Read documentation for your task

Start at /llms.txt. Each entry links to the machine-readable version of an ordinary documentation page. Follow the relevant page links and reuse documents already read in this task. Links between pages in a machine-readable version lead to the linked page’s machine-readable version.

Choose a reading path

Task	Start with	Read next when needed
Build a dashboard	Data App developer guide	The chosen layout, style, components and chart references.
Build a prepared demo	Demo quick track	Reusable starter and app verification.
Restyle an existing app	Choose a style	Theming, changed layout or operations.
Explore data	Semantic layer workflows	Business questions and relevant query contracts.
Edit a semantic layer	Semantic layer workflows	Entity authoring, validation and publication guides.

For every free-form App build, follow the question coverage and capability checklist. For a new dashboard, establish the data before choosing one layout and style. Read the guides for the layout, style and components you select. For a restyle, preserve the verified setup and query meaning; revisit chart selection when chart forms change.

Read dependencies and limitations beside the API you use. A URL fragment selects a reading position; it does not reduce the transferred document. Load example images when appearance is part of the task; you do not need every catalog image for an implementation change.

Reuse context and keep required checks

Reuse a verified project’s integration, layout and data evidence. Recheck affected contracts when the installed package, model, hosting route or requested behavior changes. Query results are evidence for their captured scope and time, not a cache of current live data.

Selective reading must preserve authentication boundaries, live metadata, complete query meaning, permission checks, operation-result handling and appropriate app verification. Reading a guide does not authorize editing or publishing a semantic layer. Keep unavailable and unverified capabilities visible rather than removing them to meet a timing goal.

A successful local build does not establish live data correctness.