Noti Mobile-first training visibility
Python SDK + iOS + Android

See every run without being chained to your desk.

Noti turns TensorBoard-style training output into a mobile monitoring workflow: charts, progress, logs, images, and inbox alerts in one app, powered by the Notiboard Python package.

Your original TensorBoard event files are still written locally. Existing log_dir behavior stays intact, and your normal TensorBoard viewing workflow does not change.

Download on the App Store Get it on Google Play Download APK
Read the docs pip install notiboard
Dashboard Run synced
Instrument Classification8 runs
Current progress82%
Unread alerts3
Charts multi-run, smoothed
Inbox push + history
Checkpoint saved Epoch 4 snapshot written to disk.
Simulated training anomaly Gradient norm spike detected.
Charts

TensorBoard-style metrics, but usable from your phone

Scalars, histograms, images, text logs, smoothing, log scale, and multi-run overlays without opening a laptop.

Compatibility

Keep your local TensorBoard logs exactly as they are

Notiboard keeps writing normal local TensorBoard files through log_dir, so your existing tensorboard --logdir ... workflow still works.

Progress

Track live training progress in real time

Push progress updates from your Python job and keep ETA, run status, and checkpoints visible while you are away from the terminal.

Download Noti your way.

The app is the front door. The Python package sends your metrics. You need both for the full workflow.

iPhone / iPad

App Store

Install the mobile client, sign in, and create your API key in Settings.

Open App Store
Android

Google Play

Use the Play release for the default hosted API and standard update flow.

Open Google Play
Direct install

APK

Need a sideloadable build for QA or internal rollout? Grab the direct APK.

Download APK
PyPI Install the SDK with pip install notiboard.
Hosted API Production default: https://notiapi.tech-webs.com.
Account flow Email/password sign-in and API keys generated from the app.
TensorBoard Local event files are still preserved; your existing viewer flow stays the same.
Quick flow

What the setup actually looks like

  • Install the Noti app on your phone.
  • Create an account and generate an API key in Settings → API Keys.
  • Install notiboard into your Python environment.
  • Swap or wrap your TensorBoard SummaryWriter with NotiWriter.
  • Open Noti on your phone to watch runs, charts, and notifications arrive.
Start with the docs

Built for training loops, not demo slides.

Noti is designed around what real ML jobs already produce: TensorBoard-style scalars, histograms, images, text logs, progress counters, and completion events. You keep your existing workflow and add mobile visibility on top.

  1. Write locally, sync remotely.
    Your training loop still writes regular TensorBoard data to local event files. Notiboard mirrors the same run to the Noti server in the background.
  2. Stay informed without polling terminals.
    Progress updates and notifications show up in the app as your job moves forward.
  3. Inspect the run anywhere.
    Open charts, logs, histograms, and image outputs directly from your phone.

Drop-in TensorBoard migration

Replace SummaryWriter with NotiWriter, keep your existing scalar and histogram calls, then add progress or notifications where you want them.

Detailed bilingual documentation

The docs page includes English and Chinese, vibe-coding prompts for migration, manual setup steps, and a complete SDK method guide.

Fits hosted or self-hosted backends

Use the production API by default, or point the SDK to your own Noti server with the noti_server argument or NOTI_SERVER.