Document sources
A document source is one PDF (or DOCX) owned by the workspace: upload it once, tag it with a file category, and reference it from any number of templates and automations; each file category holds at most one source.
Sources live at the workspace level (peers to templates, not nested inside them), so one upload of a cover sheet, terms page, or standard form serves every template and automation in the workspace. A source is the upstream half of most template work: the blank or partly-filled PDF a template builds its pages on, and the file an automation points at when ingesting recurring documents from Gmail or Drive.
A source is exactly one file. A file category (document "type") holds at most one source: the source fans out to as many templates and automations as you like, but you never put a second source under one category. A genuinely different document gets a different named category with its own source. (New here? See core concepts.)
What a source bundles
| Part | What it does |
|---|---|
| Name | How it appears in pickers and tabs (e.g. Bid Invitations) |
| Page images for the canvas plus the original for high-fidelity rendering and form filling | |
| File category | The most important setting: it declares the input category of every template that references it |
| Linked automations | Optional: automations that ingest documents of this kind |
On upload, Document Blueprint parses built-in (AcroForm) fields (text, checkboxes, radio groups, dropdowns) so templates can fill them directly (right radio selected, checkboxes checked, dropdown values chosen). Encrypted PDFs (common with government forms) are decrypted on upload so fields stay fillable.
Where to manage them
On the Templates page, the left sidebar has a DOCUMENTS / TEMPLATES switch: pick DOCUMENTS for the Document Types catalog. Each active type renders as a row (colored chip, the template that produces it, a file count) with its one source folded underneath. Each row carries a pencil (rename the category in place: chip becomes an inline field, Enter saves, Esc cancels, and only the display name changes since the ID stays fixed so all references keep working) and a trash (delete the category). You can also rename a category from the consuming template's identity dropdown.
Clicking a source row opens a split-view modal: full PDF preview left, details (category, linked automations, linked templates) right. From there: Edit source (rename, reassign category, change color); Replace PDF (in the ⋯ menu); Open original PDF; Delete source (confirmation dialog); plus clickable Automations / Templates chip rows with + Automation and + Template. If a type is deleted, its source moves to an Unassigned sources group for re-categorizing.
Uploading a source
The panel pre-fills the category from the type you started in. Name is optional; the category tag is required. Typing a new category name creates it; typing an existing name (case-insensitive) reuses it via find-or-create, never a duplicate. Second entry point: the template editor's document bar + Add menu has Upload new file, which uploads to the workspace and adds the source to the template in one step. See source tabs.
Replacing a source's PDF
For a new revision of the same document, don't re-upload. Open the source detail, then ⋯ → Replace PDF. The file swaps in place on the same source (same name, category, ID), so every reference keeps working. The swap is silent with no page-count guard, so page count can change; existing field placements stay on their old coordinates. To shift placements to a new layout, ask the chat assistant to reposition them.
How references work
A template points at sources through source tabs in its document bar: each tab references a workspace source and contributes its pages. The reference does double duty: (1) visual base, where fields are placed on the source's pages; (2) routing, where the source's file category becomes one of the template's input categories, which makes the autofill picker offer the template for files of that category. Because the category lives on the source, changing it instantly re-routes every referencing template, with no per-template re-tagging.
Automations link indirectly: enabling Extract data with a template in the flow builder's Extract Data step and picking a template links the automation to that template's source (automatic when the template has exactly one). Ingested files are tagged with the source's category and annotated with their origin. See managing automations. The catalog shows links both ways: a source's detail lists linked automations and templates as chips, and the + Add picker shows a lightning-bolt chip on any source that an automation feeds.
Deleting a source
Deletion asks for confirmation (Delete Source, "This action cannot be undone") and is deliberately orphan-tolerant: referencing templates keep their now-dangling tabs (shown as Missing source), the canvas renders blank pages where those pages would have been, and nothing is auto-repaired or flagged. It's a normal in-progress state. To repair, open the template, remove the dangling tab with its ×, then add the replacement from + Add and drag it into position. There is no in-place "re-point" control; repair is always remove-then-re-add.