#!/usr/bin/env bash
# DeepSeek-V4.1-Flash-1017 lokal ausprobieren – macOS
# Automatisch erzeugt von Tokenstrom (tokenstrom.com).
# Aufruf: bash try_mac.sh            # Ollama (Standard)
#        MODE=hf bash try_mac.sh    # Python/transformers
set -e
cd "$(dirname "$0")"
echo "== DeepSeek-V4.1-Flash-1017 =="
echo "WARNUNG: Huge (763,2 B parameters): locally, practically only on servers with multiple GPUs. Colab Free isn't enough."

if [ "${MODE:-ollama}" = "ollama" ]; then
  echo "== Ollama =="
  if ! command -v ollama >/dev/null 2>&1; then
    if command -v brew >/dev/null 2>&1; then brew install ollama; else
      echo "Please install the Ollama app first: https://ollama.com/download/mac"; exit 1; fi
  fi
  curl -s http://localhost:11434/api/tags >/dev/null 2>&1 || (ollama serve >/dev/null 2>&1 &)
  sleep 3
  MODEL="${MODEL:-}"
  MODEL="${MODEL:-hf.co/nazeshinjite/DeepSeek-V4.1-Flash-1017-ds4-GGUF}"
  echo "Downloading $MODEL (the first time this takes a few minutes, depending on size) ..."
  ollama pull "$MODEL"
  echo ""
  echo "Done! Now type your question and press Enter. Quit with /bye"
  exec ollama run "$MODEL"
fi

echo "== Python / Hugging Face =="
python3 -m venv .venv-deepseek-v4-1-flash-1017
source .venv-deepseek-v4-1-flash-1017/bin/activate
pip install -q -U pip
pip install -q -U torch transformers accelerate pillow scipy

python3 demo.py