#!/usr/bin/env bash
set -euo pipefail

# Konfiguration
HF_HOME="${HF_HOME:-/home/dschlueter/nvme2n1p7_home/huggingface}"
MODEL_REL_PATH="models/qwen3/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q4_K_M.gguf"
IMAGE="ghcr.io/ggml-org/llama.cpp:server-cuda"
CONTAINER_NAME="qwen35b-moe-uncensored"
HOST_PORT=8000
CONTAINER_PORT=8000

echo "[*] Verwende HF_HOME = $HF_HOME"
if [ ! -f "$HF_HOME/$MODEL_REL_PATH" ]; then
  echo "[!] Modell-Datei nicht gefunden: $HF_HOME/$MODEL_REL_PATH" >&2
  exit 1
fi

# Optional: altes gleichnamiges Container-Exemplar stoppen
if docker ps -a --format '{{.Names}}' | grep -q "^${CONTAINER_NAME}\$"; then
  echo "[*] Stoppe existierenden Container $CONTAINER_NAME ..."
  docker rm -f "$CONTAINER_NAME" >/dev/null 2>&1 || true
fi

echo "[*] Starte llama.cpp-Server-Container ($IMAGE) ..."

docker run -d \
  --gpus '"device=1,2"' \
  --name "$CONTAINER_NAME" \
  --restart unless-stopped \
  -e HF_HOME="/hf_home" \
  -v "$HF_HOME:/hf_home:ro" \
  -p "${HOST_PORT}:${CONTAINER_PORT}" \
  "$IMAGE" \
    -m "/hf_home/${MODEL_REL_PATH}" \
    -c 262144 \
    -n 16384 \
    --temp 0.6 --top-p 0.95 --top-k 40 --min-p 0.01 --repeat-penalty 1.05 \
    --main-gpu 0 \
    --tensor-split 0.5,0.5 \
    -ngl 999 \
    -fa on \
    --kv-unified \
    --cache-type-k q8_0 \
    --cache-type-v q8_0 \
    --batch-size 2048 \
    --ubatch-size 512 \
    --parallel 2 \
    --cont-batching \
    --jinja \
    --no-context-shift \
    --host 0.0.0.0 \
    --port "$CONTAINER_PORT"

echo "[*] Container gestartet: $CONTAINER_NAME"
echo "[*] Warte, bis HTTP-Port ${HOST_PORT} antwortet ..."

for i in {1..60}; do
  if curl -s "http://localhost:${HOST_PORT}/" >/dev/null 2>&1; then
    echo "[*] Server antwortet auf http://localhost:${HOST_PORT}/"
    break
  fi
  echo "[*] Warte (${i}/60) ..."
  sleep 2
done

sleep 5

echo "[*] Sende Test-Chat-Request an /v1/chat/completions ..."

RESPONSE="$(curl -s -X POST "http://localhost:${HOST_PORT}/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.6-35b-a3b-moe",
    "messages": [
      { "role": "system", "content": "Du bist ein hilfreicher deutscher Assistent." },
      { "role": "user",   "content": "Gib eine sehr kurze Selbstdiagnose deiner Fähigkeiten." }
    ],
    "max_tokens": 64,
    "temperature": 0.6,
    "stream": false
  }')"

echo
echo "[*] Antwort vom Server:"
echo "$RESPONSE"

echo
echo "[*] Zum Stoppen des Servers:"
echo "    docker rm -f $CONTAINER_NAME"