gandalf/app.py

"""Gandalf: Global Advanced Network Detection And Link Facilitator.

Flask web application serving the monitoring dashboard and suppression
management UI. Authentication via Authelia forward-auth headers.
All monitoring and alerting is handled by the separate monitor.py daemon.
"""
import hashlib
import html
import ipaddress
import json
import logging
import os
import re
import tempfile
import threading
import time
import uuid
from datetime import datetime, timezone
from functools import wraps
from typing import Optional
from urllib.parse import quote, urlparse

from flask import Flask, jsonify, make_response, redirect, render_template, request, send_file, url_for

import cutsheet
import db
import devices
import diagnose
import topology
from monitor import PulseClient, UnifiClient

logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s %(levelname)s %(name)s %(message)s',
)
logger = logging.getLogger('gandalf.web')

app = Flask(__name__)

_AVATAR_COLORS = ['lt-avatar--orange', 'lt-avatar--green', 'lt-avatar--purple', '']


@app.template_filter('avatar_color')
def avatar_color_filter(name: str) -> str:
    return _AVATAR_COLORS[int(hashlib.md5(name.encode()).hexdigest(), 16) % len(_AVATAR_COLORS)]  # nosec B324

_cfg = None
_cfg_lock = threading.Lock()


@app.context_processor
def inject_config():
    """Inject safe config values into all templates."""
    cfg = _config()
    return {
        'config': {
            'ticket_api': {
                'web_url': cfg.get('ticket_api', {}).get('web_url', 'http://t.lotusguild.org/ticket/'),
            }
        }
    }


@app.context_processor
def inject_maintenance():
    """The active global suppression (maintenance), shown as a banner on every page."""
    try:
        active = [s for s in db.get_active_suppressions() if s.get('target_type') == 'all']
    except Exception as e:
        logger.error(f'Reading maintenance state failed: {e}')
        active = []
    # Of several, the one that ends last decides when tickets resume; no expiry beats any
    active.sort(key=lambda s: (s.get('expires_at') is None, s.get('expires_at') or ''))
    return {'maintenance': active[-1] if active else None}


# In-memory diagnostic job store { job_id: { status, result, created_at } }
_diag_jobs: dict = {}
_diag_lock = threading.Lock()
# Per-user rate-limit: { username: [epoch_float, ...] }: cleaned inside _diag_lock
_diag_rate: dict = {}


def _purge_old_jobs_loop():
    """Background thread: remove stale diagnostic jobs and mark stuck ones done."""
    while True:
        time.sleep(120)
        cutoff = time.time() - 600
        stuck_cutoff = time.time() - 300  # 5 min: job still 'running' → thread must have crashed
        with _diag_lock:
            stale = [jid for jid, j in _diag_jobs.items() if j.get('created_at', 0) < cutoff]
            for jid in stale:
                del _diag_jobs[jid]
            for jid, j in list(_diag_jobs.items()):
                if j['status'] == 'running' and j.get('created_at', 0) < stuck_cutoff:
                    j['status'] = 'done'
                    j['result'] = {'status': 'error', 'error': 'Diagnostic abandoned: no activity for 5 minutes.'}
                    logger.error(f'Diagnostic job {jid} stuck (no activity for 5 min); marked done/error')


_purge_thread = threading.Thread(target=_purge_old_jobs_loop, daemon=True)
_purge_thread.start()


def _config() -> dict:
    global _cfg
    if _cfg is None:
        with _cfg_lock:
            if _cfg is None:
                with open('config.json') as f:
                    _cfg = json.load(f)
    return _cfg


@app.after_request
def add_security_headers(response):
    response.headers.setdefault('X-Content-Type-Options', 'nosniff')
    response.headers.setdefault('X-Frame-Options', 'DENY')
    response.headers.setdefault('Referrer-Policy', 'strict-origin-when-cross-origin')
    return response


def _monitor_stale_after() -> int:
    """Seconds without a completed check before the monitor counts as stalled.

    Five missed polls, and never under 10 minutes: one slow cycle (an SSH timeout
    to a dead host) shouldn't trip it, but a hung monitor is caught in minutes.
    """
    try:
        poll = int(_config().get('monitor', {}).get('poll_interval', 120))
    except (TypeError, ValueError):
        poll = 120
    return max(600, poll * 5)


def _monitor_age(last_check: str) -> Optional[float]:
    """Seconds since the monitor's last check, or None if it never ran or can't be read."""
    if not last_check or last_check == 'Never':
        return None
    try:
        ts = datetime.strptime(last_check, '%Y-%m-%d %H:%M:%S UTC').replace(tzinfo=timezone.utc)
    except ValueError:
        return None
    return (datetime.now(timezone.utc) - ts).total_seconds()


def _daemon_ok(last_check: str) -> bool:
    """True if the monitor checked recently enough (see _monitor_stale_after)."""
    age = _monitor_age(last_check)
    return age is not None and age < _monitor_stale_after()


# ---------------------------------------------------------------------------
# Auth helpers
# ---------------------------------------------------------------------------

def _request_source() -> str:
    """The address the request came from.

    Gunicorn listens on loopback behind the local nginx, which overwrites X-Real-IP
    with the real peer address. A peer that isn't loopback reached gunicorn directly,
    so its own address is the source and any X-Real-IP it sent is ignored.
    """
    peer = request.remote_addr or ''
    try:
        if ipaddress.ip_address(peer).is_loopback:
            return request.headers.get('X-Real-IP', '').strip()
    except ValueError:
        return ''
    return peer


def _from_trusted_proxy() -> bool:
    """True when the request came through the Authelia-protected proxy.

    Only then are the Remote-* identity headers real: anyone else on the LAN could
    set them. Fails closed: with no auth.trusted_proxies configured, nobody is trusted.
    """
    try:
        source = ipaddress.ip_address(_request_source())
    except ValueError:
        return False
    for entry in _config().get('auth', {}).get('trusted_proxies', []):
        try:
            if source in ipaddress.ip_network(entry, strict=False):
                return True
        except ValueError:
            logger.error(f'Ignoring invalid auth.trusted_proxies entry {entry!r}')
    return False


def _get_user() -> dict:
    if not _from_trusted_proxy():
        return {'username': '', 'name': '', 'email': '', 'groups': []}
    return {
        'username': request.headers.get('Remote-User', ''),
        'name': request.headers.get('Remote-Name', ''),
        'email': request.headers.get('Remote-Email', ''),
        'groups': [
            g.strip()
            for g in request.headers.get('Remote-Groups', '').split(',')
            if g.strip()
        ],
    }


def require_auth(f):
    @wraps(f)
    def wrapper(*args, **kwargs):
        user = _get_user()
        if not user['username']:
            if request.headers.get('Remote-User'):
                logger.warning(
                    f'Ignored identity headers from untrusted source {_request_source() or "unknown"!r}; '
                    f'only auth.trusted_proxies may send them'
                )
            return (
                '<h1>401: Not authenticated</h1>'
                '<p>Please access Gandalf through '
                '<a href="https://auth.lotusguild.org">auth.lotusguild.org</a>.</p>',
                401,
            )
        allowed = _config().get('auth', {}).get('allowed_groups', ['admin'])
        if not any(g in allowed for g in user['groups']):
            safe_user   = html.escape(user['username'])
            safe_groups = html.escape(', '.join(allowed))
            return (
                f'<h1>403: Access denied</h1>'
                f'<p>Your account ({safe_user}) is not in an allowed group '
                f'({safe_groups}).</p>',
                403,
            )
        return f(*args, **kwargs)
    return wrapper


def require_admin(f):
    """Decorator: require require_auth AND membership in the 'admin' group."""
    @wraps(f)
    def wrapper(*args, **kwargs):
        user = _get_user()
        if 'admin' not in user.get('groups', []):
            return jsonify({'error': 'Admin access required'}), 403
        return f(*args, **kwargs)
    return wrapper


# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------

_PAGE_LIMIT = 200  # max events returned per request


def _annotate_suppressions(events: list, suppressions: list) -> None:
    """Annotate each event dict in-place with an is_suppressed bool.

    Mirrors the suppression check order in monitor.py exactly:
      interface_down  → interface OR host
      unifi_device_*  → unifi_device
      everything else → host
    """
    for ev in events:
        etype = ev.get('event_type', '')
        name = ev.get('target_name', '')
        detail = ev.get('target_detail', '') or ''
        if etype == 'interface_down':
            ev['is_suppressed'] = (
                db.check_suppressed(suppressions, 'interface', name, detail) or
                db.check_suppressed(suppressions, 'host', name)
            )
        elif etype == 'unifi_device_offline':
            ev['is_suppressed'] = db.check_suppressed(suppressions, 'unifi_device', name, detail)
        else:
            ev['is_suppressed'] = db.check_suppressed(suppressions, 'host', name, detail)


# The monitor refreshes the open ticket list every 15 minutes; older than this and
# it has stopped working, so nothing is flagged rather than flagging on stale data.
_OPEN_TICKETS_MAX_AGE_S = 2 * 3600


def _stale_settings() -> tuple:
    """(days an alert can stay active before it counts as stale,
        hours a ticket can stay open after its alert cleared)."""
    mon = _config().get('monitor', {})
    try:
        days = float(mon.get('stale_alert_days', 3))
        hours = int(mon.get('stale_ticket_hours', 24))
    except (TypeError, ValueError):
        days, hours = 3.0, 24
    return days, hours


def _open_tickets() -> Optional[dict]:
    """{ticket_id: status} of open Tinker tickets, or None when the list is missing or out of date."""
    try:
        raw = db.get_state('open_tickets')
        data = json.loads(raw) if raw else None
        if not data or time.time() - float(data.get('checked_at', 0)) > _OPEN_TICKETS_MAX_AGE_S:
            return None
        return {str(k): v for k, v in (data.get('tickets') or {}).items()}
    except Exception as e:
        logger.error(f'Reading open ticket list failed: {e}')
        return None


def _annotate_staleness(events: list, open_tickets: Optional[dict]) -> None:
    """Add stale_days (set when an alert has been active too long) and ticket_closed
    (its ticket was closed in Tinker while the alert is still active) to each event."""
    days, _ = _stale_settings()
    now = datetime.now(timezone.utc)
    for ev in events:
        ev['stale_days'] = None
        ev['ticket_closed'] = False
        try:
            first = datetime.fromisoformat(str(ev.get('first_seen', '')).replace('Z', '+00:00'))
            if first.tzinfo is None:
                first = first.replace(tzinfo=timezone.utc)
            age = (now - first).total_seconds() / 86400
            if age >= days:
                ev['stale_days'] = int(age)
        except ValueError:
            pass
        tid = ev.get('ticket_id')
        if tid and open_tickets is not None and str(tid) not in open_tickets:
            ev['ticket_closed'] = True


def _stale_view(events: list) -> list:
    """Flag stale active alerts in place; return tickets left open after their alerts cleared."""
    open_tickets = _open_tickets()
    _annotate_staleness(events, open_tickets)
    if open_tickets is None:
        return []
    _, hours = _stale_settings()
    try:
        return db.get_cleared_tickets(list(open_tickets), older_than_hours=hours)
    except Exception as e:
        logger.error(f'Finding tickets left open failed: {e}')
        return []


def _state_json(key: str) -> dict:
    """Return a monitor_state JSON blob as a dict, or {} if missing or unreadable."""
    raw = db.get_state(key)
    if not raw:
        return {}
    try:
        return json.loads(raw)
    except Exception as e:
        logger.error(f'Failed to parse {key} JSON: {e}')
        return {}


# ---------------------------------------------------------------------------
# Page routes
# ---------------------------------------------------------------------------


@app.route('/')
@require_auth
def index():
    user = _get_user()
    events = db.get_active_events(limit=_PAGE_LIMIT)
    total_active = db.count_active_events()
    summary = db.get_status_summary()
    snapshot_raw = db.get_state('network_snapshot')
    last_check = db.get_state('last_check', 'Never')
    try:
        snapshot = json.loads(snapshot_raw) if snapshot_raw else {}
    except Exception as e:
        logger.error(f'Failed to parse network_snapshot JSON: {e}')
        snapshot = {}
    suppressions = db.get_active_suppressions()
    _annotate_suppressions(events, suppressions)
    stale_tickets = _stale_view(events)
    recent_resolved = db.get_recent_resolved(hours=24, limit=10)
    return render_template(
        'index.html',
        user=user,
        events=events,
        total_active=total_active,
        summary=summary,
        snapshot=snapshot,
        last_check=last_check,
        suppressions=suppressions,
        recent_resolved=recent_resolved,
        stale_tickets=stale_tickets,
        stale_alerts=sum(1 for e in events if e.get('stale_days') is not None or e.get('ticket_closed')),
        daemon_ok=_daemon_ok(last_check),
    )


@app.route('/links')
@require_auth
def links_page():
    """Link Debug moved to the per-device pages; keep old bookmarks working."""
    return redirect(url_for('index'))


@app.route('/device/<path:name>')
@require_auth
def device_page(name: str):
    user = _get_user()
    return render_template('device.html', user=user, device_name=name, can_edit='admin' in user['groups'])


@app.route('/inspector')
@require_auth
def inspector():
    """The port inspector moved to each switch's device page; keep old bookmarks working."""
    return redirect(url_for('index'))


@app.route('/events')
@require_auth
def events_page():
    return render_template('events.html', user=_get_user())


_HISTORY_RANGES = {'24h': 24, '7d': 24 * 7, '30d': 24 * 30, '90d': 24 * 90}
_HISTORY_PAGE = 100


@app.route('/api/events/history')
@require_auth
def api_events_history():
    """Every alert open at some point in a time window, with summaries.

    Query: range (24h|7d|30d|90d), type, target (substring of name or detail),
    severity, status (all|active|resolved), offset.
    """
    args = request.args
    hours = _HISTORY_RANGES.get(args.get('range', '7d'))
    if hours is None:
        return jsonify({'error': f'range must be one of {", ".join(_HISTORY_RANGES)}'}), 400
    severity = args.get('severity', '')
    if severity not in ('', 'critical', 'warning', 'info'):
        return jsonify({'error': 'severity must be critical, warning or info'}), 400
    status = args.get('status', 'all')
    if status not in ('all', 'active', 'resolved'):
        return jsonify({'error': 'status must be all, active or resolved'}), 400
    try:
        offset = max(int(args.get('offset', 0)), 0)
    except ValueError:
        return jsonify({'error': 'offset must be an integer'}), 400
    event_type = args.get('type', '')[:60]
    target = args.get('target', '').strip()[:255]

    result = db.get_events_history(hours, event_type, target, severity, status,
                                   limit=_HISTORY_PAGE, offset=offset)
    result['flappers'] = db.get_flappers(hours, event_type, target, severity)
    result['types'] = db.get_event_types()
    result['offset'] = offset
    result['page_size'] = _HISTORY_PAGE
    return jsonify(result)


@app.route('/suppressions')
@require_auth
@require_admin
def suppressions_page():
    user = _get_user()
    active = db.get_active_suppressions()
    history = db.get_suppression_history(limit=50)
    snapshot_raw = db.get_state('network_snapshot')
    try:
        snapshot = json.loads(snapshot_raw) if snapshot_raw else {}
    except Exception as e:
        logger.error(f'Failed to parse network_snapshot JSON: {e}')
        snapshot = {}
    return render_template(
        'suppressions.html',
        user=user,
        active=active,
        history=history,
        snapshot=snapshot,
    )


# ---------------------------------------------------------------------------
# API routes
# ---------------------------------------------------------------------------

@app.route('/api/status')
@require_auth
def api_status():
    active = db.get_active_events(limit=_PAGE_LIMIT)
    suppressions = db.get_active_suppressions()
    _annotate_suppressions(active, suppressions)
    stale_tickets = _stale_view(active)
    last_check = db.get_state('last_check', 'Never')
    return jsonify({
        'summary': db.get_status_summary(),
        'last_check': last_check,
        'events': active,
        'stale_tickets': stale_tickets,
        'total_active': db.count_active_events(),
        'daemon_ok': _daemon_ok(last_check),
    })


@app.route('/api/network')
@require_auth
def api_network():
    raw = db.get_state('network_snapshot')
    if raw:
        try:
            return jsonify(json.loads(raw))
        except Exception:
            logger.error('Failed to parse network_snapshot JSON')
    return jsonify({'hosts': {}, 'unifi': [], 'updated': None})


@app.route('/api/links')
@require_auth
def api_links():
    raw = db.get_state('link_stats')
    if raw:
        if len(raw) > 10_000_000:
            logger.error(f'link_stats exceeds 10 MB ({len(raw)} bytes); possible corruption')
            return jsonify({'error': 'Invalid cached data'}), 503
        try:
            return jsonify(json.loads(raw))
        except Exception as e:
            logger.error(f'Failed to parse link_stats JSON: {e}')
    return jsonify({'hosts': {}, 'unifi_switches': {}, 'updated': None})


@app.route('/api/modem')
@require_auth
def api_modem():
    """Latest WAN modem status plus optical history for the trend chart."""
    try:
        hours = max(1, min(int(request.args.get('hours', 168)), 2160))
    except ValueError:
        return jsonify({'error': 'hours must be an integer'}), 400

    raw = db.get_state('modem_stats')
    if not raw:
        return jsonify({'configured': False})
    try:
        data = json.loads(raw)
    except Exception as e:
        logger.error(f'Failed to parse modem_stats JSON: {e}')
        return jsonify({'configured': False})

    # A history failure shouldn't take the current readings down with it.
    try:
        # Keep roughly 200 points regardless of range; never finer than the 5 min poll.
        data['history'] = db.get_modem_history(hours, max(300, hours * 3600 // 200))
    except Exception as e:
        logger.error(f'Failed to load modem history: {e}')
        data['history'] = []
    data['wan_history'] = _state_json('wan_history')
    data['configured'] = True
    return jsonify(data)


@app.route('/api/devices')
@require_auth
def api_devices():
    """Device list for the home page search, plus cutsheet panels, jacks and cables."""
    found = devices.build_index(
        _state_json('network_snapshot'), _state_json('link_stats'), _state_json('modem_stats'),
    )
    known = {d['name'] for d in found}
    try:
        _ensure_cutsheet_schema()
        data = db.cutsheet_load()
    except Exception as e:
        logger.error(f'Cutsheet unavailable for search: {e}')
        data = {'items': [], 'cables': []}
    for item in data['items']:
        if item['name'] in known:
            continue
        kind = item['kind'].replace('_', ' ')
        found.append({
            'name': item['name'], 'kind': kind, 'status': 'unknown', 'ip': '', 'mac': '', 'model': '',
            'url': f'/cutsheet#item-{item["id"]}', 'interface_count': item.get('port_count') or 0,
            'search': ' '.join(filter(None, [item['name'], kind, item.get('location'), 'cutsheet'])).lower(),
        })
    for cable in data['cables']:
        if not cable.get('label'):
            continue
        ends = ' '.join(filter(None, [cable['a_device'], cable['b_device']]))
        found.append({
            'name': cable['label'], 'kind': 'cable', 'status': 'unknown', 'ip': '', 'mac': '',
            'model': f'{cable["a_device"]} ↔ {cable["b_device"] or "unterminated"}',
            'url': f'/cutsheet#cable-{cable["id"]}', 'interface_count': 0,
            'search': ' '.join(filter(None, [cable['label'], ends, cable.get('cable_type'),
                                             cable.get('color'), 'cable cutsheet'])).lower(),
        })
    return jsonify({'devices': found})


@app.route('/api/device/<path:name>')
@require_auth
def api_device(name: str):
    """Everything known about one device, for its detail page."""
    detail = devices.device_detail(
        name, _state_json('network_snapshot'), _state_json('link_stats'), _state_json('modem_stats'),
    )
    if detail is None:
        return jsonify({'error': f'No device named "{name}"'}), 404
    events = db.get_events_for_target(name)
    _annotate_suppressions(events, db.get_active_suppressions())
    detail['events'] = events
    try:
        detail['cutsheet'] = _cutsheet_for_device(name)
    except Exception as e:
        logger.error(f'Cutsheet unavailable for {name}: {e}')
        detail['cutsheet'] = []
    detail['inventory'] = _inventory_for(name, detail.get('model') or '')
    try:
        _ensure_cutsheet_schema()
        detail['unused_interfaces'] = db.get_unused_interfaces().get(name, {})
    except Exception as e:
        logger.error(f'Unused interfaces unavailable for {name}: {e}')
        detail['unused_interfaces'] = {}
    return jsonify(detail)


@app.route('/api/events')
@require_auth
def api_events():
    try:
        limit = min(int(request.args.get('limit', _PAGE_LIMIT)), 1000)
        offset = max(int(request.args.get('offset', 0)), 0)
    except ValueError:
        return jsonify({'error': 'limit and offset must be integers'}), 400
    status_filter = request.args.get('status', 'active')
    if status_filter not in ('active', 'resolved', 'all'):
        return jsonify({'error': 'status must be active, resolved, or all'}), 400

    result: dict = {}
    if status_filter in ('active', 'all'):
        result['active'] = db.get_active_events(limit=limit, offset=offset)
        result['total_active'] = db.count_active_events()
    if status_filter in ('resolved', 'all'):
        result['resolved'] = db.get_recent_resolved(hours=24, limit=30)
    return jsonify(result)


@app.route('/api/suppressions', methods=['GET'])
@require_auth
def api_get_suppressions():
    return jsonify(db.get_active_suppressions())


@app.route('/api/suppressions', methods=['POST'])
@require_auth
@require_admin
def api_create_suppression():
    user = _get_user()
    data = request.get_json(silent=True) or {}

    target_type = data.get('target_type', 'host')
    target_name = (data.get('target_name') or '').strip()
    target_detail = (data.get('target_detail') or '').strip()
    reason = (data.get('reason') or '').strip()
    expires_minutes = data.get('expires_minutes')  # None = manual/permanent

    if target_type not in ('host', 'interface', 'unifi_device', 'all'):
        return jsonify({'error': 'Invalid target_type'}), 400
    if target_type != 'all' and not target_name:
        return jsonify({'error': 'target_name required'}), 400
    if not reason:
        return jsonify({'error': 'reason required'}), 400
    if len(reason) > 500:
        return jsonify({'error': 'reason must be 500 characters or fewer'}), 400
    if len(target_name) > 255:
        return jsonify({'error': 'target_name must be 255 characters or fewer'}), 400
    if len(target_detail) > 255:
        return jsonify({'error': 'target_detail must be 255 characters or fewer'}), 400

    if expires_minutes is not None:
        try:
            expires_minutes = int(expires_minutes)
            if expires_minutes <= 0 or expires_minutes > 43200:
                return jsonify({'error': 'expires_minutes must be between 1 and 43200 (30 days)'}), 400
        except (ValueError, TypeError):
            return jsonify({'error': 'expires_minutes must be a valid integer'}), 400

    sup_id = db.create_suppression(
        target_type=target_type,
        target_name=target_name,
        target_detail=target_detail,
        reason=reason,
        suppressed_by=user['username'],
        expires_minutes=expires_minutes,
    )
    logger.info(
        f'Suppression #{sup_id} created by {user["username"]}: '
        f'{target_type}/{target_name}/{target_detail}: {reason}'
    )
    return jsonify({'success': True, 'id': sup_id})


@app.route('/api/suppressions/<int:sup_id>', methods=['DELETE'])
@require_auth
@require_admin
def api_delete_suppression(sup_id: int):
    user = _get_user()
    db.deactivate_suppression(sup_id)
    logger.info(f'Suppression #{sup_id} removed by {user["username"]}')
    return jsonify({'success': True})


@app.route('/api/diagnose', methods=['POST'])
@require_auth
def api_diagnose_start():
    """Start a link diagnostic job. Returns {job_id}."""
    data = request.get_json(silent=True) or {}
    switch_name = (data.get('switch_name') or '').strip()
    try:
        port_idx = int(data.get('port_idx'))
    except (TypeError, ValueError):
        return jsonify({'error': 'port_idx must be an integer'}), 400

    if not switch_name:
        return jsonify({'error': 'switch_name and port_idx required'}), 400

    # Look up switch + port in cached link_stats
    raw = db.get_state('link_stats')
    if not raw:
        return jsonify({'error': 'No link_stats data available'}), 503
    try:
        link_data = json.loads(raw)
    except Exception as e:
        logger.error(f'Failed to parse link_stats JSON in /api/diagnose: {e}')
        return jsonify({'error': 'Internal data error'}), 500

    switches = link_data.get('unifi_switches', {})
    sw = switches.get(switch_name)
    if not sw:
        return jsonify({'error': f'Switch "{switch_name}" not found'}), 404

    # Find port by port_idx
    port_data = None
    for pname, pd in sw.get('ports', {}).items():
        if pd.get('port_idx') == port_idx:
            port_data = dict(pd)
            port_data['name'] = pname
            port_data['switch_name'] = switch_name
            break
    if not port_data:
        return jsonify({'error': f'Port {port_idx} not found on switch "{switch_name}"'}), 404

    # LLDP neighbor required to know which host+iface to SSH into
    lldp = port_data.get('lldp')
    if not lldp or not lldp.get('system_name'):
        return jsonify({'error': 'No LLDP neighbor data for this port'}), 400

    server_name = lldp['system_name']
    if not re.fullmatch(r'[a-zA-Z0-9._-]+', server_name):
        logger.error(f'Refusing diagnostic: invalid server_name from LLDP: {server_name!r}')
        return jsonify({'error': 'LLDP neighbor name contains invalid characters'}), 400
    lldp_port_id = lldp.get('port_id', '')

    # Find matching host + interface in link_stats hosts
    hosts = link_data.get('hosts', {})
    server_ifaces = hosts.get(server_name)
    if not server_ifaces:
        return jsonify({'error': f'Host "{server_name}" not in link stats'}), 404

    # Match interface by LLDP port_id (exact then fuzzy)
    matched_iface = None
    if lldp_port_id and lldp_port_id in server_ifaces:
        matched_iface = lldp_port_id
    if not matched_iface and lldp_port_id:
        matched_iface = next(
            (k for k in server_ifaces if lldp_port_id in k or k in lldp_port_id),
            None
        )
    if not matched_iface:
        matched_iface = next(iter(server_ifaces), None)
    if not matched_iface:
        return jsonify({'error': 'Cannot determine server interface'}), 400

    # Resolve host IP from link_stats host data
    host_ip = (server_ifaces.get(matched_iface) or {}).get('host_ip')
    if not host_ip:
        # Fallback: use first valid IP from LLDP mgmt IPs
        for candidate in (lldp.get('mgmt_ips') or []):
            try:
                ipaddress.ip_address(candidate)
                host_ip = candidate
                break
            except ValueError:
                continue
    if not host_ip:
        return jsonify({'error': 'Cannot determine host IP for SSH'}), 400

    # Validate resolved values before passing to SSH command builder
    try:
        ipaddress.ip_address(host_ip)
    except ValueError:
        logger.error(f'Refusing diagnostic: invalid host_ip "{host_ip}" for {server_name}')
        return jsonify({'error': 'Resolved host IP is not a valid IP address'}), 400
    if not re.fullmatch(r'[a-zA-Z0-9._-]+', matched_iface):
        logger.error(f'Refusing diagnostic: invalid iface "{matched_iface}" for {server_name}')
        return jsonify({'error': 'Resolved interface name contains invalid characters'}), 400

    job_id = str(uuid.uuid4())
    requesting_user = _get_user()['username']
    now = time.time()
    with _diag_lock:
        # Rate limit: max 5 diagnostic jobs per user per minute; prune stale user entries
        stale_users = [u for u, ts in _diag_rate.items() if not ts or max(ts) < now - 3600]
        for u in stale_users:
            del _diag_rate[u]
        recent = [t for t in _diag_rate.get(requesting_user, []) if now - t < 60]
        if len(recent) >= 5:
            return jsonify({'error': 'Rate limit exceeded: max 5 diagnostics per minute'}), 429
        recent.append(now)
        _diag_rate[requesting_user] = recent
        _diag_jobs[job_id] = {
            'status': 'running', 'result': None,
            'created_at': now, 'user': requesting_user,
        }

    def _run():
        try:
            cfg = _config()
            pulse = PulseClient(cfg)
            runner = diagnose.DiagnosticsRunner(pulse)
            result = runner.run(host_ip, server_name, matched_iface, port_data)
        except Exception as e:
            logger.error(f'Diagnostic job {job_id} failed: {e}', exc_info=True)
            result = {'status': 'error', 'error': 'Diagnostic failed; check server logs.'}
        with _diag_lock:
            if job_id in _diag_jobs:
                _diag_jobs[job_id]['status'] = 'done'
                _diag_jobs[job_id]['result'] = result

    t = threading.Thread(target=_run, daemon=True)
    t.start()

    return jsonify({'job_id': job_id})


@app.route('/api/diagnose/<job_id>', methods=['GET'])
@require_auth
def api_diagnose_poll(job_id: str):
    """Poll a diagnostic job. Returns {status, result}."""
    current_user = _get_user()['username']
    with _diag_lock:
        job = _diag_jobs.get(job_id)
        if not job:
            return jsonify({'error': 'Job not found'}), 404
        if job.get('user') != current_user:
            return jsonify({'error': 'Forbidden'}), 403
        snapshot = {'status': job['status'], 'result': job.get('result')}
    return jsonify(snapshot)


@app.route('/api/avatar')
@require_auth
def api_avatar():
    """Serve the current user's LDAP avatar photo (JPEG), cached to disk."""
    username = _get_user()['username'].strip()
    if not username:
        return '', 404

    ldap_cfg = _config().get('ldap', {})
    if not ldap_cfg.get('host') or not ldap_cfg.get('bind_dn'):
        logger.warning('Avatar unavailable: config.json has no ldap.host / ldap.bind_dn')
        return '', 404

    # Build a safe cache filename from the username (alphanumeric + - _ .)
    safe_name = re.sub(r'[^a-zA-Z0-9._-]', '_', username)
    cache_dir = os.path.abspath(
        ldap_cfg.get('cache_dir', os.path.join(tempfile.gettempdir(), 'gandalf_avatars'))
    )
    os.makedirs(cache_dir, exist_ok=True)
    cache_file = os.path.abspath(os.path.join(cache_dir, f'user_{safe_name}.jpg'))
    sentinel   = os.path.abspath(os.path.join(cache_dir, f'user_{safe_name}.none'))
    # Guard against path escape (shouldn't happen with sanitised safe_name, but be explicit)
    if not cache_file.startswith(cache_dir + os.sep) or not sentinel.startswith(cache_dir + os.sep):
        logger.error(f'Avatar path escape detected for user {username!r}')
        return '', 404
    try:
        cache_ttl = int(ldap_cfg.get('cache_ttl', 3600))
    except (ValueError, TypeError):
        logger.warning('Invalid cache_ttl in ldap config; using default 3600')
        cache_ttl = 3600

    now = time.time()

    # Serve cached image if fresh
    if os.path.exists(cache_file) and now - os.path.getmtime(cache_file) < cache_ttl:
        return send_file(cache_file, mimetype='image/jpeg',
                         max_age=cache_ttl, conditional=True)

    # Skip LDAP if we already know this user has no avatar
    try:
        if os.path.exists(sentinel) and now - os.path.getmtime(sentinel) < cache_ttl:
            logger.info(f'Avatar for {username!r}: cached "no photo" marker is still valid')
            return '', 404
    except OSError:
        pass

    # Query lldap
    bind_pw = ldap_cfg.get('bind_pw', '')
    if not bind_pw:
        logger.error('LDAP bind_pw not configured: avatar lookup disabled')
        return '', 404

    avatar_data = None
    conn = None
    try:
        import ldap3
        server = ldap3.Server(ldap_cfg['host'], port=int(ldap_cfg.get('port', 3890)))
        conn   = ldap3.Connection(server,
                                  user=ldap_cfg['bind_dn'],
                                  password=bind_pw,
                                  auto_bind=True, receive_timeout=5)
        safe_uid = ldap3.utils.conv.escape_filter_chars(username)
        conn.search(ldap_cfg.get('user_base', 'ou=people,dc=example,dc=com'),
                    f'(uid={safe_uid})', attributes=['avatar'])
        if conn.entries and conn.entries[0]['avatar'].value:
            avatar_data = conn.entries[0]['avatar'].value
    except ImportError:
        logger.error('ldap3 not installed; run pip install ldap3')
        return '', 404
    except Exception as e:
        logger.error(f'LDAP avatar lookup failed for {username}: {e}')
        return '', 404
    finally:
        if conn is not None:
            try:
                conn.unbind()
            except Exception:
                pass

    if not avatar_data or len(avatar_data) < 100:
        logger.warning(f'Avatar for {username!r}: LDAP has no photo (or no user with that uid)')
        with open(sentinel, 'w'):
            pass
        return '', 404

    # Validate JPEG magic bytes (FF D8 FF)
    if isinstance(avatar_data, str):
        avatar_data = avatar_data.encode('latin-1')
    if avatar_data[:3] != b'\xFF\xD8\xFF':
        logger.warning(f'Non-JPEG avatar data for {username}')
        with open(sentinel, 'w'):
            pass
        return '', 404

    with open(cache_file, 'wb') as f:
        f.write(avatar_data)
    if os.path.exists(sentinel):
        os.unlink(sentinel)

    resp = make_response(avatar_data)
    resp.headers['Content-Type']  = 'image/jpeg'
    resp.headers['Cache-Control'] = f'private, max-age={cache_ttl}'
    return resp


# ---------------------------------------------------------------------------
# Cutsheets
# ---------------------------------------------------------------------------

_cutsheet_schema_ready = False
_cutsheet_schema_lock = threading.Lock()


def _ensure_cutsheet_schema() -> None:
    """Create the cutsheet tables on first use, so deploys need no migration step."""
    global _cutsheet_schema_ready
    if _cutsheet_schema_ready:
        return
    with _cutsheet_schema_lock:
        if not _cutsheet_schema_ready:
            db.ensure_cutsheet_schema()
            _cutsheet_schema_ready = True


def _node_json(node) -> dict:
    if node is None:
        return {'device': '', 'port': '', 'side': '', 'text': 'unterminated'}
    return {'device': node[0], 'port': node[1], 'side': node[2], 'text': cutsheet.describe_node(node)}


def _cutsheet_catalog(link_stats: dict, items: list) -> list:
    """Devices and their ports, for the cable form's suggestions."""
    catalog = []
    for name, sw in sorted((link_stats.get('unifi_switches') or {}).items()):
        ports = sorted((p for p in (sw.get('ports') or {}).items() if p[1].get('port_idx') is not None),
                       key=lambda kv: kv[1]['port_idx'])
        catalog.append({'name': name, 'kind': 'switch', 'ports': [
            {'value': str(p['port_idx']), 'text': f'{p["port_idx"]} · {pname}'} for pname, p in ports]})
    for name, ifaces in sorted((link_stats.get('hosts') or {}).items()):
        catalog.append({'name': name, 'kind': 'host',
                        'ports': [{'value': i, 'text': i} for i in sorted(ifaces)]})
    known = {c['name'] for c in catalog}
    for item in items:
        if item['name'] in known:
            continue
        ports = [{'value': str(n), 'text': str(n)} for n in range(1, (item.get('port_count') or 0) + 1)]
        catalog.append({'name': item['name'], 'kind': item['kind'], 'ports': ports})
    return catalog


def _cutsheet_view() -> dict:
    """Every cutsheet record with its traced path and plan-vs-actual verdict."""
    _ensure_cutsheet_schema()
    data = db.cutsheet_load()
    link_stats = _state_json('link_stats')
    switches = link_stats.get('unifi_switches') or {}
    items = {i['name']: i for i in data['items']}
    results = cutsheet.evaluate(data['cables'], items, switches)
    for cable in data['cables']:
        r = results[cable['id']]
        cable['verdict'] = r['verdict']
        cable['detail'] = r['detail']
        cable['problem'] = r['verdict'] in cutsheet.PROBLEM_VERDICTS
        cable['path'] = [
            {'cable_id': h} if isinstance(h, int) else _node_json(h) for h in r['path']['hops']
        ]
        cable['a_text'] = cutsheet.describe_node(cutsheet._node(cable, 'a'))
        cable['b_text'] = cutsheet.describe_node(cutsheet._node(cable, 'b'))
    data['suggestions'] = cutsheet.undocumented(data['cables'], switches, items)
    data['labels'] = cutsheet.labels_to_make(data['cables'], items, data['port_labels'])
    data['next_label'] = cutsheet.next_label(data['cables'])
    data['catalog'] = _cutsheet_catalog(link_stats, data['items'])
    data['racks'] = [r['name'] for r in _racks(data['items'])]
    snapshot = _state_json('network_snapshot')
    data['devices'] = sorted(set(switches) | set(snapshot.get('hosts') or {})
                             | {u['name'] for u in snapshot.get('unifi') or [] if u.get('name')}, key=str.lower)
    data['verdicts'] = cutsheet.VERDICTS
    data['chassis'] = cutsheet.CHASSIS
    data['options'] = {'kinds': cutsheet.ITEM_KINDS, 'passive_kinds': cutsheet.PASSIVE_KINDS,
                       'statuses': cutsheet.STATUSES, 'cable_types': cutsheet.CABLE_TYPES}
    data['updated'] = link_stats.get('updated')
    return data


def _cutsheet_for_device(name: str) -> list:
    """The cables that end on one device, for its port panel."""
    view = _cutsheet_view()
    # Cables may name the device or a cutsheet item linked to it
    names = {name} | {i['name'] for i in view['items'] if i.get('device') == name}
    out = []
    for cable in view['cables']:
        for end in ('a', 'b'):
            if cable[f'{end}_device'] in names and cable['status'] != 'retired':
                out.append({
                    'port': cable[f'{end}_port'], 'side': cable[f'{end}_side'], 'id': cable['id'],
                    'label': cable['label'], 'status': cable['status'], 'verdict': cable['verdict'],
                    'detail': cable['detail'], 'problem': cable['problem'], 'path': cable['path'],
                })
    return out


def _json_body() -> dict:
    """The request's JSON object. Requiring a JSON content type also keeps another
    site's plain form post from reaching these write endpoints (no CORS preflight)."""
    if not request.is_json:
        raise cutsheet.ValidationError('Send the request as JSON')
    data = request.get_json(silent=True)
    if not isinstance(data, dict):
        raise cutsheet.ValidationError('Request body must be a JSON object')
    return data


def _cutsheet_write(fn):
    """Turn validation and conflict errors from a write into 400 and 409 responses."""
    @wraps(fn)
    def wrapper(*args, **kwargs):
        try:
            _ensure_cutsheet_schema()
            return fn(*args, **kwargs)
        except cutsheet.ValidationError as e:
            return jsonify({'error': str(e)}), 400
        except db.CutsheetConflict as e:
            return jsonify({'error': str(e)}), 409
    return wrapper


def _cutsheet_items() -> dict:
    return {i['name']: i for i in db.cutsheet_load()['items']}


@app.route('/cutsheet')
@require_auth
def cutsheet_page():
    user = _get_user()
    return render_template('cutsheet.html', user=user, can_edit='admin' in user['groups'])


@app.route('/api/cutsheet')
@require_auth
def api_cutsheet():
    view = _cutsheet_view()
    view['history'] = db.cutsheet_history(limit=100)
    return jsonify(view)


# ---------------------------------------------------------------------------
# Inventory (HomeBox) links
# ---------------------------------------------------------------------------

_INVENTORY_DEFAULT = 'https://inventory.lotusguild.org'


def _inventory_base() -> str:
    return ((_config().get('inventory') or {}).get('base_url') or _INVENTORY_DEFAULT).rstrip('/')


def _inventory_for(name: str, model: str) -> dict:
    """The device's inventory record link, or a search to find it."""
    base = _inventory_base()
    template = (_config().get('inventory') or {}).get('search_url') or '{base}/items?q={q}'
    try:
        _ensure_cutsheet_schema()
        url = db.get_device_link(name)
    except Exception as e:
        logger.error(f'Inventory link unavailable for {name}: {e}')
        url = ''
    return {'url': url, 'search_url': template.format(base=base, q=quote(model or name, safe=''))}


def _validate_inventory_url(value) -> str:
    """Only links into the configured inventory site: this URL is shown as a link to every viewer."""
    url = str(value or '').strip()
    if not url:
        return ''
    if len(url) > 512:
        raise cutsheet.ValidationError('Inventory link is too long')
    base = urlparse(_inventory_base())
    got = urlparse(url)
    if got.scheme != 'https' or got.netloc.lower() != base.netloc.lower() or got.username or got.password:
        raise cutsheet.ValidationError(f'Inventory link must start with {base.scheme}://{base.netloc}/')
    return url


@app.route('/api/device/<path:name>/inventory', methods=['PUT'])
@require_auth
@require_admin
@_cutsheet_write
def api_device_inventory(name: str):
    """Set or clear (empty url) the device's inventory link."""
    if len(name) > 255:
        raise cutsheet.ValidationError('Device name is too long')
    url = _validate_inventory_url(_json_body().get('url'))
    db.set_device_link(name, url, _get_user()['username'])
    logger.info(f'{_get_user()["username"]} set inventory link for {name}: {url or "(removed)"}')
    return jsonify({'success': True, 'url': url})


# ---------------------------------------------------------------------------
# Rack elevation
# ---------------------------------------------------------------------------

@app.route('/rack')
@require_auth
def rack_page():
    user = _get_user()
    return render_template('rack.html', user=user, can_edit='admin' in user['groups'])


def _racks(items: list) -> list:
    """Every rack, in order: those saved on the rack page, then any other name an item uses.

    With none saved, racks come from config: "racks": [{"name": "Main", "units": 42}, ...], or
    one rack named rack.name (default "Rack 1") with rack.units (default 42). Those have no id
    until saved. An item with no rack belongs to the first.
    """
    def units(value, default=42):
        try:
            return max(1, min(int(value), 60))
        except (TypeError, ValueError):
            return default

    cfg = _config()
    single = cfg.get('rack') or {}
    default_units = units(single.get('units', 42))
    racks = [{'id': r['id'], 'name': r['name'], 'units': r['units'], 'notes': r.get('notes') or ''}
             for r in db.cutsheet_racks()]
    if not racks:
        racks = [{'id': None, 'name': str(r.get('name') or '').strip(), 'notes': '',
                  'units': units(r.get('units', default_units), default_units)}
                 for r in cfg.get('racks') or [] if isinstance(r, dict) and str(r.get('name') or '').strip()]
    if not racks:
        racks = [{'id': None, 'name': str(single.get('name') or 'Rack 1'), 'units': default_units, 'notes': ''}]
    named = {r['name'] for r in racks}
    for item in items:
        if item.get('rack') and item['rack'] not in named:
            named.add(item['rack'])
            racks.append({'id': None, 'name': item['rack'], 'units': default_units, 'notes': ''})
    return racks


_physical_interfaces = devices.physical_interfaces


def _device_status(name: str, snapshot: dict) -> str:
    host = (snapshot.get('hosts') or {}).get(name)
    if host:
        return host.get('status') or 'unknown'
    for d in snapshot.get('unifi') or []:
        if d.get('name') == name:
            return 'up' if d.get('connected') else 'down'
    return 'unknown'


@app.route('/api/rack')
@require_auth
def api_rack():
    """One rack face: placed cutsheet items, live switch ports, and each port's cutsheet state.

    Query: rack (name; default the first rack), face (front|rear).
    """
    face = request.args.get('face', 'front')
    if face not in cutsheet.RACK_VIEWS:
        return jsonify({'error': f'face must be one of {", ".join(cutsheet.RACK_VIEWS)}'}), 400
    view = _cutsheet_view()
    link_stats = _state_json('link_stats')
    switches = link_stats.get('unifi_switches') or {}
    unifi_info = link_stats.get('unifi_devices') or {}
    snapshot = _state_json('network_snapshot')
    online = {d['name']: d.get('connected', True) for d in snapshot.get('unifi') or []}
    known = set(switches) | set(online) | set(snapshot.get('hosts') or {})

    racks = _racks(view['items'])
    default = racks[0]['name']
    current = next((r for r in racks if r['name'] == request.args.get('rack', default)), None)
    if current is None:
        return jsonify({'error': 'No such rack'}), 404

    def rack_of(item):
        return item.get('rack') or default

    for r in racks:
        racked = [i for i in view['items'] if not i.get('shelf') and i.get('rack_u') and rack_of(i) == r['name']]
        r['count'] = len(racked)
        # How many items each view shows: rear-post gear is what "behind" and "rear" add
        r['rear'] = sum(1 for i in racked if i.get('mount') in ('rear', 'full'))
        r['behind'] = sum(1 for i in racked if i.get('mount') == 'rear')
    for cable in view['cables']:
        cable['run'] = [h['cable_id'] for h in cable['path'] if 'cable_id' in h]
    marks = cutsheet.port_marks(view['cables'])
    labels: dict = {}
    for pl in view['port_labels']:
        labels.setdefault(pl['item_name'], {})[str(pl['port'])] = pl['label']

    def dev(item):
        # An item can be named after the product ("UDM-Pro") and linked to the live device
        return item.get('device') or item['name']

    def url(item):
        return f'/device/{quote(dev(item), safe="")}' if dev(item) in known else f'/cutsheet#item-{item["id"]}'

    on_shelf: dict = {}
    for item in view['items']:
        if item.get('shelf'):
            on_shelf.setdefault(item['shelf'], []).append(item)

    here = [i for i in view['items'] if not i.get('shelf') and rack_of(i) == current['name']]
    layout = cutsheet.rack_layout(here, current['units'], face)
    out = []
    for item in layout['placed']:
        name, device = item['name'], dev(item)
        entry = {k: item.get(k) for k in ('id', 'name', 'kind', 'port_count', 'rack_u', 'rack_height',
                                          'rack_top', 'location', 'notes')}
        entry['device'] = device
        entry['linked'] = device in known
        entry['mount'] = item.get('mount') or 'front'
        entry['chassis'] = item.get('chassis') or ''
        entry['through'] = item.get('through', False)
        entry['url'] = url(item)
        entry['status'] = _device_status(device, snapshot) if device in known else ''
        # Cables may name either the item or the device it is linked to
        entry['ports'] = {port: m for (d, port), m in marks.items() if d in (name, device)}
        entry['labels'] = labels.get(name, {})
        model = next((u.get('model') for u in snapshot.get('unifi') or [] if u.get('name') == device), '')
        if device in switches or model:
            sw = switches.get(device) or {}
            entry['switch'] = {'model': sw.get('model') or model or '', 'ports': sw.get('ports') or {},
                               'outlets': (unifi_info.get(device) or {}).get('outlets') or [],
                               'online': online.get(device, True)}
        if item['kind'] == 'shelf':
            # Left to right: by position on the shelf, then by name
            entry['contents'] = [
                {'id': c['id'], 'name': c['name'], 'kind': c['kind'], 'url': url(c),
                 'status': _device_status(dev(c), snapshot) if dev(c) in known else '',
                 'model': next((u.get('model') for u in snapshot.get('unifi') or [] if u.get('name') == dev(c)), '')
                 or '', 'notes': c.get('notes') or '', 'chassis': c.get('chassis') or '',
                 'slot': c.get('slot'), 'interfaces': _physical_interfaces(dev(c), snapshot),
                 'device': dev(c), 'ports': (switches.get(dev(c)) or {}).get('ports') or {}}
                for c in sorted(on_shelf.get(name, []), key=lambda c: (c.get('slot') or 999, c['name'].lower()))
            ]
        out.append(entry)
    return jsonify({
        'racks': racks, 'rack': current['name'], 'rack_id': current['id'], 'units': current['units'],
        'notes': current['notes'], 'face': face, 'items': out,
        'overlaps': layout['overlaps'], 'outside': layout['outside'],
        'unplaced': sorted(i['name'] for i in here if not i.get('rack_u')),
        'updated': link_stats.get('updated'),
    })


# ---------------------------------------------------------------------------
# Topology
# ---------------------------------------------------------------------------

@app.route('/topology')
@require_auth
def topology_page():
    return render_template('topology.html', user=_get_user())


@app.route('/api/topology')
@require_auth
def api_topology():
    """Every device and physical cable, for the topology page (see topology.build_graph)."""
    events = db.get_active_events(limit=_PAGE_LIMIT)
    _annotate_suppressions(events, db.get_active_suppressions())
    try:
        _ensure_cutsheet_schema()
        sheet = db.cutsheet_load()
    except Exception as e:
        logger.error(f'Cutsheet unavailable for topology: {e}')
        sheet = {'items': [], 'cables': []}
    graph = topology.build_graph(
        _state_json('network_snapshot'), _state_json('link_stats'), _state_json('modem_stats'),
        events, sheet['items'], sheet['cables'], _state_json('known_uplinks'),
    )
    return jsonify(graph)


@app.route('/api/cutsheet/racks', methods=['POST'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_create_rack():
    """Add a rack. `previous` is the name of the default rack this replaces, when the first is saved."""
    body = _json_body()
    rack = cutsheet.validate_rack(body)
    previous = str(body.get('previous') or '')[:100] if not db.cutsheet_racks() else ''
    rack_id = db.cutsheet_save_rack(None, rack, _get_user()['username'], previous=previous)
    return jsonify({'success': True, 'id': rack_id})


@app.route('/api/cutsheet/racks/<int:rack_id>', methods=['PUT'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_update_rack(rack_id: int):
    rack = cutsheet.validate_rack(_json_body())
    db.cutsheet_save_rack(rack_id, rack, _get_user()['username'])
    return jsonify({'success': True})


@app.route('/api/cutsheet/racks/<int:rack_id>', methods=['DELETE'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_delete_rack(rack_id: int):
    if not db.cutsheet_delete_rack(rack_id, _get_user()['username']):
        return jsonify({'error': 'No such rack'}), 404
    return jsonify({'success': True})


# ---------------------------------------------------------------------------
# Unused interfaces
# ---------------------------------------------------------------------------

_IFACE_RE = re.compile(r'^[A-Za-z0-9_.:@-]{1,64}$')


@app.route('/api/device/<path:name>/interfaces/<iface>/unused', methods=['PUT', 'DELETE'])
@require_auth
@require_admin
@_cutsheet_write
def api_interface_unused(name: str, iface: str):
    """Mark a host interface unused (PUT {reason}) or in use again (DELETE).

    Unused interfaces raise no link-down alerts and don't count toward the host's status;
    marking one quietly closes any link-down alert it has open.
    """
    if len(name) > 255 or not _IFACE_RE.match(iface):
        raise cutsheet.ValidationError('Not a valid host or interface name')
    user = _get_user()['username']
    if request.method == 'DELETE':
        found = db.clear_unused_interface(name, iface)
        logger.info(f'{user} marked {name} {iface} in use again')
        return jsonify({'success': True, 'changed': found})
    reason = str(_json_body().get('reason') or '').strip()[:255]
    db.set_unused_interface(name, iface, reason, user)
    logger.info(f'{user} marked {name} {iface} unused: {reason or "(no reason)"}')
    return jsonify({'success': True})


@app.route('/api/cutsheet/cables', methods=['POST'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_create_cable():
    cable = cutsheet.validate_cable(_json_body(), _cutsheet_items())
    cable_id = db.cutsheet_create_cable(cable, _get_user()['username'])
    return jsonify({'success': True, 'id': cable_id})


@app.route('/api/cutsheet/cables/<int:cable_id>', methods=['PUT'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_update_cable(cable_id: int):
    cable = cutsheet.validate_cable(_json_body(), _cutsheet_items())
    if not db.cutsheet_update_cable(cable_id, cable, _get_user()['username']):
        return jsonify({'error': 'No such cable'}), 404
    return jsonify({'success': True})


@app.route('/api/cutsheet/cables/<int:cable_id>', methods=['DELETE'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_delete_cable(cable_id: int):
    if not db.cutsheet_delete_cable(cable_id, _get_user()['username']):
        return jsonify({'error': 'No such cable'}), 404
    return jsonify({'success': True})


@app.route('/api/cutsheet/items', methods=['POST'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_create_item():
    item = cutsheet.validate_item(_json_body(), _cutsheet_items())
    item_id = db.cutsheet_create_item(item, _get_user()['username'])
    return jsonify({'success': True, 'id': item_id})


@app.route('/api/cutsheet/items/<int:item_id>', methods=['PUT'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_update_item(item_id: int):
    item = cutsheet.validate_item(_json_body(), _cutsheet_items())
    if not db.cutsheet_update_item(item_id, item, _get_user()['username']):
        return jsonify({'error': 'No such item'}), 404
    return jsonify({'success': True})


@app.route('/api/cutsheet/items/<int:item_id>', methods=['DELETE'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_delete_item(item_id: int):
    if not db.cutsheet_delete_item(item_id, _get_user()['username']):
        return jsonify({'error': 'No such item'}), 404
    return jsonify({'success': True})


@app.route('/api/cutsheet/items/<int:item_id>/ports/<port>', methods=['PUT'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_port_label(item_id: int, port: str):
    data = _json_body()
    label = str(data.get('label') or '').strip()
    if len(label) > 64:
        raise cutsheet.ValidationError('Port label must be 64 characters or fewer')
    if len(port) > cutsheet.PORT_MAX:
        raise cutsheet.ValidationError('Port is too long')
    if not db.cutsheet_set_port_label(item_id, port, label, bool(data.get('applied')), _get_user()['username']):
        return jsonify({'error': 'No such item'}), 404
    return jsonify({'success': True})


_live_ports_cache: dict = {'at': 0.0, 'data': None, 'client': None}
_live_ports_lock = threading.Lock()
_LIVE_PORTS_TTL = 2.0


@app.route('/api/cutsheet/live-ports')
@require_auth
@require_admin
def api_cutsheet_live_ports():
    """Current up/down state of every switch port, read from UniFi now.

    Trace mode polls this while someone unplugs a cord, to see which port dropped.
    The monitor's cached state is minutes old, so it can't be used for this.
    """
    with _live_ports_lock:
        now = time.time()
        if _live_ports_cache['data'] is None or now - _live_ports_cache['at'] > _LIVE_PORTS_TTL:
            if _live_ports_cache['client'] is None:
                _live_ports_cache['client'] = UnifiClient(_config()['unifi'])
            raw = _live_ports_cache['client'].get_switch_ports()
            if raw is None:
                return jsonify({'error': 'The UniFi controller did not answer'}), 502
            _live_ports_cache['data'] = {
                sw: {str(p['port_idx']): {'up': bool(p.get('up')), 'name': name}
                     for name, p in (d.get('ports') or {}).items() if p.get('port_idx') is not None}
                for sw, d in raw.items()
            }
            _live_ports_cache['at'] = now
        return jsonify({'switches': _live_ports_cache['data'], 'at': _live_ports_cache['at']})


@app.route('/api/cutsheet/cables.csv')
@require_auth
def api_cutsheet_export():
    _ensure_cutsheet_schema()
    resp = make_response(cutsheet.cables_to_csv(db.cutsheet_load()['cables']))
    resp.headers['Content-Type'] = 'text/csv; charset=utf-8'
    resp.headers['Content-Disposition'] = 'attachment; filename="gandalf-cutsheet.csv"'
    return resp


@app.route('/api/cutsheet/import', methods=['POST'])
@require_auth
@require_admin
@_cutsheet_write
def api_cutsheet_import():
    data = _json_body()
    text = data.get('csv')
    if not isinstance(text, str) or not text.strip():
        raise cutsheet.ValidationError('Paste or choose a CSV file')
    if len(text) > 1_000_000:
        raise cutsheet.ValidationError('CSV is larger than 1 MB')
    cables, errors = cutsheet.parse_csv(text, _cutsheet_items())
    if errors:
        return jsonify({'error': 'Nothing was imported', 'errors': errors[:50]}), 400
    if data.get('dry_run'):
        return jsonify({'success': True, 'dry_run': True, 'rows': len(cables)})
    result = db.cutsheet_import_cables(cables, _get_user()['username'])
    return jsonify(dict(result, success=True))


@app.route('/health')
def health():
    """Health check endpoint (no auth). Checks DB and monitor freshness."""
    checks = {}
    overall = 'ok'

    # DB connectivity
    try:
        db.get_state('last_check')
        checks['db'] = 'ok'
    except Exception as e:
        logger.error(f'Health check db error: {e}')
        checks['db'] = 'error'
        overall = 'degraded'

    # Monitor freshness: a hung or crashed monitor stops updating last_check
    try:
        last_check = db.get_state('last_check', '')
        age_s = _monitor_age(last_check)
        if not last_check:
            checks['monitor'] = 'no data yet'
        elif age_s is None or age_s > _monitor_stale_after():
            checks['monitor'] = 'stale'
            checks['monitor_age_s'] = int(age_s) if age_s is not None else None
            overall = 'degraded'
        else:
            checks['monitor'] = 'ok'
            checks['monitor_age_s'] = int(age_s)
    except Exception as e:
        logger.error(f'Health check monitor error: {e}')
        checks['monitor'] = 'error'
        overall = 'degraded'

    status_code = 200 if overall == 'ok' else 503
    return jsonify({'status': overall, 'service': 'gandalf', 'checks': checks}), status_code


if __name__ == '__main__':
    app.run(debug=True, host='0.0.0.0', port=5000)  # nosec B201 B104: dev runner only; production uses gunicorn