Get it on Google Play

How Wertek AI Studio works

BETA feature. Studio is enabled per account; today it's visible only to selected beta accounts. Ask Wertek AI about availability on your plan.

How it works (concept)

Wertek AI Studio is an engineering analysis lab inside the web platform: Python notebooks that run in your browser over your organization's data (energy, drives, vibration sensors, incidents). Nothing to install and no server to run — you open a cell, write Python, and it executes locally with scientific libraries already loaded (pandas, numpy, scipy, scikit-learn, matplotlib).

What makes it Wertek AI is the wertek bridge: an object available in every cell that brings your real data in as a DataFrame and can export a finding back to the platform (an alert rule, an event, a work order). In other words: explore the data and turn the result into something operational, in the same place.

What it is NOT: it doesn't run anything on your equipment, it can't see other organizations' data (each account is isolated), and it doesn't replace the diagnostics module or the CMMS — it's the exploratory layer on top of that data. Each cell has a 30-second execution limit (built for analysis, not heavy jobs).

First 15 minutes

  1. Open it: in app.wertek.ai, side menu → Studio (marked BETA). You'll see three tabs: Notebooks, Datasets, and Artifacts.
  2. Create a notebook: the + button, blank or from a guided template (Vibration Trend, Energy Anomaly, Correlation, Fault Classifier, RUL Estimation). Templates ship with sample data so you can see the flow; for your real data use the wertek bridge (below).
  3. Rename it: click the title at the top, or the pencil icon in the list. Export the whole notebook to .ipynb (opens in Jupyter/VS Code) with the Export button.
  4. Bring in your real data in a cell:
    import wertek
    df = await wertek.query("energy_readings", asset="TAB-PTAR-01", days=30)
    df.head()
    
  5. Plot what you loaded:
    wertek.plot_trend(df, "power_kw", title="Power — last 30 days")
    

The wertek bridge — what you can call in a cell

Every method that fetches data is async: use await. They return a pandas DataFrame.

Fetch data

  • await wertek.query(table, asset=None, days=30, columns=None, limit=10000) — query a source table directly. Ex: await wertek.query("vfd_readings", asset="AST_DRIVE", days=7).
  • await wertek.load_dataset(name) — load a saved Dataset (defined in the Datasets tab) as a complete table.

Analyze and plot (no await)

  • wertek.plot_trend(df, column, title=None) — time series (auto-detects the created_at column).
  • wertek.plot_spectrum(df, column, title=None) — frequency spectrum (FFT).
  • wertek.correlation_matrix(df, target=None) — correlation heatmap, or bars against a target column.
  • wertek.align_timeseries(df_sensors, df_labels, tolerance="5min") — join two series on nearest timestamp (to cross measurements with events).

Export to production (create a draft artifact)

  • await wertek.export_alert(name, asset=None, condition=None, severity="warning") — create an alert rule.
  • await wertek.export_iaes_event(asset_id, event_type="asset.health", severity="warning", diagnosis="") — publish an event.
  • await wertek.create_wo(asset_id, title, priority="medium", type="predictive") — create a pre-filled work order.

You also have pandas (pd), numpy (np), and matplotlib.pyplot (plt) available directly, no import needed.

Typical flow: from exploration to operation

  1. Fetch the data with wertek.query(...) (or save a reusable Dataset in the Datasets tab and load it with load_dataset).
  2. Explore: df.describe(), df.head(), plot trend/spectrum, look for correlations.
  3. Export the finding with wertek.export_alert(...) → it shows up in the Artifacts tab as a draft.
  4. Deploy it: in Artifacts, the Deploy button. A deployed alert rule becomes a real automation rule — a "View in Automations" link appears (Settings → Automations), where it lives and is edited like any other rule. Studio captures the trigger (condition/asset/severity); the actions (notify, create WO) you tune in Automations.
  5. Share: export the notebook to .ipynb, or a Dataset to CSV, from their download icons.

Common errors and fixes

  • "The cell stopped / the output vanished" — there's a 30-second per-cell limit. Narrow the range (days=) or limit=, or split the work across cells. It's not meant to crunch millions of rows at once.
  • "I forgot await and got a weird object (coroutine)" — the methods that fetch or export data (query, load_dataset, export_*, create_wo) are async: always await wertek.query(...).
  • "The DataFrame comes back empty" — the asset has no data in that window, or the table/asset name doesn't match. Verify the exact asset tag (Assets page) and try a wider window.
  • "I deployed an alert but can't find it in Studio" — that's correct: deploying an alert artifact turns it into an automation rule. Find it in Settings → Automations (the artifact itself shows the link).
  • "My rule fires on everything" — if you exported the alert without a condition, the rule becomes any condition. Edit it in Automations to narrow condition, severity, or asset.
  • "I don't see Studio in the menu" — it's a BETA feature enabled per account. If it's missing, email support to request beta access.

What's next

  • Save reusable queries: define Datasets in the Datasets tab (table + asset + range) and load them with wertek.load_dataset("name") in any notebook.
  • From analysis to a live alert: see how rules are managed in CMMS/automations and assisted interpretation in the Copilot.
  • Support: for beta access or questions, [email protected] or the platform chat — see general FAQ.

Ready to give your assets a digital identity?

Register unlimited assets. Manage what matters.