Update start.sh

start.sh

@@ -1,31 +1,25 @@
 #!/bin/bash
 set -e
 
-# --- CONFIGURATION ---
 echo "------------------------------------------------"
 echo "🛠️  Setting up Environment"
 echo "------------------------------------------------"
 
-# Create model directory
 mkdir -p /app/models
 
-download_model () {
+download_file () {
     REPO_ID=$1
     FILENAME=$2
-    LOCAL_NAME=$3
+    DEST_PATH=$3
 
-    if [ -z "$LOCAL_NAME" ]; then LOCAL_NAME="$FILENAME"; fi
-    FILE_PATH="/app/models/$LOCAL_NAME"
-
-    if [ ! -f "$FILE_PATH" ]; then
-        echo "⬇️  Downloading $LOCAL_NAME ..."
+    if [ ! -f "$DEST_PATH" ]; then
+        echo "⬇️  Downloading $FILENAME ..."
         python3 -c "
 from huggingface_hub import hf_hub_download
-import os, sys
+import shutil, sys
 try:
-    hf_hub_download(repo_id='$REPO_ID', filename='$FILENAME', local_dir='/app/models')
-    if '$FILENAME' != '$LOCAL_NAME':
-        os.rename('/app/models/$FILENAME', '/app/models/$LOCAL_NAME')
+    path = hf_hub_download(repo_id='$REPO_ID', filename='$FILENAME')
+    shutil.copy(path, '$DEST_PATH')
 except Exception as e:
     print(f'Error downloading $FILENAME: {e}')
     sys.exit(1)
@@ -33,23 +27,22 @@ except Exception as e:
     fi
 }
 
-# 1. TEXT: Qwen 2.5 Coder
-download_model "Qwen/Qwen2.5-Coder-14B-Instruct-GGUF" "qwen2.5-coder-14b-instruct-q4_k_m.gguf"
-
-# 2. IMAGE: Qwen 2.5 VL (correct case-sensitive filename)
-download_model "ggml-org/Qwen2.5-VL-7B-Instruct-GGUF" "Qwen2.5-VL-7B-Instruct-Q4_K_M.gguf" "qwen2.5-vl-7b-instruct-q4_k_m.gguf"
+# 1. TEXT/CODE: Qwen 2.5 Coder 14B (single file, text-only)
+download_file "Qwen/Qwen2.5-Coder-14B-Instruct-GGUF" "qwen2.5-coder-14b-instruct-q4_k_m.gguf" "/app/models/qwen2.5-coder-14b-instruct-q4_k_m.gguf"
 
-# 3. AUDIO: Qwen 2 Audio
-download_model "gaianet/Qwen2-Audio-7B-Instruct-GGUF" "Qwen2-Audio-7B-Instruct-Q4_K_M.gguf" "qwen2-audio-7b-instruct-q4_k_m.gguf"
+# 2. MULTIMODAL (Image + Audio): Qwen 2.5 Omni 7B
+# Needs subdirectory with model + mmproj
+mkdir -p /app/models/qwen2.5-omni-7b
+download_file "ggml-org/Qwen2.5-Omni-7B-GGUF" "Qwen2.5-Omni-7B-Q4_K_M.gguf" "/app/models/qwen2.5-omni-7b/Qwen2.5-Omni-7B-Q4_K_M.gguf"
+download_file "ggml-org/Qwen2.5-Omni-7B-GGUF" "mmproj-Qwen2.5-Omni-7B-Q8_0.gguf" "/app/models/qwen2.5-omni-7b/mmproj-Qwen2.5-Omni-7B-Q8_0.gguf"
 
 echo "------------------------------------------------"
 echo "🚀 Starting Server in Router Mode"
 echo "------------------------------------------------"
 
-# Start in ROUTER MODE - serve all models from /app/models
 exec /usr/local/bin/llama-server \
     --models-dir /app/models \
-    --models-max 3 \
+    --models-max 2 \
     --host 0.0.0.0 \
     --port 7860 \
     -c 8192 \