#!/bin/bash
set -e

INSTANCES=${LLAMA_INSTANCES:-2}
BASE_PORT=${LLAMA_BASE_PORT:-9000}
READY_TIMEOUT=${LLAMA_READY_TIMEOUT:-600}

echo "Starting $INSTANCES llama-server instances on ports $BASE_PORT-$((BASE_PORT+INSTANCES-1))"

for ((i=0; i<INSTANCES; i++)); do
    PORT=$((BASE_PORT + i))
    echo "Starting instance $((i+1))/$INSTANCES on port $PORT..."
    LLAMA_ARG_PORT=$PORT /app/llama-server \
        > /var/log/llama-server-$PORT.log 2>&1 &
    sleep 3
done

echo "Waiting for servers to be ready..."
for ((i=0; i<INSTANCES; i++)); do
    PORT=$((BASE_PORT + i))
    elapsed=0
    while [ $elapsed -lt $READY_TIMEOUT ]; do
        if curl -s http://127.0.0.1:$PORT/health > /dev/null 2>&1; then
            echo "Instance on port $PORT is ready"
            break
        fi
        sleep 5
        elapsed=$((elapsed + 5))
    done
    if [ $elapsed -ge $READY_TIMEOUT ]; then
        echo "ERROR: Server on port $PORT did not become ready after ${READY_TIMEOUT}s"
    fi
done

echo "All instances ready. Monitoring logs..."
tail -f /var/log/llama-server-*.log &
wait