start.sh

#!/usr/bin/env bash
set -euo pipefail

echo "== Container start =="
echo "ENV=${ENV:-dev}"
echo "CHROMA_ROOT=${CHROMA_ROOT:-/data/chroma}"

# Ensure Chroma root exists
mkdir -p "${CHROMA_ROOT}"

# Decide whether ingestion is needed (if any level folder missing or empty)
_need_ingest=0
for level in low mid high; do
  lvl_dir="${CHROMA_ROOT}/${level}"
  if [ ! -d "$lvl_dir" ] || [ -z "$(ls -A "$lvl_dir" 2>/dev/null || true)" ]; then
    _need_ingest=1
  fi
done

if [ "${_need_ingest}" -eq 1 ]; then
  echo "No (or empty) Chroma data found → running ingestion..."
  # Ingest PDFs from /app/pdfs/{low,mid,high} into ${CHROMA_ROOT}/{low,mid,high}
  python -m ragg.ingest_all || echo "WARNING: ingestion returned non-zero exit"
else
  echo "Chroma already present → skipping ingestion."
fi

# Start the API
exec gunicorn --workers 2 --threads 4 --timeout 120 -b 0.0.0.0:7860 verification:app