Upload documents

You can upload standalone PDF, Word, and PowerPoint files — project documents that didn’t arrive by email. ddx PMO reads each file, pulls out the risks, issues, decisions, actions, stakeholders, and charter sections it mentions, and files them in the right registers. The original file is kept and linked to everything derived from it.

Uploading

1. Open your project and go to the Import page (under the project’s ingest area).
2. Click the drop zone, or drag files onto it.
3. Drop one or more .pdf, .docx, or .pptx files. You can also drop a .zip containing a mix of documents, emails (.eml), and meeting notes (.md) — they’re sorted out for you.

Each file can be up to 200 MB, and you can select several at once. A progress bar shows each file uploading, then “Server is parsing + queueing classification”.

Note

Spreadsheets (.xlsx) are not accepted — there’s no text reader for them yet, so a workbook would be stored but invisible to the classifier. Upload one on its own and you get an error pointing you at the markdown-extraction route; if one is inside a .zip, the import summary lists it as skipped rather than storing a file nobody can read.

What happens to the content

Once a document is uploaded, ddx PMO extracts its content (reading PDFs natively — both layout and text — and Word/PowerPoint as text) and classifies it — works out what it’s about and which register each point belongs in. This runs in the background; the Classification status panel on the Import page shows files moving from queued, to classifying, to classified.

Because you uploaded it, a document is treated as a trusted source: what the classifier finds is applied straight into the registers, rather than waiting in the inbox for you to approve it the way emailed items do. The point is that importing, say, a charter actually populates the project — its sections, stakeholders and risks — without extra clicks.

Two safety checks can withdraw that trust for a file: if its text looks like it’s trying to manipulate the extractor (a prompt-injection attempt), or if a quoted “excerpt” can’t be found in the real document. If either fires, that file’s items fall back to needing your review in the inbox instead of auto-applying.

See the Glossary for artefact and register, and Evidence & relationships for how each extracted point links back to the source text.

Where uploaded documents go

• They appear in the project Inbox with a Document badge, and you can filter the inbox by the Document source.
• The file itself is kept and downloadable from the Attachments view, which lists how many artefacts were extracted from each file.
• From there, the attachment’s detail page shows the register artefacts the classifier extracted from it, so you can trace each point back to the document it came from.
• A charter you import also populates the Charter page.

Re-uploading and replacing

Uploading the same file again is recognised as a duplicate and skipped — files are matched by their exact contents, so nothing is processed twice. To force a file to be read again (for example after the system has been improved), tick Replace existing items before uploading; this destroys the matching items’ existing artefacts and re-runs the full pipeline for them.

Tip

Bringing in older documents as background? Tick This is historical data so long-overdue or years-old points are imported already retired — kept in the knowledge graph but out of My work and the live registers. You can un-retire any of them later.