Backend slots (velo-text/api-report)

Optional Node/server entry. It does not import the editor, PDF/ODT/DOCX writers, or any npm dependency. The browser must not load this module (and must not load velo-text/export or velo-text/backend) if you care about payload size.

We will not add Puppeteer, LibreOffice WASM, pdf-lib, or anything over ~2 MB gzip. Fill and export stay on the AST + our writers.

The generated catalog of public names is API report. This entry is the slot map for those node types.

Front vs back

App	Import
Editor UI	velo-text/vanilla or velo-text/editor-web + theme CSS
Fill tags + export	velo-text/api-report + velo-text/export + velo-text/backend

Do not import from "velo-text" on the server if you want to keep the editor out of the graph; the main barrel still re-exports the host.

Tags

Same identifiers the editor already stores:

Slot kind	tag
variable	path (customer.name)
image / inline-image	assetId
table	repeat.path if present, else node id
table-repeat	repeat.path (collection for cloned rows)
equation / equation-block	node id (latex on the slot)
columns	node id

pointer is a JSON-pointer-like path into the AST (/root/children/0/...) so a host can show “this field is the first paragraph variable”.

Fill

import { reportSlots, dataFromSlotValues, assetsFromSlotValues } from "velo-text/api-report";
import { exportDocument } from "velo-text/export";
import { createBufferSink } from "velo-text/backend";

const slots = reportSlots(document);
// slots → your form / job: values keyed by slot.tag

const data = dataFromSlotValues({
  "customer.name": "Ada",
  items: [{ name: "Widget" }],
});
const assets = assetsFromSlotValues({
  logo: { mediaType: "image/png", data: pngBytes },
});

const { sink } = createBufferSink();
await exportDocument({ document, data, assets, format: "pdf", sink });

The pixel-true preview is previewPdf (same exportPdf as the API). reportSlots only lists inject tags.