQUEST_TIMEOUT_SECS: "120"
+      GATEWAY_REQUEST_TIMEOUT_SECS: "120"
+      GATEWAY_MAX_TOOL_ITERATIONS: "10"
       LOG_LEVEL: "info"
     secrets:
       - mysql-password
+      - source: llm-api-key
+        target: /run/secrets/llm-api-key
     networks:
       - goclaw-net
     ports:
@@ -119,6 +147,9 @@ services:
         published: 18789
         protocol: tcp
         mode: ingress
+    volumes:
+      # Docker socket for docker_exec tool
+      - /var/run/docker.sock:/var/run/docker.sock
     deploy:
       replicas: 2
       update_config: