#!/bin/sh
set -eu

model=${WHISPER_MODEL:-/models/ggml-small.bin}
model_url=${WHISPER_MODEL_URL:-https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.bin}
model_dir=$(dirname "$model")

mkdir -p "$model_dir"
chown 10004:10004 "$model_dir"

if [ ! -s "$model" ]; then
  partial="${model}.part"
  echo "Downloading Whisper model to persistent storage"
  # All writes below must use the volume owner. The container deliberately
  # drops CAP_DAC_OVERRIDE, so even uid 0 cannot rename a file in the
  # whisper-owned directory after capabilities have been removed.
  if [ ! -s "$partial" ]; then
    gosu whisper rm -f "$partial"
    gosu whisper curl --fail --location --retry 5 --retry-delay 5 \
      --output "$partial" "$model_url"
  fi
  gosu whisper mv "$partial" "$model"
fi

server_port=${WHISPER_SERVER_PORT:-8085}
server_log=/tmp/whisper-server.log

gosu whisper /opt/whisper.cpp/build/bin/whisper-server \
  --model "$model" \
  --threads "${WHISPER_THREADS:-8}" \
  --language "${WHISPER_LANGUAGE:-de}" \
  --host 127.0.0.1 \
  --port "$server_port" \
  --inference-path /inference \
  --no-gpu >"$server_log" 2>&1 &

server_pid=$!
worker_pid=""
cleanup() {
  [ -z "$worker_pid" ] || kill "$worker_pid" 2>/dev/null || true
  kill "$server_pid" 2>/dev/null || true
}
trap cleanup EXIT INT TERM

# Der Modell-Load dauert beim Containerstart kurz. Der HTTP-Wrapper wird erst
# freigegeben, wenn whisper-server tatsächlich Antworten annimmt.
i=0
until curl --fail --silent "http://127.0.0.1:${server_port}/" >/dev/null; do
  i=$((i + 1))
  if ! kill -0 "$server_pid" 2>/dev/null; then
    cat "$server_log"
    exit 1
  fi
  if [ "$i" -ge 180 ]; then
    echo "whisper-server did not become ready" >&2
    cat "$server_log"
    exit 1
  fi
  sleep 1
done

export WHISPER_SERVER_URL="http://127.0.0.1:${server_port}"
gosu whisper python /app/stt_worker.py &
worker_pid=$!
wait "$worker_pid"