User guide¶
The guide is organized around a successful project lifecycle:
- install an engine and run the smallest local quickstart;
- author metadata and choose a provider;
- construct a project runner with explicit runtime paths;
- run, replay or maintain a stage;
- validate logs and state before handing a build to an external deployer.
Choose a starting point¶
| Goal | Start here |
|---|---|
| First local pipeline | Installation, then the Polars quickstart. |
| Author metadata | Metadata guide. |
| Prepare, inspect, and build a project | CLI preparation walkthrough, then the CLI project workflow. |
| Run an explicit stage | Run a stage. |
| Use SQL files, replay or maintenance | File metadata, replay, and maintenance. |
| Operate a deployment | Platform recipes and smoke project and logging/troubleshooting. |
Project and runtime boundary¶
datacoolie.yml describes a project for the CLI. A runner is the executable
boundary: it chooses the engine and platform, constructs a Driver, supplies
providers and paths, and calls run, replay or maintenance operations. The
Driver does not read datacoolie.yml or a build manifest at runtime.
For exact precedence and ownership rules, use the runtime configuration guide. For the complete CLI contract, use the CLI reference.
Find a task¶
- Metadata authoring — section wrappers, sources, destinations, transforms, load strategies and validation.
- Providers — file, database and API metadata boundaries.
- Run, replay and maintenance — explicit Driver lifecycle actions.
- Platform recipes — choose a host, prepare the smoke project and verify output.
- Logging and troubleshooting — runtime output and failure diagnosis.