# Start with Ptah

Install Ptah, get a verified result, choose a schema workflow, or adopt a database that already exists.

Source: https://docs.ptah.run/v0.8.1/start/overview/

import { LinkCard, CardGrid } from '@astrojs/starlight/components';

Start with the state you have, not with a command namespace. A disposable
database is the shortest route to a first result. A shared or existing database
needs a workflow decision before anything changes.

## Common starting points

<CardGrid>
  <LinkCard title="Install Ptah" href="../install/" description="Install and verify the native CLI, Atlas-compatible binary, and language server." />
  <LinkCard title="Get a first result" href="../quick-start/" description="Apply and verify one schema against disposable SQLite in about five minutes." />
  <LinkCard title="Create a first migration" href="../quick-start-migrations/" description="Write, seal, apply, and verify one versioned migration pair." />
  <LinkCard title="Adopt an existing database" href="../adopt-an-existing-database/" description="Inspect what exists before choosing a baseline or desired-schema workflow." />
</CardGrid>

## One decision before a shared database

Choose how changes reach the database:

- **Versioned migrations** keep ordered SQL files in the repository and replay
  them across environments.
- **Direct schema changes** compute and apply the difference from the desired
  schema without creating a migration file.

[Choose a workflow](../choose-a-workflow/) compares those paths and the cases
where a desired schema generates versioned migration files.

Persistent embedding and model output use a different lifecycle. Start with
[Inference migrations](../../inference/overview/) when the active data must stay
available while Ptah builds and verifies a candidate generation.

## Look up exact behavior

Use the [reference landing](../../reference/overview/) for generated command,
flag, configuration, format, rule, capability, and exit-code references.