start_server.sh

#!/bin/bash
PROJECT_ROOT="/root/claude_tests/NEODEMO4"
PYTHON_BIN="$PROJECT_ROOT/venv/bin/python3"
export MODEL_PATH="$PROJECT_ROOT/model_assets/mistral-7b-v0.1.Q4_K_M.gguf"
export LD_LIBRARY_PATH="$PROJECT_ROOT/venv/lib/python3.12/site-packages/llama_cpp/lib:$LD_LIBRARY_PATH"

echo "Stopping any existing server instance..."
pkill -f api.py && sleep 2

echo "Starting Mistral-7B Inference Server..."
nohup $PYTHON_BIN $PROJECT_ROOT/api.py > $PROJECT_ROOT/server.log 2>&1 &
PID=$!
echo $PID > $PROJECT_ROOT/server.pid

echo "Waiting for server to bind to port 8000..."
max_retries=30
count=0
while ! grep -q "Starting Uvicorn server on 0.0.0.0:8000" "$PROJECT_ROOT/server.log" && [ $count -lt $max_retries ]; do
    sleep 2
    ((count++))
done

if [ $count -eq $max_retries ]; then
    echo "Timeout waiting for server to start. Check $PROJECT_ROOT/server.log"
    exit 1
fi

echo "Server is up! Testing connectivity..."
curl -s http://localhost:8000/health || echo "Health check failed, but process is running."