feat: consolidate GPU IP with single GPU_TAILSCALE_IP variable

- Replace COMFYUI_BACKEND_HOST and SUPERVISOR_BACKEND_HOST with GPU_TAILSCALE_IP
- Update LiteLLM config to use os.environ/GPU_TAILSCALE_IP for vLLM models
- Add GPU_TAILSCALE_IP env var to LiteLLM service
- Configure qwen-2.5-7b and llama-3.1-8b to route through orchestrator

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

ai/compose.yaml

@@ -101,6 +101,7 @@ services:
       ANTHROPIC_API_KEY: ${ANTHROPIC_API_KEY}
       LITELLM_MASTER_KEY: ${AI_LITELLM_API_KEY}
       DATABASE_URL: postgresql://${AI_DB_USER}:${AI_DB_PASSWORD}@ai_postgres:5432/litellm
+      GPU_TAILSCALE_IP: ${GPU_TAILSCALE_IP}
       # LITELLM_DROP_PARAMS: 'true'  # DISABLED: Was breaking streaming
       NO_DOCS: 'true'
       NO_REDOC: 'true'
@@ -206,7 +207,7 @@ services:
     restart: unless-stopped
     environment:
       TZ: ${TIMEZONE:-Europe/Berlin}
-      COMFYUI_BACKEND_HOST: ${COMFYUI_BACKEND_HOST}
+      COMFYUI_BACKEND_HOST: ${GPU_TAILSCALE_IP}
       COMFYUI_BACKEND_PORT: ${COMFYUI_BACKEND_PORT:-8188}
     volumes:
       - ./comfyui-nginx.conf:/etc/nginx/nginx.conf.template:ro
@@ -239,7 +240,7 @@ services:
     restart: unless-stopped
     environment:
       TZ: ${TIMEZONE:-Europe/Berlin}
-      SUPERVISOR_BACKEND_HOST: ${SUPERVISOR_BACKEND_HOST}
+      SUPERVISOR_BACKEND_HOST: ${GPU_TAILSCALE_IP}
       SUPERVISOR_BACKEND_PORT: ${SUPERVISOR_BACKEND_PORT:-9001}
     volumes:
       - ./supervisor-nginx.conf:/etc/nginx/nginx.conf.template:ro

ai/litellm-config.yaml

@@ -33,7 +33,7 @@ model_list:
   - model_name: qwen-2.5-7b
     litellm_params:
       model: hosted_vllm/openai/qwen-2.5-7b  # hosted_vllm/openai/ for vLLM via orchestrator
-      api_base: http://100.121.199.88:9000/v1  # RunPod GPU via Tailscale
+      api_base: http://os.environ/GPU_TAILSCALE_IP:9000/v1  # RunPod GPU via Tailscale
       api_key: dummy
       rpm: 1000
       tpm: 100000
@@ -45,7 +45,7 @@ model_list:
   - model_name: llama-3.1-8b
     litellm_params:
       model: hosted_vllm/openai/llama-3.1-8b  # hosted_vllm/openai/ for vLLM via orchestrator
-      api_base: http://100.121.199.88:9000/v1  # RunPod GPU via Tailscale
+      api_base: http://os.environ/GPU_TAILSCALE_IP:9000/v1  # RunPod GPU via Tailscale
       api_key: dummy
       rpm: 1000
       tpm: 100000