#!/usr/bin/env bash
# Start Qwen3-Embedding-8B embedding service via sentence-transformers FastAPI on port 8200
# API: http://localhost:8200/v1/embeddings

set -euo pipefail

PORT=8200
LOG_FILE="/lsinfo/ai/hellotax_ai/llm_service/logs/embedding-${PORT}.log"
PID_FILE="/lsinfo/ai/hellotax_ai/llm_service/logs/embedding-${PORT}.pid"

mkdir -p "$(dirname "$LOG_FILE")"

# Check if already running
if [[ -f "$PID_FILE" ]] && kill -0 "$(cat "$PID_FILE")" 2>/dev/null; then
  echo "Embedding service already running (PID $(cat "$PID_FILE"))"
  exit 0
fi

echo "Starting sentence-transformers embedding server (Qwen3-Embedding-8B) on port ${PORT}..."

export EMBEDDING_MODEL_PATH=/lsinfo/ai/hellotax_ai/llm_service/base_models/Qwen3-Embedding-8B
export SERVED_MODEL_NAME=Qwen3-Embedding-8B
export CUDA_VISIBLE_DEVICES=0

nohup /lsinfo/ai/hellotax_ai/llm_service/venv_embed/bin/python \
  /lsinfo/ai/hellotax_ai/llm_service/servers/embedding_server.py \
  --host 0.0.0.0 --port "${PORT}" \
  > "${LOG_FILE}" 2>&1 &

echo $! > "${PID_FILE}"
echo "PID $(cat "$PID_FILE") — log: ${LOG_FILE}"

# Wait for service to be ready
echo -n "Waiting for embedding service to be ready"
for i in $(seq 1 60); do
  if curl -sf http://localhost:${PORT}/health > /dev/null 2>&1; then
    echo " ready."
    exit 0
  fi
  echo -n "."
  sleep 3
done
echo " timeout. Check log: ${LOG_FILE}"
exit 1