The same ERP extract usually feeds more than one dashboard. Handle that by attaching a copy of the file to each dashboard and you have signed up for a monthly chore: re-upload it everywhere, and hope nobody misses a copy.
Data sources are datasets that belong to your organization rather than to a single report. You upload the dataset once, point as many reports at it as you like, and every one of them reads whatever version is current. Refreshing the data is one upload.
#Key Concepts
Data source — A named dataset (currently SQLite) owned by your organization. Each one has a Slug — a short lowercase identifier like shopify, used to reference it in queries — plus a Title, an optional Description, notes, and a current version.
Managed data source — A Threaded-built cache (for example your process data or submitted form records) that stays pinned at the top of the library. Title and Description are set by the platform. Managed caches show a Rebuild schedule and Process Run History instead of notes and version history, and cannot be deleted. Use Create or Rebuild when you want Threaded to refresh that cache.
Version — Every upload to a data source creates a new numbered version. Exactly one version is current at a time, and that is the one every report reads. A new upload becomes current immediately, so there is no draft or staging step.
Binding — The link between a report and a data source. A report can read up to eight at once. Threaded keeps bindings in sync with the queries in the report’s layout, and editors can adjust them by hand from the report’s Data source bindings section.
#Who can use it
Anyone in your organization can view data sources and their versions. Creating a data source, uploading a version, rolling back, and deleting require an edit role — Owner or Editor.
#How It Works
#Opening the Data sources page
Go to Reporting → Data sources in the navigation. The library is a table with the columns Title, Slug, Description, Build, Updated, and Update. Threaded-managed caches stay pinned at the top with a Threaded mark; your uploaded sources are sortable below.
Use the Open sidebar button on a row to inspect data source details. View as full page opens the same detail on its own page. Click a Title or Description cell to start inline editing (editors only). The View button on a title cell opens a preview of the dataset.
#Creating a data source
Click Add Data source and fill in the dialog:
- Enter a Title. The Slug fills in automatically — change it if you want something shorter.
- Add a Description noting where the data comes from, for example “ERP nightly dump.” Future readers will want this.
- Optionally use Add data file (optional) to upload the first version right away.
- Click Create.
Slugs must be lowercase letters, digits, and underscores, starting with a letter. Because the slug is the name your queries use, database keywords are rejected — order, select, main, and query will not work, so reach for something like sales_orders instead.
#Data source details
The sidebar and full page show the same information, which varies by source type.
Uploaded data sources:
- Title, Description, and Slug (copyable). Editors can edit Title and Description in place.
- Notes — optional working notes, including @mentions of other Threaded objects.
- Linked reports — chips for every report that currently binds this source. Click a chip to open that report’s sidebar.
- Version history — every upload, showing revision number and file name. Rollback appears on non-current revisions for editors.
Preview opens the dataset’s tables in a browser you can page through without leaving Threaded. Download saves the current file. Upload new version uploads a new revision. Delete data source permanently removes the source and all of its versions.
Managed data sources (Threaded-built caches):
- Title and Description (read-only, set by the platform).
- Rebuild schedule — the cron schedule with last and next run times.
- Linked reports — chips for every report that currently binds this cache.
- Process Run History — a log of each automated rebuild with status and result.
Preview opens the dataset’s tables. Download saves the current file. Create or Rebuild refreshes the cache. Managed caches cannot be deleted and do not have notes or version history.
#Uploading a new version
On a source you uploaded, use Upload in the library’s Update column or Upload new version on the detail page. The file becomes current as soon as it finishes, and every report reading that source picks it up immediately. For a managed cache, use Create or Rebuild instead of uploading a file.
#Rolling back
If a new version turns out to be wrong, use Rollback on an older revision in version history to promote it back to current. Nothing is re-uploaded — Threaded just points at the older file again — so the change takes effect for every report at once.
#Deleting a data source
Delete data source on the detail page permanently removes the source and all of its versions. Any report still bound to it will fail to load its data, so unbind it from those reports first if they need to keep working. Managed caches cannot be deleted.
#Binding data sources to a report
Most of the time you will not set bindings by hand. When a report’s layout is saved, Threaded reads the queries inside it and binds exactly the data sources those queries use. The result is visible under Data source bindings on that report’s Report details page, listed by title and slug — see Manufacturing Reports for the full authoring workflow.
Editors can override the bindings at any time with the Bound data sources picker in that section. Do that when a report needs a source Threaded could not detect from the layout, or when you are preparing a report before its queries exist.
#Querying data sources with the AI Assistant
Ask the AI Assistant to inspect a data source’s schema or run read-only SELECT queries against it before it writes any report queries. This runs through Threaded MCP, so that connection needs to be live — for most organizations it connects on its own.
When a report reads more than one data source, qualify table names with the slug, for example shopify.orders or threaded.parts. Each bound source is attached under its own slug when the report loads.
Example:
“What tables and columns does the shopify data source have?”
“Run a sample query against threaded joined with shopify on part number.”
#Why This Matters
Data sources turn “upload this file to every report that needs it” into “upload it once.” A single refresh brings every dashboard current, the version history tells you exactly which file a report was reading and when, and because bindings follow the queries a report actually runs, a report cannot quietly drift away from the data it claims to show.