> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sphinx.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Ingest and train on data and documents

> Use Deep Update to train Sphinx on your data and documents, follow progress, and review proposed knowledge changes.

Use **Deep Update** to train Sphinx on your data and documents. Add files and instructions, and Sphinx prepares knowledge base changes for your review.

## Before you start

* Sign in to the workspace you want to update.
* Have the data files, documents, or instructions you want Sphinx to learn from.

## Start a Deep Update

**1. Open Deep Update**

Select **Ingest → New Ingest → Deep Update**.

**2. Describe the outcome**

In **Prompt**, explain what to add or change, who needs the knowledge, and which details to keep. For a correction, name the affected pages and explain the correct rule. See [Write a good ingest prompt](https://docs.sphinx.ai/ingest-prompts.md).

```text theme={null}
Here are our quarterly sales review decks, pipeline reports,
and leadership meeting notes. Learn how we present information
to our sales leadership team: the metrics we emphasize, how we
explain performance against targets, and how we communicate
risks, opportunities, and recommended actions.

Use this context to help our agents prepare future sales
leadership updates that follow our team's conventions.
```

**3. Attach your sources**

Drag files or a folder into the upload area, or use **Choose files** or **Choose folder**. You can also submit instructions without files to update existing knowledge.

Review **Notebook mode** and **Source timestamps** if those settings matter to your inputs. **Use file last-modified dates** is selected by default in the demo shown below; make sure those dates reflect the source history you intend to provide.

**4. Create the ingest**

Review the prompt and attachments, then click **Create ingest** once. Open the job to follow its progress.

![Deep Update form with Prompt, Choose files, Choose folder, and Advanced options](https://mintcdn.com/sphinx-977f1673/O9wCzUTcMLhLJ368/images/ui/deep-update.jpg?fit=max\&auto=format\&n=O9wCzUTcMLhLJ368\&q=85\&s=59d1bfdfa60f77d2588b0744a4d809bb)

Deep Update accepts instructions, files, or both.

## Follow a job

Open **Ingest** to see jobs and their statuses. Use the execution-status filter to narrow the list, then open a job for details.

![Demo account ingest list with synthetic data showing named jobs and suggested, accepted, rejected, and cancelled statuses](https://mintcdn.com/sphinx-977f1673/O9wCzUTcMLhLJ368/images/ui/ingest.jpg?fit=max\&auto=format\&n=O9wCzUTcMLhLJ368\&q=85\&s=8bd470ee9a364aebc3f62f86916fc470)

The Ingest list includes pending proposals and accepted, rejected, or cancelled jobs.

When the job offers **Review Changes**, open the proposal and [review its differences](https://docs.sphinx.ai/knowledge-base/review-changes.md). A **suggested** update is waiting for a decision; it has not been accepted into your knowledge base yet.

If a job fails or appears stuck, open its details and note the job ID and error message before contacting [Sphinx support](mailto:support@sphinx.ai).

## Train programmatically

Use [`sphinx_learn.fit()`](https://docs.sphinx.ai/python-sdk/fit) to train programmatically from Python. It uploads your documents and starts a training ingest, like a Deep Update with files used for training. See the [Python quickstart](https://docs.sphinx.ai/python-sdk/quickstart.md) for a first example.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.