deploy qwen3.5 4b

litellm/litellm.yaml

@@ -52,6 +52,12 @@ data:
           model: openai/qwen3.6-27b
           api_base: http://llamacpp-qwen36-27b.llamacpp/v1
           api_key: "sk-no-auth"
+      # Small, fast model intended for Home Assistant voice Assist.
+      - model_name: qwen3.5-4b
+        litellm_params:
+          model: openai/qwen3.5-4b
+          api_base: http://llamacpp-qwen35-4b.llamacpp/v1
+          api_key: "sk-no-auth"
     router_settings:
       fallbacks:
         - deepseek-v4-flash-0731: [qwen3.6-27b]

llamacpp/README.md

@@ -15,6 +15,7 @@ to the NUCBox (`nodeSelector: {kubernetes.io/arch: amd64, hardware: high-memory}
 | Alias                    | Model                          | GGUF                                                  | Service                                              | Args ref |
 |--------------------------|--------------------------------|-------------------------------------------------------|------------------------------------------------------|----------|
 | `deepseek-v4-flash-0731` | DeepSeek-V4-Flash-0731 (MoE)   | unsloth/DeepSeek-V4-Flash-0731-GGUF (UD-IQ1_M, ~87 GiB) | `llamacpp-deepseek-v4-flash-0731.llamacpp:80`        | [args-deepseek-v4-flash-0731.md](args-deepseek-v4-flash-0731.md) |
+| `qwen3.5-4b`             | Qwen3.5-4B (MTP)              | unsloth/Qwen3.5-4B-MTP-GGUF (`UD-Q4_K_XL`)           | `llamacpp-qwen35-4b.llamacpp:80`                    | — |
 
 DeepSeek-V4-Flash-0731 is a Mixture-of-Experts model (256 experts, 6 active per
 token) with MLA attention, so only a small fraction of the weights is computed
@@ -150,3 +151,8 @@ to VRAM).
    directory recursively.)
 4. Check the VRAM budget table above — at ~87 GiB this model nearly fills the
    90 GiB pool on its own, so co-locating another large model is not possible.
+
+The `qwen3.5-4b` alias is exposed through LiteLLM as a low-latency option for
+Home Assistant voice Assist. Select this model in the Home Assistant
+conversation/voice Assist provider configuration; the LiteLLM endpoint is
+`http://litellm-service.litellm:80/v1` from inside the cluster.

llamacpp/deployment-qwen35-4b.yaml

@@ -0,0 +1,138 @@
+# Qwen3.5-4B (MTP, UD-Q4_K_XL) served by llama.cpp for low-latency assistants.
+# The model is fully offloaded to the NUCBox Radeon 8060S. Model weights are
+# stored on the shared llamacpp-models hostPath PVC.
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: llamacpp-qwen35-4b
+  namespace: llamacpp
+  labels:
+    app: llamacpp
+    model: qwen3.5-4b
+spec:
+  replicas: 1
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      app: llamacpp
+      model: qwen3.5-4b
+  template:
+    metadata:
+      labels:
+        app: llamacpp
+        model: qwen3.5-4b
+    spec:
+      nodeSelector:
+        kubernetes.io/arch: amd64
+        hardware: high-memory
+      initContainers:
+      - name: fetch-model
+        image: alpine:3.20
+        command: ["/bin/sh", "-c"]
+        args:
+        - |
+          set -e
+          if [ -s "/models/$MODEL_FILE" ]; then
+            echo "Model $MODEL_FILE already present — skipping download."
+            exit 0
+          fi
+          apk add --no-cache curl
+          echo "Downloading $MODEL_FILE from $MODEL_URL ..."
+          curl -fL --retry 5 --retry-delay 5 -o "/models/$MODEL_FILE.partial" "$MODEL_URL"
+          mv "/models/$MODEL_FILE.partial" "/models/$MODEL_FILE"
+          echo "Download complete: $(ls -lh /models/$MODEL_FILE)"
+        env:
+        - name: MODEL_URL
+          value: "https://huggingface.co/unsloth/Qwen3.5-4B-MTP-GGUF/resolve/main/Qwen3.5-4B-UD-Q4_K_XL.gguf"
+        - name: MODEL_FILE
+          value: "Qwen3.5-4B-UD-Q4_K_XL.gguf"
+        volumeMounts:
+        - name: models
+          mountPath: /models
+      containers:
+      - name: llama-server
+        image: ghcr.io/ggml-org/llama.cpp:server-vulkan
+        imagePullPolicy: IfNotPresent
+        args:
+        - -m
+        - /models/Qwen3.5-4B-UD-Q4_K_XL.gguf
+        - --alias
+        - qwen3.5-4b
+        - --host
+        - 0.0.0.0
+        - --port
+        - "8080"
+        - --jinja
+        - -ngl
+        - "999"
+        - -c
+        - "8192"
+        - -np
+        - "1"
+        - --cont-batching
+        - --cache-type-k
+        - q8_0
+        - --cache-type-v
+        - q8_0
+        - --threads
+        - "8"
+        ports:
+        - name: http
+          containerPort: 8080
+        resources:
+          requests:
+            cpu: "1000m"
+            memory: 2Gi
+          limits:
+            cpu: "4000m"
+            memory: 8Gi
+        readinessProbe:
+          httpGet:
+            path: /health
+            port: 8080
+          initialDelaySeconds: 30
+          periodSeconds: 10
+          failureThreshold: 6
+        livenessProbe:
+          httpGet:
+            path: /health
+            port: 8080
+          initialDelaySeconds: 120
+          periodSeconds: 30
+          failureThreshold: 5
+        securityContext:
+          privileged: true
+        volumeMounts:
+        - name: models
+          mountPath: /models
+          readOnly: true
+        - name: dri
+          mountPath: /dev/dri
+      volumes:
+      - name: models
+        persistentVolumeClaim:
+          claimName: llamacpp-models
+      - name: dri
+        hostPath:
+          path: /dev/dri
+          type: Directory
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: llamacpp-qwen35-4b
+  namespace: llamacpp
+  labels:
+    app: llamacpp
+    model: qwen3.5-4b
+spec:
+  type: ClusterIP
+  selector:
+    app: llamacpp
+    model: qwen3.5-4b
+  ports:
+  - name: http
+    port: 80
+    targetPort: 8080

llamacpp/deployment-qwen36-27b.yaml

@@ -71,7 +71,7 @@ spec:
         - -c
         - "131072"
         - -np
-        - "2"
+        - "1"
         - --cont-batching
         - --cache-type-k
         - q8_0