# Data Platform Extension

Every Data Platform notebook ships with the **Data Platform extension**: a companion panel inside JupyterLab that brings the platform to your notebook. Browse the **documentation**, explore your **datasets and buckets**, run **read-only SQL**, open **ready-to-use notebooks** wired to your own data, and **export a notebook as a custom action** when it is ready to run as a job. Nothing to install, no credentials to configure: it works the moment your [notebook](/en/product/dpe/notebooks/index.md) starts.

![dpx-overview](picts/dpx-overview.png)

?> Browsing, previewing and querying are **read-only**: they never modify your data. The only things the extension creates are the notebooks you generate from the gallery, saved in your home folder, and the [custom actions you export](#export-a-notebook-as-a-custom-action).

* [Open the extension](#open-the-extension)
* [Documentation](#documentation)
* [Data: browse your datasets](#data-browse-your-datasets)
* [SQL: query without leaving Jupyter](#sql-query-without-leaving-jupyter)
* [Buckets: explore your files](#buckets-explore-your-files)
* [Notebook gallery](#notebook-gallery)
* [Export a notebook as a custom action](#export-a-notebook-as-a-custom-action)
* [Search everything](#search-everything)

---

## Open the extension

There are two ways to open it, depending on how you like to work:

- **Side panel:** click the **Data Platform icon** in the left sidebar of JupyterLab. The panel opens next to your notebook, so you can explore data and write code side by side. Use the **expand button** in its top bar to switch to a full-width tab.
- **Full view:** click the **Data Platform** tile in the JupyterLab launcher (under *Other*).

![dpx-side-panel](picts/dpx-main-panel.png)

The top bar is always available: a **global search** field, a **refresh** button that reloads the platform data (new tables show up immediately), and the **language** switch.

?> The extension is available in **English, French and Japanese**: use the **EN / FR / JP** switch in the top bar to change the language of the whole panel at any time.

---

## Documentation

The **Documentation** tab shows the live platform catalog: complete action scenarios, SDK snippets, a PySpark section and a FAQ. It is the same content that powers the helpers of the platform console, always up to date.

Every snippet has a **copy button** and a link to the full documentation page. Combined with the table pin (see [Data](#data-browse-your-datasets) below), the snippets are filled with your real dataset and table names: what you copy runs as-is in your notebook.

![dpx-documentation](picts/dpx-documentation.png)

---

## Data: browse your datasets

The **Data** tab lists the datasets of your project, the same ones you see in the [Lakehouse Manager](/en/product/lakehouse-manager/datasets/index.md). Open a dataset to browse its tables: the search stays instant even with thousands of tables.

![dpx-data](picts/dpx-data.png)

Expand a table to get everything you usually open the console for:

![dpx-data](picts/dpx-data-expanded.png)

| Action | What you get |
|---|---|
| **Schema** | Column names and types, in a compact scrollable grid |
| **Count rows** | The live row count of the table |
| **Preview** | The first 20 rows, with a **Copy for notebook** button that gives you the matching pandas code |
| **Copy as** | Ready-to-paste snippets: `connect()` call, SQL SELECT, filtered `select()`, pandas DataFrame |
| **Open in SQL** | Jumps to the [SQL tab](#sql-query-without-leaving-jupyter) with a query prefilled on this table |
| **Analyze (gallery)** | Jumps to the [gallery](#notebook-gallery) with this table preselected |
| **Use in snippets** | Pins the table: every documentation and gallery snippet is filled with its real name |

---

## SQL: query without leaving Jupyter

The **SQL** tab is a lightweight editor to query a dataset in read-only SQL. Pick the dataset, type a SELECT, press **Run** (or Cmd/Ctrl+Enter): the results show up right below, with the execution time.

![dpx-sql](picts/dpx-sql.png)

A few guardrails keep it safe and fast:

- only **SELECT** queries are accepted: nothing can be modified from here;
- a row limit is added automatically if you forget one;
- **Copy for notebook** on any result gives you the exact SDK code to reproduce the query in a notebook cell.

---

## Buckets: explore your files

The **Buckets** tab is a file explorer over the buckets of your project: navigate folders, see file sizes. Every file comes with a copy-ready snippet to read it from your notebook.

![dpx-buckets](picts/dpx-buckets.png)

---

## Notebook gallery

The **Gallery** is a set of ready-to-use notebooks, organized in packs. When you click **Create & open**, a pop-up lets you choose:

- **wire it to a real table of your project**: the notebook opens already connected to your data, ready to run top to bottom;
- **keep the placeholders**: you get a generic template to adapt later.

The notebook is created in your home folder and opens immediately as a JupyterLab tab.

![dpx-gallery](picts/dpx-gallery.png)

| Pack | Notebooks |
|---|---|
| **Getting started** | Data Platform 101: the guided tour of the platform on your project's real data |
| **Explore & assess** | Explore a table, Data quality report, Cross-dataset analytics |
| **Pipelines** | ETL table to table, Ingest an API into the DWH, Load bucket files into the DWH |
| **ML & Spark** | ML quickstart, Segmentation, Anomaly detection, Time-series forecast, PySpark starter |
| **AI (OVHcloud AI Endpoints)** | LLMs on your data, Talk to your data (SQL agent) |

The gallery adapts to your notebook image: the ML notebooks appear on **Data Science** and **TensorFlow** notebooks, and the PySpark starter on **PySpark** notebooks.

?> The **AI pack** uses [OVHcloud AI Endpoints](https://www.ovhcloud.com/en/public-cloud/ai-endpoints/) with your own API key. The rows you select are sent to the AI Endpoints API, so mind data sensitivity. See also [Jupyter AI Extension](/en/product/dpe/notebooks/notebooks-ai-extension.md) for the chat assistant.

---

## Export a notebook as a custom action

Once your notebook code is ready to run as a job, the extension can turn it into a [Custom action](/en/product/dpe/actions/custom/index.md) without leaving JupyterLab, and without copying code around.

![dpx-export-menu](picts/dpx-export-menu.png)

Start the export from the extension, then fill in the export form:

1. Select the **notebook** to export.
2. Pick the **repository** of the Data Processing Engine the action is exported to.
3. Choose whether to **create a new custom action** or **replace an existing action** with this export.
4. Give the action a **name**, and pick its type: **Custom Python** or **Custom PySpark**.
5. Click **Preview**.

![dpx-export-setup](picts/dpx-export-setup.png)

The preview sums up what will be created: target repository and version, action name and type, entry function, code size, and the dependencies detected from the notebook's **pip installs**, automatically converted into action dependencies. Below the summary you can review the generated code: non-code cells are dropped, and the code is wrapped in a `def customfunc(event):` entry function to match the action format.

![dpx-export-preview](picts/dpx-export-preview.png)

Happy with the result? Click **Create action**. The action is then available in the [Actions](/en/product/dpe/actions/index.md) screen like any other custom action: configure it, run it, or schedule it in a [workflow](/en/product/dpe/workflows/index.md).

![dpx-export-confirmation](picts/dpx-export-confirmation.png)

---

## Search everything

The **search bar** at the top looks across the whole extension at once: documentation snippets, datasets, tables you have browsed, and buckets. Results are grouped by type; clicking a table opens it directly with its details expanded.

![dpx-search](picts/dpx-search.png)