Fix: add LiteLLM proxy so Claude Code can use qwen2.5-coder:32b

docker-compose.yml

@@ -22,22 +22,37 @@ services:
       - OLLAMA_HOST=0.0.0.0
       - OLLAMA_KEEP_ALIVE=24h
 
+  litellm:
+    image: ghcr.io/berriai/litellm:main-latest
+    container_name: litellm
+    restart: unless-stopped
+    depends_on:
+      - ollama
+    ports:
+      - "4000:4000"
+    volumes:
+      - ./litellm-config.yaml:/app/config.yaml
+    command: ["--config", "/app/config.yaml", "--port", "4000"]
+
   claude-code:
     image: node:20-slim
     container_name: claude-code
     restart: unless-stopped
     depends_on:
-      - ollama
+      - litellm
     environment:
-      - ANTHROPIC_BASE_URL=http://ollama:11434/v1
+      # Point Claude Code at LiteLLM which fully speaks the Anthropic API
+      - ANTHROPIC_BASE_URL=http://litellm:4000
       - ANTHROPIC_API_KEY=ollama
-      - CLAUDE_CODE_DEFAULT_MODEL=qwen2.5-coder:32b
     volumes:
       - workspace:/workspace
+      - claude_config:/root/.claude
     working_dir: /workspace
     command: >
       sh -c "npm install -g @anthropic-ai/claude-code &&
-             echo 'Claude Code installed. Run: claude' &&
+             mkdir -p /root/.claude &&
+             echo '{\"model\": \"qwen2.5-coder:32b\"}' > /root/.claude/settings.json &&
+             echo 'Claude Code ready. Run: claude' &&
              tail -f /dev/null"
     stdin_open: true
     tty: true
@@ -47,3 +62,5 @@ volumes:
     driver: local
   workspace:
     driver: local
+  claude_config:
+    driver: local

litellm-config.yaml

@@ -0,0 +1,8 @@
+model_list:
+  - model_name: qwen2.5-coder:32b
+    litellm_params:
+      model: ollama/qwen2.5-coder:32b
+      api_base: http://ollama:11434
+
+general_settings:
+  master_key: ollama