Dockerfile

FROM python:3.11-slim

RUN apt-get update && apt-get install -y --no-install-recommends \
    ffmpeg \
    build-essential \
    git \
    && rm -rf /var/lib/apt/lists/*

RUN pip install --upgrade pip setuptools wheel

# Create appuser early so all downloads happen as them
RUN useradd -m -u 1000 appuser

# Set cache dirs to appuser-owned locations from the start
ENV HOME=/home/appuser
ENV HF_HOME=/home/appuser/.cache/huggingface
ENV TORCH_HOME=/home/appuser/.cache/torch

WORKDIR /app
RUN chown appuser:appuser /app

# Install dependencies as root (fine for pip)
COPY backend/requirements.txt ./backend/requirements.txt
RUN pip install --no-cache-dir -r backend/requirements.txt

# Pre-download models as appuser so file ownership is correct
USER appuser

RUN python -c "\
from transformers import MarianMTModel, MarianTokenizer; \
print('Downloading translation model...'); \
MarianTokenizer.from_pretrained('Helsinki-NLP/opus-mt-en-mul'); \
MarianMTModel.from_pretrained('Helsinki-NLP/opus-mt-en-mul'); \
print('Done.')"

RUN python -c "\
from speechbrain.inference.TTS import Tacotron2; \
from speechbrain.inference.vocoders import HIFIGAN; \
print('Downloading TTS models...'); \
Tacotron2.from_hparams(source='Sunbird/sunbird-lug-tts', savedir='/home/appuser/tts_model'); \
HIFIGAN.from_hparams(source='speechbrain/tts-hifigan-ljspeech', savedir='/home/appuser/vocoder'); \
print('Done.')"

RUN python -c "\
import whisper; \
print('Downloading Whisper...'); \
whisper.load_model('base'); \
print('Done.')"

RUN python -c "import nltk; nltk.download('punkt'); nltk.download('punkt_tab')"

# Copy app code
USER root
COPY backend/ ./backend/
COPY frontend/ ./frontend/
RUN chown -R appuser:appuser /app

USER appuser
WORKDIR /app/backend

EXPOSE 7860

CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]