entrypoint.sh

#!/bin/bash
set -e

echo "==> Starting Ollama server on port 7860..."
export OLLAMA_HOST=0.0.0.0:7860
export OLLAMA_MODELS=/home/user/.ollama/models

# Start ollama in foreground
ollama serve &
OLLAMA_PID=$!

# Wait for ollama to be ready
echo "==> Waiting for Ollama to be ready..."
MAX_RETRIES=30
COUNT=0
until curl -s http://localhost:7860/api/version > /dev/null 2>&1; do
    COUNT=$((COUNT + 1))
    if [ $COUNT -ge $MAX_RETRIES ]; then
        echo "ERROR: Ollama did not start in time."
        exit 1
    fi
    echo "  ... attempt $COUNT/$MAX_RETRIES"
    sleep 2
done

echo "==> Ollama is ready!"

# Pull model if not cached (fallback in case build layer failed)
if ! ollama list | grep -q "granite4"; then
    echo "==> Pulling granite4:350m..."
    ollama pull granite4:350m
fi

echo "==> Model available:"
ollama list

echo "==> Ollama API running at http://0.0.0.0:7860"
echo "==> Endpoints:"
echo "    POST /api/generate"
echo "    POST /api/chat"
echo "    GET  /api/tags"
echo "    POST /api/embeddings"

# Keep process alive
wait $OLLAMA_PID