Increase default context size from 8192 to 16384 tokens

Fixes 400 error when image requests exceed 8192 token context window.
Aligns shell script and Server Control node defaults with GGUF node auto-start.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

nodes/inputs/archai3d_qwenvl_gguf.py

@@ -844,7 +844,7 @@ class ArchAi3D_QwenVL_Server_Control:
                     "tooltip": "GPU layers (99=all on GPU, lower values use more CPU to save VRAM)"
                 }),
                 "context_size": ("INT", {
-                    "default": 8192,
+                    "default": 16384,
                     "min": 1024,
                     "max": 32768,
                     "step": 1024,
@@ -900,7 +900,7 @@ class ArchAi3D_QwenVL_Server_Control:
             return False, str(e)
 
     def control(self, action, model_size, quantization="Q4_K_M (Smaller, ~5GB)",
-                gpu_layers=99, context_size=8192,
+                gpu_layers=99, context_size=16384,
                 flash_attention="auto (Recommended)", kv_cache_type="f16 (Best Quality)",
                 parallel_slots=2, trigger=None, cache_to_local_ssd=True):
         """Control the llama-server with quality and speed optimizations."""

start_qwenvl_server.sh

@@ -24,7 +24,7 @@
 # ============================================================================
 
 # Configuration
-CTX="${CTX:-8192}"
+CTX="${CTX:-16384}"
 GPU_LAYERS="${GPU_LAYERS:-99}"  # 99 for all layers on GPU
 
 # ============================================================================