start.sh

#!/bin/bash
exec 2>&1

echo "!!! [SYSTEM]: Starting Ollama..."
ollama serve > /dev/null 2>&1 &

until curl -s http://localhost:11434/api/tags > /dev/null; do
    sleep 2
done

# ✅ Router model
if ! ollama list | grep -q "qwen3:0.6b"; then
    echo "!!! [SYSTEM]: Pulling router model qwen3:0.6b..."
    ollama pull qwen3:0.6b
else
    echo "!!! [SYSTEM]: Router model qwen3:0.6b ready ✅"
fi

# ✅ Core model — pull if missing
if ! ollama list | grep -q "qwen3.5:4b"; then
    echo "!!! [SYSTEM]: Pulling core model qwen3.5:4b (first boot, ~2.5GB)..."
    ollama pull qwen3.5:0.8b
    echo "!!! [SYSTEM]: Core model ready ✅"
else
    echo "!!! [SYSTEM]: Core model qwen3.5:4b ready ✅"
fi

# llama.cpp install if needed
if [ "$USE_LLAMA_CPP" = "true" ]; then
    if ! python3 -c "import llama_cpp" 2>/dev/null; then
        echo "!!! [SYSTEM]: Installing llama-cpp-python..."
        pip3 install llama-cpp-python --no-cache-dir --break-system-packages \
            --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu \
            || pip3 install llama-cpp-python --no-cache-dir --break-system-packages
    fi
else
    echo "!!! [SYSTEM]: USE_LLAMA_CPP=false — using Ollama"
fi

echo "!!! [SYSTEM]: Starting Brain App..."
exec python3 -u app.py