> 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.

# Web app guide

Use the Umnai Decision Intelligence Platform to connect source data, prepare datasets, train and deploy models, and inspect their predictions and explanations. This guide takes you through each task in the web interface, with the settings, checks and example results needed to follow the workflow.

Follow the numbered steps to complete a task. Use the accompanying screenshots to locate controls and compare the example results with your own resources. Replace example names, paths and values with those appropriate to your account and data.

## Guide areas

The guide covers storage connectors, schemas, ingestion, dataset onboarding, workspaces, training, evaluation, deployment and explanations. It describes the resources you create, how they relate to one another and how to check their status. For programmatic access and advanced configuration, use the [REST API guide](/guides).

| Area            | Covers                                                                                                                     | Start with                                                    |
| --------------- | -------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------- |
| Setup           | Sign-in and dashboard navigation.                                                                                          | [Sign in and explore the dashboard](/ui-guides/setup/sign-in) |
| Data Lab        | Source preparation, storage connectors, processing compute, feature transform services, schemas, ingestion and onboarding. | [Data Lab overview](/ui-guides/data/overview)                 |
| Model Lab       | Workspaces, constraints, training, trained models and evaluation.                                                          | [Model Lab overview](/ui-guides/models/overview)              |
| Serving         | Deployment configuration and deployed models.                                                                              | [Serving overview](/ui-guides/deployment/overview)            |
| Explanation Lab | Local explanations, attributions, rules, dependencies and generated interpretations.                                       | [Explanation views](/ui-guides/explanations/overview)         |

The sections follow the application's main areas. Read each overview to understand the resources and their relationships before following the app procedures. Within Data Lab, use Ingestion to import source data and Onboarding to prepare datasets. Within Model Lab, use Training to configure and run training jobs. Model evaluation requires a running deployment: deploy the model in Serving, then return to Model Lab to evaluate it. For the underlying model and explanation terminology, see [Concepts](/concepts).

## Before you begin

Read the prerequisites before opening a create form. Review units carefully: the ingestion creation form uses seconds, while the onboarding and training creation forms use minutes. Existing job records may display seconds together with a conversion to minutes.

## How resources fit together

| Resource                                               | Saved choices or output                                 | Used by                        |
| ------------------------------------------------------ | ------------------------------------------------------- | ------------------------------ |
| Storage connector and schema                           | Source location and field definitions                   | Schema retrieval and ingestion |
| Ingestion configuration and processing compute         | Interpretation of source data and processing machine    | Ingestion job                  |
| Datasource                                             | Imported data                                           | Onboarding job                 |
| Onboarding configuration                               | Prediction task and preprocessing choices               | Onboarding job                 |
| Dataset                                                | Prepared training data                                  | Training job                   |
| Workspace, training configuration and training compute | Project context, learning settings and training machine | Training job                   |
| Trained model                                          | Output of training                                      | Deployment                     |
| Deployment configuration                               | Serving resource settings                               | Deployment                     |
| Deployed model                                         | Running serving resource                                | Evaluation and explanations    |

A configuration saves reusable choices. A job executes those choices and produces or updates a resource. Check the output status and references before moving to the next task.

## Example context

The examples use Adult Income and payments fraud data. They illustrate different parts of the workflow and do not represent a single processing lineage. Before running a job or interpreting a result, confirm the selected dataset, model and deployment and follow their resource references.

Some screenshots focus on part of a larger page; scroll within the application to see the remaining controls. Names, settings and results in a screenshot are examples. Check the values in your own account before continuing.