Manufacturing Reports

Most manufacturing teams rely on spreadsheets and one-off exports when they need a shared view of performance or operations. Someone pulls an ERP extract, builds a chart, and emails it around. A month later there are four copies of that chart, two of them stale, and no agreement on which numbers are right.

Manufacturing Reports brings those dashboards into Threaded. A report is a live layout that reads from datasets your organization maintains, so everyone opens the same report in the same place, and refreshing the data updates every report that uses it.

#Key Concepts

Manufacturing Report — A named report your organization can open from the Reports page. Each report has a title, an optional description, and an identifying slug. A status chip shows Draft until the report has a layout, then Published.

Shell — The interactive HTML layout that viewers open. This is the dashboard itself: charts, tables, filters, and the queries behind them. A report accumulates numbered shell revisions over time, so you can see what changed and go back if you need to.

Data source — A named, org-level dataset (currently SQLite) that any report can read from: an ERP export, a curated extract, or anything else your team maintains. Data sources are reusable — the same dataset can back several reports at once, and uploading a new revision updates all of them. Manage them from Reporting → Data sources.

Data source binding — The link between a report and the data sources it reads. A report can bind up to eight data sources at once, and it always reads each one’s current revision, so a refreshed dataset shows up in the report immediately.

Planning document — One Markdown document per report describing the charts, metrics, and queries the dashboard should use. It is where the intent behind a report lives, and it can be replaced at any time — before the first shell exists or long after.

Report artifacts — Shell HTML, the planning document, generated insights, and any auxiliary files are stored on the report itself, not in the general Admin → Media library. Datasets are the exception: they belong on the Data sources page and are bound to the report.

#Who can use it

Anyone in your organization can browse and open reports. Creating a report, uploading artifacts, adding shell revisions, and deleting reports require an edit role — Owner or Editor.

If your organization does not have Manufacturing Reports enabled yet, the Reports page explains how to get access.

#How It Works

#Opening the Reports list

Go to Reporting → Reports in the navigation to open the report library.

Each row shows the report title, a Draft or Published chip, the description when there is one, and when it was last updated. Click a published row to open the report viewer. Click a draft row to open Report details. Use the Open sidebar button to inspect details without leaving the list — the sidebar shows View report for published reports and View as full page to open Report details.

Use the star on the right of each row to Pin report or Unpin report. Pinned reports sit in a Pinned section at the top of the list (the same sort applies inside each section). The pinned section is hidden when empty or when a search query filters all pinned reports away. Pinning is per-user — it does not change the report or appear for other users. Anyone who can view reports can pin them.

Use Search reports to filter by title, and Sort by to order the list by name, creation date, or last updated. The list opens sorted by Last updated (newest).

Editors get Add Report above the list, which opens the Create report dialog. Enter a title (and optional description), then click Build with AI to create the draft, open Report details, and start the AI authoring workflow there.

#Reading a report

The report viewer fills the workspace so you can focus on the dashboard. Along the top you get the report title and description, Open in new tab for a full-window view, and a menu with Pin report (or Unpin report), Report details, and Delete report. Editors also get Build with AI on a draft (no shell yet) or Update with AI once the report has a shell — that is the main place to revise a report with the assistant.

Generate insights runs an AI pass over the report’s data and writes the result back as an insights artifact. It requires at least one bound data source and only runs against the report’s latest shell.

#Report details

Open Report details from the viewer menu, the sidebar’s View as full page, or by clicking a draft row in the list. It has four sections:

• Overview — title, description, slug, who created the report, and Last updated with the reason it changed, such as shell rev 4 or data source shopify v2. Editors can edit the title and description in place.
• Shell revisions — every revision with its file name and number, marked current and latest. View opens that revision, and Rollback… makes an earlier revision current again.
• Data source bindings — the data sources this report reads, shown as title and slug. Editors change them with the Bound data sources picker; Manage org data sources jumps to the Data sources page.
• Artifacts — every file stored on this report, with its role, revision, size, and upload date. Drag files in to upload, and use View, Download, Ingest shell, or Delete on individual rows.

Editors can Build with AI from the draft guidance alert when a report still has no shell. Use the viewer for ongoing updates.

#Working with revisions

Opening a specific shell revision with View puts that revision in the URL, so you can bookmark or share exactly what you are looking at.

#Building a report with the AI Assistant

You do not write the dashboard HTML yourself. Describe what you want to see and the AI Assistant plans the report, queries your data, and builds the layout. Creating a report with Add Report → Build with AI creates the draft, opens Report details, and starts the assistant there. To revise an existing report, open it in the viewer and click Update with AI (or Build with AI if it is still a draft).

The assistant reads your data sources through Threaded MCP, which connects on its own for most organizations. If it is not connected, the assistant will show you a Connect card in the conversation. See Connecting Threaded MCP.

#Query your data sources

Ask the assistant to inspect a data source’s schema or run read-only SELECT queries against it. There is no SQL console in the report UI — the conversation is the query surface. When a report reads more than one data source, tables are qualified by the source’s slug, for example shopify.orders.

Query results are how you decide what is worth charting. The finished dashboard runs the same queries in your browser when you open it.

Example:

“Show me the schema and a few sample rows from the shopify data source.”

#Author a report end to end

1. Create a draft with Add Report → Build with AI (opens Report details with the assistant), or ask the assistant to create one. To revise an existing report, open it in the viewer.
2. From the report viewer, click Build with AI or Update with AI, or ask the assistant in your own words to author the report. Drafts that open on Report details also have Build with AI in the draft guidance alert.
3. Answer its questions about your goals and the elements you want. It checks your existing data sources first, so there is no upload step if the data is already there.
4. Review the planning document it drafts. The assistant saves it as the report’s Planning artifact, then builds or revises the HTML shell.
5. Threaded sets the report’s data source bindings when the new shell is saved, reading them from the queries in the shell itself. Check the result under Data source bindings.
6. The assistant updates the planning document with notes on what it actually built. Open Artifacts → Planning → View to read it any time.

If your organization has no data sources yet, create one on the Data sources page first, or ask the assistant to design the report without live data and fill it in later.

Example:

“Create a production KPI report from the shopify data source.”

“Update this report to add a filter by site and refresh the planning doc.”

Bindings are kept in sync from the queries Threaded can read in the shell. If a report ends up reading a source that is not bound — or you want to bind one by hand — set it yourself with the Bound data sources picker on Report details.

For more example prompts, see AI Assistant Use Cases. For managing datasets directly, see Manufacturing Report Data Sources.

#Automating with the CLI

Day-to-day browsing and AI-assisted authoring happen in the app. The Threaded CLI covers automation and scripted workflows — uploading large files, managing datasets, or generating revisions outside a conversation. Report commands run as threaded task report:… and dataset commands as threaded task reports:data-source:….

#Why This Matters

Manufacturing Reports narrows the gap between “someone exported a chart” and “everyone trusts the numbers.” Numbered shell revisions give you a record of what the dashboard showed and when, shared data sources mean a single refresh updates every report that depends on them, and access follows the same organization and role structure as the rest of your process — not an email thread with a spreadsheet attached.