> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.umnai.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.umnai.com/_mcp/server.

# Overview

Views generate structured outputs from datasources, datasets, models, and deployed models.

Use them to inspect data, understand model structure, query deployed models, and explain predictions. Each view is generated from the artefact it belongs to.

For deployed models, views are also the inference interface: generating a deployed model view can return predictions, explanations, diagnostics, or evaluation outputs depending on the selected view.

## How views work

Choose the datasource, dataset, model, or deployed model you want to inspect.

Select the view that matches the question you want to answer.

Call the view endpoint for that artefact. Some requests return immediately, while larger requests may be accepted for asynchronous processing.

View responses use a shared structure, with view-specific data inside the response payload.

## Choose a view

| View family                                     | Use                                                                                                |
| ----------------------------------------------- | -------------------------------------------------------------------------------------------------- |
| [Data views](./data-views/)                     | Inspect datasource and dataset lineage, distributions, and data readiness.                         |
| [Model views](./model-views/)                   | Inspect the structure of a trained model.                                                          |
| [Deployed model views](./deployed-model-views/) | Query deployed models and generate predictions, explanations, diagnostics, and evaluation outputs. |

## Generate one or more views

View endpoints can generate one or multiple views in a single request.

The platform supports view generation for:

| Artefact       | Endpoint                                        |
| -------------- | ----------------------------------------------- |
| Datasource     | `POST /datasources/{datasourceId}/views`        |
| Dataset        | `POST /datasets/{datasetId}/views`              |
| Model          | `POST /models/{modelId}/views`                  |
| Deployed model | `POST /deployed-models/{deployedModelId}/views` |

A successful request can return the generated view directly. If the view is accepted for asynchronous processing, the response includes a `Location` header that points to the endpoint where the result can be retrieved.

## Response structure

Most views follow a common response structure built around tabular data.

The exact fields depend on the view, but the shared structure makes view responses easier to parse, display, and compare across workflows.

See [Response structure](./views-standards) for the common response format.

## Caching and freshness

Views may be cached to improve response time when the same view is requested again.

Most requests should leave cache control unset. Use `X-Views-Cache-Control` only when you need to control freshness or prevent a generated view from being stored.

| Header value         | Effect                                                           |
| -------------------- | ---------------------------------------------------------------- |
| `no-cache`           | Bypasses the cache and generates the view again.                 |
| `no-store`           | Prevents the request or response from being stored in the cache. |
| `no-cache, no-store` | Generates a fresh view and does not store the result.            |