Merge pull request #38 from sirius0xdev/feat/optimize-vllm-cost-perf

feat: optimize cost via L4 spot instances and vLLM performance tuning

apps/base/customer1/openclaw/vllm-gemma.yaml

@@ -24,7 +24,7 @@ spec:
           effect: "NoSchedule"
       containers:
         - name: vllm-brain
-          image: vllm/vllm-openai:latest
+          image: vllm/vllm-openai:v0.4.2
           command: ["python3", "-m", "vllm.entrypoints.openai.api_server"]
           env:
             - name: HF_TOKEN
@@ -44,6 +44,8 @@ spec:
             - --tool-call-parser=qwen3_coder
             - --reasoning-parser=qwen3
             - --enable-prefix-caching
+            - --enable-chunked-prefill
+            - --kv-cache-dtype=fp8
             ##- --limit-mm-per-prompt='{"image"=1, "video"=0}'  # Optional: limit vision if not heavily using images yet 
           ports:
             - containerPort: 8000

apps/base/customer1/openclaw/vllm-l4.yaml

@@ -24,7 +24,7 @@ spec:
           effect: "NoSchedule"
       containers:
         - name: vllm-brain-l4
-          image: vllm/vllm-openai:latest
+          image: vllm/vllm-openai:v0.4.2
           command: ["python3", "-m", "vllm.entrypoints.openai.api_server"]
           env:
             - name: HF_TOKEN

modules/nodepool-gpu.tf

@@ -23,7 +23,7 @@ resource "google_container_node_pool" "gpu_pool" {
     }
 
     
-    spot = false 
+    spot = true 
 
     oauth_scopes = [
       "https://www.googleapis.com/auth/cloud-platform"