entrypoint_serve.sh

#!/bin/bash
set -euo pipefail

# -------------------------------------------------------------------
# Default values (can be overridden via docker-compose envs or .env)
# -------------------------------------------------------------------
MODEL_NAME="${MODEL_NAME:-loan_default_model}"
MODEL_ALIAS="${MODEL_ALIAS:-staging}"
PORT="${PORT:-5001}"

echo "🚀 Starting Serve container"
echo "   MODEL_NAME: $MODEL_NAME"
echo "   MODEL_ALIAS: $MODEL_ALIAS"
echo "   PORT: $PORT"
echo "   MLFLOW_TRACKING_URI: ${MLFLOW_TRACKING_URI:-not_set}"

# -------------------------------------------------------------------
# Ensure mlflow CLI exists
# -------------------------------------------------------------------
if ! command -v mlflow >/dev/null 2>&1; then
  echo "❌ mlflow command not found in PATH"
  exit 1
fi

# -------------------------------------------------------------------
# Start MLflow model serving
# -------------------------------------------------------------------
exec mlflow models serve \
    -m "models:/${MODEL_NAME}@${MODEL_ALIAS}" \
    --host 0.0.0.0 \
    --port "${PORT}" \
    --no-conda