pulse_ops #

A modern, Flutter-native developer toolkit for in-app network inspection, performance & memory monitoring, and crash diagnostics — built for iOS & Android, with a beautiful dark Material 3 UI.

PulseOps in action — floating launcher, expandable inspector, cURL export, and log sharing

PulseOps ships with four focused capabilities:

🌐 Network Inspector — a Dio interceptor that records every request, pretty-prints JSON, exports cURL, retries calls, and presents it all in a developer-grade dark inspector.
⚡ Performance Monitoring — real-time FPS tracking, frame drop & jank detection, startup time measurement, and API latency charts.
🧠 Memory Monitoring — RSS memory tracking, spike detection, leak detection via FlutterMemoryAllocations, widget lifecycle logs, and rebuild frequency tracking.
💥 Crash Diagnostics — pluggable bridge to Firebase Crashlytics (or any backend) with rich breadcrumbs and automatic attachment of recent API activity to every crash report.

✨ Highlights #

🎨 Beautiful dark, Material 3 inspector with monospace JSON viewer and syntax highlighting
🔌 One-line Dio integration — works with GET, POST, PUT, PATCH, DELETE, and multipart/form-data
🔍 Search, filter by method / status family / slow / failed — live and composable
📋 Copy buttons everywhere — headers, body, full cURL
↻ Retry requests from the inspector with your real Dio client
⚡ Real-time FPS monitor — sparkline chart, dropped-frame list, startup time, and API latency bar chart
🧠 Memory monitor — RSS sparkline, spike warnings, leak list, widget lifecycle breakdown, and rebuild frequency — all live
💾 Persistent network store — FileBackedNetworkStore survives app restarts with zero extra setup
📡 Unified event exporter — one interface to forward every failed request and every crash to your own analytics backend
📳 Shake to open — shake the device to launch the inspector
🔒 Sanitization for secrets / tokens / passwords before storage or upload
🧭 Breadcrumb trail with bounded ring buffer
💥 Backend-agnostic crash reporter — wire Crashlytics, Sentry, or your own logger via a thin PulseCrashReporter interface
🛡️ Production-safe — disabled in release builds by default
🪶 Lightweight — no Firebase or Isar at runtime; pure Dart + Dio + Riverpod core

🚀 Quick start #

1. Add the dependency #

dependencies:
  pulse_ops: ^1.3.0
  dio: ^5.4.0

2. Initialize in `main()` #

import 'package:dio/dio.dart';
import 'package:flutter/material.dart';
import 'package:pulse_ops/pulse_ops.dart';

Future<void> main() async {
  WidgetsFlutterBinding.ensureInitialized();

  await PulseOps.initialize(
    crashlytics: true,
    enableInRelease: false,
    sanitizeKeys: ['token', 'password', 'authorization'],
  );

  final dio = Dio()..interceptors.add(PulseOps.instance.dioInterceptor);

  runApp(PulseOps.instance.wrap(retryDio: dio, child: const MyApp()));
}

That's it. A draggable floating button appears in debug builds; tap it to open the inspector. Use the 🧠 toolbar button for memory, ⚡ for performance.

3. (Optional) Open the inspector imperatively #

PulseOps.instance.openInspector(context, retryDio: dio);

🌐 Network Inspector #

Every Dio call routed through PulseOps.instance.dioInterceptor is captured as a NetworkRecord and pushed into an in-memory ring buffer (configurable via PulseOpsConfig.maxRecords).

The inspector UI provides:

Surface	Contents
Timeline list	Newest-first list of requests with method chip, host, path, timestamp, duration, status chip
Overview tab	Status, timing, request/response sizes, error details
Headers tab	Sanitized request + response headers with copy-all
Request tab	Query params and request body with syntax-highlighted JSON
Response tab	Highlighted response body and error banner
cURL tab	One-tap copy of the full `curl` command

The list supports:

🔎 Live search across URL, method, status, host, and error message
🎯 Filter by HTTP method (GET / POST / PUT / PATCH / DELETE)
⚠️ Failed only toggle — show only errored requests
🐢 Slow only toggle — requests exceeding slowRequestThresholdMs
🔢 Status-family chips — 2xx / 3xx / 4xx / 5xx

Persistent store #

To keep network history across app restarts, swap in FileBackedNetworkStore:

final store = FileBackedNetworkStore(maxRecords: 200);
await store.initialize();              // loads previous session from disk

await PulseOps.initialize(
  networkStore: store,
);

Retrying a request #

Pass your authenticated Dio instance to wrap(retryDio:) or openInspector(retryDio:). The retry button in the app bar reissues the captured request via that client.

Multipart support #

FormData payloads are described (field names, file names, sizes) rather than serialized — useful for inspecting uploads without breaking streams.

⚡ Performance Monitoring #

The performance screen is available from the ⚡ icon in the inspector toolbar. It requires no additional dependencies — everything uses Flutter's built-in WidgetsBinding.addTimingsCallback.

What's tracked #

Metric	Description
Startup time	Wall-clock time from `PulseOps.initialize()` to the first rendered frame
Current FPS	Rolling average over the last 15 frames
Dropped frames	Frames taking > 16 ms (one refresh period at 60 Hz)
Severe jank	Frames taking > 33 ms (two full refresh periods)
API latency	Per-request bar chart for the last 40 completed calls

FPS chart #

A gradient-filled sparkline shows FPS over the last N frames (default 300). The line turns green (≥ 55 fps), yellow (≥ 40 fps), or red (< 40 fps). Reference grid lines are drawn at 60 fps and 30 fps.

Latency chart #

Each bar represents one completed request, coloured:

🟢 Green — within the slow-request threshold
🟡 Yellow — exceeds slowRequestThresholdMs
🔴 Red — the request failed

A dashed threshold line marks the slow boundary.

🧠 Memory Monitoring #

The memory screen is available from the 🧠 icon in the inspector toolbar. It uses dart:io's ProcessInfo.currentRss for RSS sampling and Flutter's FlutterMemoryAllocations for object lifecycle — no extra dependencies.

Metrics #

Metric	Description
Current RSS	Process resident set size, sampled every 2 s (configurable)
Peak RSS	Highest RSS reading since monitoring started
Memory spikes	Samples >20% above the 10-sample rolling average
Potential leaks	Objects created but not disposed after 30 s
Widget lifecycle	Created / active / disposed counts per type
Rebuild counts	How many times each widget type has been rebuilt

Leak detection #

PulseOps subscribes to FlutterMemoryAllocations and tracks the lifecycle of every ChangeNotifier, AnimationController, TextEditingController, and other disposable Flutter objects. Any object that remains alive for > 30 s without being disposed appears in the Potential Leaks list with its age.

Rebuild tracking #

Call PulseOps.instance.memoryStore.recordRebuild(widgetType) at the top of your build() method to track rebuild frequency:

@override
Widget build(BuildContext context) {
  PulseOps.instance.memoryStore.recordRebuild('MyHeavyWidget');
  return ...;
}

Counts appear colour-coded in the Memory screen — red (> 50), yellow (> 20).

📡 Unified Event Exporter #

Implement PulseEventExporter and pass it to PulseOps.initialize to receive every failed API call and every crash in one place, so you can forward them to Mixpanel, Amplitude, your own backend, or any analytics sink:

class MyAnalyticsExporter implements PulseEventExporter {
  @override
  Future<void> onFailedRequest(NetworkRecord record) async {
    await MyAnalytics.track('api_error', {
      'url': record.url,
      'method': record.method,
      'status': record.statusCode,
      'error': record.error,
      'duration_ms': record.duration.inMilliseconds,
    });
  }

  @override
  Future<void> onCrash(
    Object error,
    StackTrace? stackTrace, {
    required String? reason,
    required bool fatal,
    required List<Breadcrumb> breadcrumbs,
    required List<NetworkRecord> recentRequests,
  }) async {
    await MyAnalytics.trackCrash(
      error.toString(),
      fatal: fatal,
      context: {'breadcrumb_count': breadcrumbs.length},
    );
  }
}

await PulseOps.initialize(
  eventExporter: MyAnalyticsExporter(),
);

This is separate from PulseCrashReporter — the exporter fires for both crashes and failed requests, making it ideal for a unified observability pipeline.

💥 Crash Diagnostics #

PulseOps decouples itself from any specific crash backend via the PulseCrashReporter interface, so the package itself does not depend on firebase_crashlytics. You wire that up in your app.

Example adapter for Firebase Crashlytics #

import 'package:firebase_crashlytics/firebase_crashlytics.dart';
import 'package:pulse_ops/pulse_ops.dart';

class FirebaseCrashReporterAdapter implements PulseCrashReporter {
  FirebaseCrashReporterAdapter(this._c);
  final FirebaseCrashlytics _c;

  @override
  Future<void> recordNonFatal(Object error,
      {StackTrace? stackTrace, String? reason, Map<String, dynamic>? context}) async {
    await _attach(context);
    await _c.recordError(error, stackTrace, reason: reason, fatal: false);
  }

  @override
  Future<void> recordFatal(Object error,
      {StackTrace? stackTrace, Map<String, dynamic>? context}) async {
    await _attach(context);
    await _c.recordError(error, stackTrace, fatal: true);
  }

  @override
  Future<void> attachBreadcrumbs(List<Breadcrumb> breadcrumbs) async {
    for (final b in breadcrumbs) {
      await _c.log(b.toString());
    }
  }

  @override
  Future<void> attachNetworkHistory(List<NetworkRecord> records) async {
    final summary = records.take(20).map((r) =>
        '${r.method} ${r.endpoint} -> ${r.statusCode ?? r.status.name}').join('\n');
    await _c.setCustomKey('pulse_ops_recent_requests', summary);
  }

  @override
  Future<void> setCustomKey(String key, Object value) =>
      _c.setCustomKey(key, value);

  Future<void> _attach(Map<String, dynamic>? context) async {
    if (context == null) return;
    for (final e in context.entries) {
      await _c.setCustomKey(e.key, e.value.toString());
    }
  }
}

Then pass it in:

await PulseOps.initialize(
  crashReporter: FirebaseCrashReporterAdapter(FirebaseCrashlytics.instance),
);

Example adapter for Sentry #

import 'package:pulse_ops/pulse_ops.dart';
import 'package:sentry_flutter/sentry_flutter.dart';

class SentryCrashReporterAdapter implements PulseCrashReporter {
  const SentryCrashReporterAdapter();

  @override
  Future<void> recordNonFatal(Object error,
      {StackTrace? stackTrace, String? reason, Map<String, dynamic>? context}) async {
    await Sentry.captureException(
      error,
      stackTrace: stackTrace,
      hint: Hint.withMap({
        if (reason != null) 'reason': reason,
        if (context != null) ...context,
      }),
    );
  }

  @override
  Future<void> recordFatal(Object error,
      {StackTrace? stackTrace, Map<String, dynamic>? context}) async {
    await Sentry.captureException(
      error,
      stackTrace: stackTrace,
      withScope: (scope) => scope.setTag('fatal', 'true'),
    );
  }

  @override
  Future<void> attachBreadcrumbs(List<Breadcrumb> breadcrumbs) async {
    for (final b in breadcrumbs) {
      await Sentry.addBreadcrumb(
        SentryBreadcrumb(
          message: b.message,
          level: _sentryLevel(b.level),
          timestamp: b.timestamp,
          data: b.data,
        ),
      );
    }
  }

  @override
  Future<void> attachNetworkHistory(List<NetworkRecord> records) async {
    final summary = records.take(20).map((r) =>
        '${r.method} ${r.endpoint} -> ${r.statusCode ?? r.status.name}').join('\n');
    await Sentry.configureScope(
      (scope) => scope.setContexts('pulse_ops_recent_requests', {'log': summary}),
    );
  }

  @override
  Future<void> setCustomKey(String key, Object value) async {
    await Sentry.configureScope((scope) => scope.setTag(key, value.toString()));
  }

  SentryLevel _sentryLevel(BreadcrumbLevel level) {
    switch (level) {
      case BreadcrumbLevel.debug:   return SentryLevel.debug;
      case BreadcrumbLevel.info:    return SentryLevel.info;
      case BreadcrumbLevel.warning: return SentryLevel.warning;
      case BreadcrumbLevel.error:   return SentryLevel.error;
    }
  }
}

Then initialize Sentry first, then PulseOps:

await SentryFlutter.init(
  (options) => options.dsn = 'YOUR_DSN',
  appRunner: () async {
    await PulseOps.initialize(
      crashReporter: const SentryCrashReporterAdapter(),
    );
    runApp(PulseOps.instance.wrap(child: MyApp()));
  },
);

What gets attached to crashes #

Whenever an error is reported through PulseOps — automatically for failed HTTP requests, or manually via PulseOps.instance.recordError(...):

The breadcrumb trail (default 50 entries) is forwarded.
The last 20 network records are summarized and attached as context.
Any additional extra map you pass is merged in.

Adding your own breadcrumbs #

PulseOps.instance.log('User opened checkout', data: {'cart_size': 4});

Reporting errors manually #

try {
  await doRiskyThing();
} catch (e, st) {
  await PulseOps.instance.recordError(e, st, reason: 'checkout pipeline');
}

⚙️ Configuration #

const PulseOpsConfig(
  // — General —
  enableInRelease: false,                  // keep disabled in prod builds
  showOverlay: true,                       // floating launcher button

  // — Network —
  maxRecords: 200,                         // request ring-buffer capacity
  sanitizeKeys: ['authorization', ...],    // keys redacted before storage
  slowRequestThresholdMs: 2000,            // ms to flag a request as slow
  captureFailedRequestsAsCrashEvents: true,
  attachNetworkHistoryToCrashes: true,

  // — Performance —
  enableFpsMonitor: true,                  // frame-timing subscriber
  fpsFrameBufferSize: 300,                 // frames kept in memory

  // — Memory —
  enableMemoryMonitor: true,               // RSS polling + FlutterMemoryAllocations
  memorySampleIntervalSeconds: 2,          // polling interval
  memorySnapshotBufferSize: 120,           // snapshots kept (~4 min at 2 s)

  // — Overlay / UX —
  enableShakeToOpen: true,                 // shake gesture to open inspector
  shakeThreshold: 22.0,                    // m/s² to trigger a shake
  inspectorPresentation: InspectorPresentation.bottomSheet, // or .fullScreen

  // — Crash —
  maxBreadcrumbs: 50,
)

You can pass it directly to PulseOps.initialize(config: ...), or use the shorthand named args enableInRelease, sanitizeKeys, crashlytics.

🏗 Architecture #

lib/
├── pulse_ops.dart                         # public exports
└── src/
    ├── core/                              # facade + config + PulseEventExporter
    ├── network/
    │   ├── interceptor/                   # PulseDioInterceptor
    │   ├── models/                        # NetworkRecord (with toJson/fromJson)
    │   ├── store/                         # InMemoryNetworkStore, FileBackedNetworkStore
    │   └── utils/                         # CurlBuilder, Sanitizer, LogExporter
    ├── performance/
    │   ├── frame_metric.dart              # FrameMetric value type
    │   ├── performance_store.dart         # ring-buffer + stream
    │   └── fps_tracker.dart              # WidgetsBinding timing subscriber
    ├── memory/
    │   ├── memory_snapshot.dart           # RSS snapshot value type
    │   ├── tracked_object.dart            # object lifecycle record
    │   ├── memory_store.dart             # ring-buffer + leak map + rebuild counts
    │   └── memory_monitor.dart           # RSS polling + FlutterMemoryAllocations
    ├── crash/                             # breadcrumbs + reporter + bridge
    ├── ui/
    │   ├── inspector/                     # screens, tabs, widgets
    │   ├── performance/                   # PerformanceScreen + charts
    │   ├── memory/                        # MemoryScreen + RSS chart
    │   ├── overlay/                       # draggable launcher + shake detector
    │   └── theme/                         # dark Material 3 theme
    └── providers/                         # Riverpod scope

The design follows clean architecture principles: the network layer is plain Dart with no Flutter imports, the UI consumes data only through Riverpod providers, and crash/export backends are injected via interfaces. This makes it trivial to:

swap the in-memory store for FileBackedNetworkStore or any custom sink
substitute the crash reporter for Sentry, Bugsnag, or a custom logger
forward every event to your own backend via PulseEventExporter
embed the inspector inside a debug menu without using the overlay

🧪 Testing #

The package ships with a full test suite covering the sanitizer, cURL builder, in-memory store, breadcrumb trail, Dio interceptor (success / failure / sanitization paths), and the facade.

flutter test

🛣 Roadmap #

✅ Memory monitoring — RSS, leaks, lifecycle, rebuild tracking (v1.3)
✅ Persistent network store + unified event exporter (v1.3)
✅ Real-time FPS monitor, frame drop detection, API latency chart (v1.2)
✅ Shake-to-open, expandable bottom sheet, log export (v1.1)
❌ HTTP/2 + http package interceptor adapter
❌ Log inspector (debugPrint / Logger)
❌ Per-host throttling visualizer
❌ GC pressure & heap breakdown charts

📄 License #

MIT — see LICENSE.

pulse_ops 1.3.0 pulse_ops: ^1.3.0 copied to clipboard

Metadata