Pipeline¶

Once a tool has produced canonical IR, the same core pipeline takes over.

Processing Stages¶

1. Validate¶

The IR is checked against the shared schema. This is where malformed JSON, invalid enum values, or structural mismatches fail early.

Importers are expected to provide canonical IDs and source metadata before this point.

2. Resolve Settings¶

The pipeline turns partial settings into a full resolved settings object using canonical defaults plus any inline or external config.

3. Group Artboards¶

Artboards are grouped into responsive sets when their names or input metadata indicate that they belong together.

4. Compute Render-Ready Values¶

This is where the core prepares the document for emitters:

• breakpoints
• responsive behavior
• text/style normalization
• asset path behavior
• emitter-ready layer structure

5. Emit¶

Finally, an emitter writes output files:

• HTML fragment
• Standalone HTML
• Svelte component
• React component

Export surfaces can then wrap the emitted files, ir.json, assets, warnings, and manifest.json into a common bundle.

The pipeline is shared so that importers and plugins do not each reinvent rendering.

Why This Split Is Useful¶

It lets you:

• add new input surfaces without rewriting emitters
• add new emitters without touching source-tool extraction
• inspect ir.json and debug failures at a stable boundary
• inspect manifest.json to understand what a bundle contains