Update renovate/renovate Docker tag to v43.56.1

apps/llama/configs/config.yaml

@@ -5,15 +5,15 @@ logToStdout: "both" # proxy and upstream
 hooks:
   on_startup:
     preload:
-      - "Qwen3-VL-2B-Instruct-GGUF:Q4_K_M"
+      - "Qwen3.5-0.8B-GGUF-nothink:Q4_K_XL"
 
 groups:
-  qwen-vl-always:
+  always:
     persistent: true
     exclusive: false
     swap: false
     members:
-      - "Qwen3-VL-2B-Instruct-GGUF:Q4_K_M"
+      - "Qwen3.5-0.8B-GGUF-nothink:Q4_K_XL"
 
 models:
   "DeepSeek-R1-0528-Qwen3-8B-GGUF":
@@ -505,6 +505,7 @@ models:
       /app/llama-server
         -hf mradermacher/Qwen3.5-35B-A3B-heretic-GGUF:Q4_K_M
         --mmproj-url https://huggingface.co/unsloth/Qwen3.5-35B-A3B-GGUF/resolve/main/mmproj-F16.gguf
+        --mmproj /root/.cache/llama.cpp/unsloth_Qwen3.5-35B-A3B-GGUF_mmproj-F16.gguf
         --ctx-size 16384
         --temp 1.0
         --min-p 0.00
@@ -519,6 +520,7 @@ models:
       /app/llama-server
         -hf mradermacher/Qwen3.5-35B-A3B-heretic-GGUF:Q4_K_M
         --mmproj-url https://huggingface.co/unsloth/Qwen3.5-35B-A3B-GGUF/resolve/main/mmproj-F16.gguf
+        --mmproj /root/.cache/llama.cpp/unsloth_Qwen3.5-35B-A3B-GGUF_mmproj-F16.gguf
         --ctx-size 16384
         --temp 1.0
         --min-p 0.00
@@ -529,12 +531,13 @@ models:
         --chat-template-kwargs "{\"enable_thinking\": false}"
 
   "Qwen3-VL-2B-Instruct-GGUF:Q4_K_M":
-    ttl: 0
+    ttl: 600
     cmd: |
       /app/llama-server
         -hf unsloth/Qwen3-VL-2B-Instruct-GGUF:Q4_K_M
-        --ctx-size 16384
-        --predict 4096
+        --ctx-size 6144
+        --cache-type-k q8_0
+        --cache-type-v q8_0
         --temp 0.7
         --top-p 0.8
         --top-k 20
@@ -557,3 +560,17 @@ models:
         --repeat-penalty 1.0
         --no-warmup
         --port ${PORT}
+
+  "Qwen3.5-0.8B-GGUF-nothink:Q4_K_XL":
+    ttl: 0
+    cmd: |
+      /app/llama-server
+        -hf unsloth/Qwen3.5-0.8B-GGUF:Q4_K_XL
+        --ctx-size 16384
+        --temp 0.6
+        --top-p 0.95
+        --top-k 20
+        --min-p 0.00
+        --no-warmup
+        --port ${PORT}
+        --chat-template-kwargs "{\"enable_thinking\": false}"

apps/llama/deployment.yaml

@@ -17,7 +17,7 @@ spec:
       containers:
         - name: llama-swap
           # TODO: make renovate update the image tag
-          image: ghcr.io/mostlygeek/llama-swap:v195-vulkan-b8148
+          image: ghcr.io/mostlygeek/llama-swap:v197-vulkan-b8202
           imagePullPolicy: IfNotPresent
           command:
             - /app/llama-swap

apps/renovate/cronjob.yaml

@@ -15,7 +15,7 @@ spec:
             - name: renovate
               # Update this to the latest available and then enable Renovate on
               # the manifest
-              image: renovate/renovate:43.55.1-full
+              image: renovate/renovate:43.56.1-full
               envFrom:
                 - secretRef:
                     name: renovate-gitea-token