Enh model prompting and sequential processing

Dockerfile

@@ -13,13 +13,16 @@ WORKDIR $HOME/app
 ARG IS_LOCAL=false
 ENV IS_LOCAL=${IS_LOCAL}
 
-# Install Python dependencies first (better layer caching)
+# Install Python dependencies based on mode (better layer caching)
 COPY --chown=user requirements.txt .
-# Install local mode dependencies if IS_LOCAL is true
 COPY --chown=user requirements-dev.txt .
-RUN pip install --upgrade pip && pip install --no-cache-dir -r requirements.txt && \
+RUN pip install --upgrade pip && \
     if [ "$IS_LOCAL" = "true" ]; then \
+        echo "Installing LOCAL mode dependencies (MedAlpaca-13b)"; \
         pip install --no-cache-dir -r requirements-dev.txt; \
+    else \
+        echo "Installing CLOUD mode dependencies (NVIDIA/Gemini APIs)"; \
+        pip install --no-cache-dir -r requirements.txt; \
     fi
 
 # Copy the application

build.sh

@@ -60,8 +60,8 @@ elif [ "$1" = "cloud" ]; then
 else
     # Interactive mode selection
     echo "Please select the runtime mode:"
-    echo "1) Local mode (MedAlpaca-13b) - No API costs, complete privacy"
-    echo "2) Cloud mode (NVIDIA/Gemini APIs) - Faster processing, requires API keys"
+    echo "1) Local mode (MedAlpaca-13b) - No API costs, complete privacy, heavy dependencies"
+    echo "2) Cloud mode (NVIDIA/Gemini APIs) - Faster processing, requires API keys, lightweight"
     echo ""
     
     while true; do
@@ -83,9 +83,13 @@ if prompt_yes_no "Would you like to build/rebuild the Docker image?"; then
     echo ""
     if [ "$MODE" = "local" ]; then
         echo "🏠 Building in LOCAL mode (MedAlpaca-13b)..."
+        echo "   - Installing local LLM dependencies (torch-cuda, accelerate, bitsandbytes, etc.)"
+        echo "   - Excluding cloud API dependencies (google-genai, google-auth, etc.)"
         docker build --build-arg IS_LOCAL=true -t medai-processing:local .
     else
         echo "☁️  Building in CLOUD mode (NVIDIA/Gemini APIs)..."
+        echo "   - Installing cloud API dependencies (google-genai, google-auth, etc.)"
+        echo "   - Excluding heavy local LLM dependencies (torch-cuda, accelerate, bitsandbytes, etc.)"
         docker build --build-arg IS_LOCAL=false -t medai-processing:cloud .
     fi
     

requirements-dev.txt

@@ -1,6 +1,18 @@
 # Local mode dependencies for MedAlpaca-13b inference
+# Core framework dependencies
+fastapi
+uvicorn[standard]
+python-dotenv
+huggingface_hub
+requests
+orjson
+ftfy
+langid
+
+# Local LLM dependencies
 torch>=2.0.0
 torch-cuda>=2.0.0
+transformers>=4.30.0
 accelerate>=0.20.0
 bitsandbytes>=0.41.0
 peft>=0.4.0
@@ -8,3 +20,8 @@ datasets>=2.14.0
 evaluate>=0.4.0
 scipy>=1.10.0
 scikit-learn>=1.3.0
+sentencepiece
+sacremoses
+
+# Vietnamese translation (Opus model fallback)
+# Note: google-genai and google-auth packages are excluded for local mode

requirements.txt

@@ -1,17 +1,25 @@
+# Cloud mode dependencies for NVIDIA/Gemini API integration
+# Core framework dependencies
 fastapi
 uvicorn[standard]
 python-dotenv
 huggingface_hub
 requests
+orjson
+ftfy
+langid
+
+# Cloud API dependencies
 google-genai
 google-api-python-client
 google-auth
 google-auth-httplib2
 google-auth-oauthlib
-orjson
-ftfy
-langid
+
+# Vietnamese translation (Opus model fallback)
 transformers
 torch
 sentencepiece
 sacremoses
+
+# Note: Heavy local LLM dependencies (accelerate, bitsandbytes, etc.) are excluded for cloud mode