#!/usr/bin/env bash
# email-channel supervisor.
#
# Streamable HTTP MCP server on port 18810 (Phase C dual-listener:
# also runs the Microsoft 365 polling loop in the same bun process).
# Bun + restart-on-crash. Locked with flock so multiple invocations are safe.
set -euo pipefail

LOCK_FILE="/tmp/email-channel.lock"
LOG_FILE="/tmp/email-channel.log"
HEALTH_FILE="/tmp/email-channel-health.json"
ENV_FILE="$HOME/.config/email-channel.env"
RESTART_DELAY=3
STABILITY_THRESHOLD=60
SERVER_DIR="$HOME/repos/exult-agent/tools/email-channel"

cleanup() {
  rm -f "$LOCK_FILE" 2>/dev/null || true
}
trap cleanup EXIT

# Single-instance guard
if [ -f "$LOCK_FILE" ]; then
  other=$(cat "$LOCK_FILE" 2>/dev/null || true)
  if [ -n "$other" ] && kill -0 "$other" 2>/dev/null; then
    echo "[$(date)] another instance running (pid $other). Exiting." >&2
    exit 0
  fi
  rm -f "$LOCK_FILE"
fi
echo $$ > "$LOCK_FILE"

log_msg() { echo "[$(date '+%Y-%m-%dT%H:%M:%S%z')] $1" | tee -a "$LOG_FILE"; }

BUN_BIN="$(command -v bun || echo "$HOME/.bun/bin/bun")"
[ -x "$BUN_BIN" ] || { log_msg "ERROR: bun binary not found"; exit 1; }

[ -d "$SERVER_DIR" ] || { log_msg "ERROR: server dir missing: $SERVER_DIR"; exit 1; }

# Load env (MS365 creds, poll interval, MCP_BEARER_TOKEN)
if [ -f "$ENV_FILE" ]; then
  set -a
  # shellcheck disable=SC1090
  source "$ENV_FILE"
  set +a
else
  log_msg "ERROR: env file missing: $ENV_FILE"
  exit 1
fi

# Required env sanity
for k in MS365_CLIENT_ID MS365_CLIENT_SECRET MS365_TENANT_ID MS365_MAILBOX MCP_BEARER_TOKEN; do
  if [ -z "${!k:-}" ]; then
    log_msg "ERROR: $k not set in $ENV_FILE"
    exit 1
  fi
done
: "${EMAIL_POLL_INTERVAL:=20}"
export MS365_CLIENT_ID MS365_CLIENT_SECRET MS365_TENANT_ID MS365_MAILBOX \
       EMAIL_POLL_INTERVAL MCP_BEARER_TOKEN

consecutive_failures=0
last_failure_ts=0
restarts_today=0

backoff() {
  if [ "$consecutive_failures" -ge 5 ]; then echo 60
  elif [ "$consecutive_failures" -ge 3 ]; then echo 15
  else echo "$RESTART_DELAY"; fi
}

while true; do
  start_ts=$(date +%s)
  log_msg "starting bun server (port=18810 failures=$consecutive_failures)"

  "$BUN_BIN" "$SERVER_DIR/server.ts" >>"$LOG_FILE" 2>&1 || true

  end_ts=$(date +%s)
  dur=$((end_ts - start_ts))

  if [ "$dur" -gt "$STABILITY_THRESHOLD" ]; then
    consecutive_failures=0
    log_msg "session ran ${dur}s (stable)"
  else
    if [ "$last_failure_ts" -gt 0 ] && [ $((end_ts - last_failure_ts)) -gt 300 ]; then
      consecutive_failures=1
    else
      consecutive_failures=$((consecutive_failures + 1))
    fi
    last_failure_ts="$end_ts"
    log_msg "session ran ${dur}s (consecutive failure #$consecutive_failures)"
  fi

  restarts_today=$((restarts_today + 1))
  cat >"$HEALTH_FILE" <<EOF
{"restarts_today":$restarts_today,"last_restart":"$(date -u +%FT%TZ)","last_duration_seconds":$dur,"consecutive_failures":$consecutive_failures}
EOF

  delay=$(backoff)
  log_msg "restart in ${delay}s"
  sleep "$delay"
done