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

export PORT="${PORT:-7860}"
export API_PORT="${API_PORT:-7860}"
export DATABASE_URL="${DATABASE_URL:-postgresql+psycopg://docqa:docqa@127.0.0.1:5432/docqa}"
export REDIS_URL="${REDIS_URL:-redis://127.0.0.1:6379/0}"

# Force local Ollama mode in Hugging Face Space.
export OPENAI_BASE_URL="${OPENAI_BASE_URL:-}"
export OPENAI_API_KEY="${OPENAI_API_KEY:-}"
export OPENAI_MODEL="${OPENAI_MODEL:-}"

export LOCAL_LLM_BASE_URL="${LOCAL_LLM_BASE_URL:-http://127.0.0.1:11434/v1}"
export LOCAL_LLM_API_KEY="${LOCAL_LLM_API_KEY:-ollama}"
export LOCAL_LLM_MODEL="${LOCAL_LLM_MODEL:-qwen3:0.6b}"

export OLLAMA_HOST="${OLLAMA_HOST:-127.0.0.1:11434}"
export OLLAMA_MODELS="${OLLAMA_MODELS:-/opt/ollama/models}"
export OLLAMA_KEEP_ALIVE="${OLLAMA_KEEP_ALIVE:-5m}"
export OLLAMA_NUM_PARALLEL="${OLLAMA_NUM_PARALLEL:-1}"
export OPENAI_TIMEOUT_SECONDS="${OPENAI_TIMEOUT_SECONDS:-360}"
export LLM_TRACE_ENABLED="${LLM_TRACE_ENABLED:-true}"
export LLM_TRACE_MAX_CHARS="${LLM_TRACE_MAX_CHARS:-20000}"
export CHUNK_VERIFICATION_TRACE_ENABLED="${CHUNK_VERIFICATION_TRACE_ENABLED:-true}"
export CHUNK_VERIFICATION_MAX_WINDOWS="${CHUNK_VERIFICATION_MAX_WINDOWS:-4}"

export EMBEDDING_MODEL="${EMBEDDING_MODEL:-infgrad/Jasper-Token-Compression-600M}"
export EMBEDDING_DEVICE="${EMBEDDING_DEVICE:-cpu}"
export EMBEDDING_BATCH_SIZE="${EMBEDDING_BATCH_SIZE:-2}"
export EMBEDDING_TRUST_REMOTE_CODE="${EMBEDDING_TRUST_REMOTE_CODE:-true}"

# Avoid duplicate migrations because this script runs alembic explicitly.
export RUN_MIGRATIONS_ON_START=false

cleanup() {
  echo "Stopping background services..."
  jobs -p | xargs -r kill || true
}
trap cleanup EXIT

POSTGRES_VERSION="$(ls /etc/postgresql | sort -V | tail -n 1)"

echo "Starting PostgreSQL..."
pg_ctlcluster "${POSTGRES_VERSION}" main start || true

su postgres -c "psql -tc \"SELECT 1 FROM pg_roles WHERE rolname='docqa'\" | grep -q 1 || psql -c \"CREATE USER docqa WITH PASSWORD 'docqa';\""
su postgres -c "psql -tc \"SELECT 1 FROM pg_database WHERE datname='docqa'\" | grep -q 1 || createdb -O docqa docqa"
su postgres -c "psql -c \"ALTER USER postgres WITH PASSWORD 'postgres';\""
su postgres -c "psql -c \"ALTER USER docqa WITH PASSWORD 'docqa';\""
su postgres -c "psql -c \"GRANT ALL PRIVILEGES ON DATABASE docqa TO docqa;\""

echo "Starting Redis..."
redis-server --daemonize yes --bind 127.0.0.1 --port 6379 --save '' --appendonly no

echo "Starting Ollama..."
ollama serve > /tmp/ollama.log 2>&1 &
OLLAMA_PID=$!

echo "Waiting for Ollama API..."
for i in $(seq 1 120); do
  if curl -fsS "http://127.0.0.1:11434/api/tags" >/dev/null 2>&1; then
    echo "Ollama API is ready."
    break
  fi

  if ! kill -0 "$OLLAMA_PID" >/dev/null 2>&1; then
    echo "Ollama exited early. Logs:"
    cat /tmp/ollama.log || true
    exit 1
  fi

  sleep 1
done

if ! curl -fsS "http://127.0.0.1:11434/api/tags" >/dev/null 2>&1; then
  echo "Ollama did not become ready. Logs:"
  cat /tmp/ollama.log || true
  exit 1
fi

if ! ollama list | awk 'NR > 1 {print $1}' | grep -Fx "${LOCAL_LLM_MODEL}" >/dev/null 2>&1; then
  echo "Model ${LOCAL_LLM_MODEL} not found; pulling at runtime..."
  ollama pull "${LOCAL_LLM_MODEL}"
fi

echo "Testing Ollama OpenAI-compatible endpoint..."
curl -fsS "http://127.0.0.1:11434/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"${LOCAL_LLM_MODEL}\",
    \"messages\": [{\"role\": \"user\", \"content\": \"Reply exactly: OLLAMA_OK /no_think\"}],
    \"max_tokens\": 32,
    \"stream\": false
  }" >/tmp/ollama-smoke.json || {
    echo "Ollama smoke test failed."
    cat /tmp/ollama.log || true
    cat /tmp/ollama-smoke.json || true
    exit 1
  }
tail -n +1 -F /tmp/ollama.log | sed -u 's/^/[ollama] /' &
OLLAMA_LOG_TAIL_PID=$!

echo "Ollama smoke test passed."

echo "Running database migrations..."
alembic upgrade head

echo "Starting Celery worker..."
./scripts/start-worker.sh &

echo "Starting FastAPI..."
exec ./scripts/start-api.sh