Add Reference Documents to a Project

This guide explains how to attach reference documents to a project — the material that describes what your data means, and the spreadsheets and CSVs a data quality report can assess.

Overview

Reference documents are the supporting material for an integration project — the documents that describe what the source data means, as opposed to the data itself. Typical examples:

Document type	What it describes
Data dictionary	The field names, types, and meanings of a source or target system
API specification	An API contract, such as an OpenAPI or WSDL specification
Sample extract	An example export that shows the shape of real data
Requirements	Client or regulatory requirements the delivery must satisfy
Other	Any supporting context that doesn't fit the categories above

Attaching these documents gives the platform the business context it can't infer from the raw data alone.

Where to find it

1. Go to Operate → Projects and open your project
2. Select the Data sources tab
3. Find the Reference documents card

Upload a document

1. On the Reference documents card, select Upload
2. In the Upload reference document dialog, choose a file — PDF, XLSX, DOCX, MD, or CSV
3. Choose the Document type that best describes it (see the table above)
4. Choose what the document applies to — the whole project, or one of its connected systems (each listed by its role). The per-system choice appears once the project has connectors linked.
5. Select Upload document

The document appears in the list on the card, showing its file name, type, and size, grouped by what it applies to. Anyone with access to the project can download a document with its download button; if you have edit rights, you can also delete a document.

What a document applies to

The applies to choice tells the platform which system a document describes:

Applies to	Use it for
Entire project	Context that holds for the whole delivery, such as requirements
A connected system	Material describing one linked connector, listed by its role — a source's data dictionary, the target's API specification, or an other system's notes

The choice records which system the document describes, so the delivery team can tell a source's data dictionary from the target's API specification at a glance.

A document keeps its association

If the source or target a document applies to is later removed from the project, the document stays in the list under Removed system — and reattaches automatically if you add the same system again.

Uploading a file to assess for data quality

When you upload a spreadsheet (.xlsx) or CSV, you can turn on Assess for data quality. The file is added as a reference document as usual, and also appears on the Data quality tab, where you choose the checks it is held to and generate its data quality report.

Removing the report on the Data quality tab keeps the document. When you delete an assessed document, the confirmation asks whether to remove its data quality report as well — leave Also remove its data quality assessment unticked to keep the report.

Check what the report reads from a document

Select the preview button on a document to see the text the report reads from it. Use this when a report says something about your source that doesn't look right — the preview shows whether the platform read the document as you'd expect.

Each document shows one of:

What you see	What it means
The document's text	This is exactly what the analysis works from
Sample of a large document	The document was long, so the report reads a representative sample rather than every row or page
Not read automatically	The file type isn't one the report reads. It's still listed by name and type, and an AI assistant connected to the platform can read the original in full
Could not be read	The file is password-protected, damaged, or a scan of pages with no selectable text
No text stored yet	The document was uploaded before previews were available. The report still reads it — upload it again to see its text here

Choosing the right type

The document type isn't just a label — it tells whoever picks the document up how to read it. A file marked Data dictionary carries authoritative field definitions; one marked Sample extract is example data. Picking the right type keeps the project's material navigable as it grows.

Related guides

• Projects — what a project is and where reference documents fit
• Profile a SQL Server Database — survey a source database
• Generate a Data Quality Report — assess a profiled source, or a document flagged for assessment