#!/bin/bash

exec /app/llama-server $EMBEDDING_MEDIUM_MODEL \
  --embeddings --pooling mean \
  --flash-attn auto --threads -1 --threads-http -1 \
  --timeout 600 --host 0.0.0.0 --port 8098 &
PID=$!

cleanup() {
    echo "Stopping llama-server..."
    kill $PID 2>/dev/null
    wait $PID 2>/dev/null
    exit 0
}

trap cleanup SIGTERM SIGINT

wait $PID