Update Dockerfile

Dockerfile

@@ -1,28 +1,38 @@
-# Use official Python image
+# Use a lightweight Python base image
 FROM python:3.10-slim
+
+# Disable CUDA since we're using CPU-only
 ENV TORCH_DISABLE_CUDA=1
+ENV TRANSFORMERS_CACHE=/app/hf_cache
+ENV HF_HOME=/app/hf_cache
 
 # Set working directory
 WORKDIR /app
 
-# Copy requirements and install dependencies
+# Copy dependency list
 COPY requirements.txt .
 
-# Upgrade pip and install pinned dependencies
-RUN pip install --no-cache-dir --upgrade pip \
-    && pip install --no-cache-dir -r requirements.txt
+# Install dependencies efficiently
+RUN apt-get update && apt-get install -y git && \
+    pip install --no-cache-dir -r requirements.txt && \
+    rm -rf /var/lib/apt/lists/*
 
-# Copy all code and model files
+# Copy application code
 COPY . .
 
-# Set Transformers cache to a writable folder
-ENV TRANSFORMERS_CACHE=/tmp/hf_cache
-
-# Expose FastAPI port
-EXPOSE 7860
+# Make cache folder writable
+RUN mkdir -p /app/hf_cache && chmod -R 777 /app/hf_cache
 
-# Precompute embeddings at build time (optional: adjust if dataset is large)
+# Expose port
+EXPOSE 8080
 
+# Preload model embeddings (optional)
+# Commented out to speed up Cloud Build — will load at runtime in main.py
+# RUN python -c "import torch, pandas as pd; from sentence_transformers import SentenceTransformer; \
+# model = SentenceTransformer('./muril_combined_multilingual_model'); \
+# df = pd.read_csv('./muril_multilingual_dataset.csv').dropna(subset=['question','answer']); \
+# embeddings = model.encode(df['answer'].tolist(), convert_to_tensor=True); \
+# torch.save(embeddings, './answer_embeddings.pt'); print('✅ Precomputed embeddings saved');"
 
-# Run FastAPI with uvicorn
-CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "]()
+# Run the FastAPI app with uvicorn
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8080"]