This page explains how AIO links each trace to the data it read and wrote, and where you see the quality of that data in the UI.
What it is
An agent's answer depends on the data behind it. AIO records which tables, files, and endpoints each trace read and wrote. It matches them to assets in the Acceldata catalog (ADOC) and shows each asset's data quality (DQ) policy results next to the trace. When an answer looks wrong, you can check whether the agent read data that was failing a policy.
Each item a span reads or writes is a data reference. A data reference has:
- A namespace and a name. For example, the namespace
postgres://orders-db.acme.com:5432and the nameorders.public.customers. - A direction: read or write.
- A kind: table, view, file, endpoint, or collection.
How it works

- Your code produces data references. A reference reaches AIO in one of two ways, and the UI labels each one with its origin:
- Detected: the agent queries a database through a supported database instrumentation, so the span carries the SQL statement. AIO parses the statement and records one reference for each table it uses.
- Declared: your code callsdeclare_assetsto name data the SQL can't show, such as a file or an endpoint.
To set up either one, see Record the data your agent uses.
2. AIO matches each reference to a catalog asset. AIO processes new spans in the background. By default it checks for new spans every 30 seconds and waits 60 seconds for late spans to arrive. It then looks up each reference in the catalog. The lookup finds one asset, several assets, or none. If AIO hasn't asked the catalog yet or couldn't reach it, the reference stays unchecked for now.
3. AIO shows the quality of each matched asset. When you open a trace or session, AIO reads each matched asset's DQ policy results from the catalog and sums them up as a status.
4. You see the assets in the lists and on detail pages. The Assets column appears in the lists. The Assets used panel appears on a trace or session.
See assets in the lists
The traces, sessions, and steps lists have an Assets column. It's visible by default and isn't sortable. Each cell shows:
- A count of the assets.
- Their short names, which are the last part of each dotted name. For example,
customersfororders.public.customers. The tooltip shows the full names. - How many assets were written, when any were.
The cell shows "—" when the trace has no references, while the column loads, or if it fails to load.
On the traces list, the Assets filter narrows the list by data access, for example to traces that called tools but recorded no data. For more on filtering, see Explore traces.
See the assets a trace or session used
The trace detail page and the session detail page show an Assets used panel. It appears only when the trace has data references, and it's open by default.
The panel starts with a summary such as "3 read, 1 written, 1 failing". Under the summary, a line shows the time the quality results are from.
Each row shows:
- read or write pills.
- The asset's name and namespace.
- Origin pills: Declared, Detected, or both.
- The number of spans that touched the asset.
- The asset's match result or quality status.
Select a row to open the first span that touched the asset. This also clears any zoom on the waterfall.
Read the match result
When a reference didn't match exactly one asset, the row shows why. The following table describes each message.
| Row text | Meaning |
|---|---|
| not in catalog | No catalog asset matched the reference. |
| several catalog matches | More than one catalog asset matched. |
| catalog not checked | AIO hasn't looked up the reference yet, or couldn't reach the catalog. |
Read the quality status
For a matched asset, the row shows a quality button. It has a status pill and a summary of how many policies are passing and when they were checked. The following table describes each status.
| Status | Meaning |
|---|---|
| passing | The policy's latest result was successful. |
| warning | The policy's latest result was a warning. |
| failing | The policy's latest result was an error. |
| couldn't run | The policy's run errored or was aborted, or it returned some other result. |
Expand the quality button to list each policy. Each one shows its status, rule name, policy type (for example, "Data cadence"), score to one decimal place, and how long before the read it ran (for example, "12m before"). Assets with a failing policy start out expanded.
A row can also show:
- no policies: the asset has no DQ policies. To find and fix these assets across a project, see Check quality coverage.
- checking quality…: the results are still loading.
- quality unavailable: AIO couldn't get the results from the catalog.
The panel loads references from the first 1,000 rows only. When a trace has more, the panel shows "Showing the references from the first 1000 rows. Later ones are not loaded."
When quality isn't shown
If the catalog request fails, the panel shows "Data quality is not shown:" followed by the reason, for example "Data quality is not shown: Unable to connect to catalog". The message "Unable to connect to catalog" means the catalog couldn't be reached. Any other reason is the error message the catalog returned. A catalog failure affects only the quality results. The rest of the page still works.
The quality results you see depend on what the catalog lets you see. If the catalog refuses the request, the panel shows the catalog's error message.
See assets on spans
The trace waterfall also shows data references:
- Waterfall rows: a span that read or wrote data shows a pill with its number of assets. The pill shows as failing, with the title "An asset was failing a policy when read", when one of its assets had a failing policy.
- Span detail pane: an Assets block, such as Assets (2), lists each asset with its direction, name, and origin. The origin's tooltip says how the reference was captured. Under each asset, the pane shows the namespace and then one of these: the match result, the quality status with its summary, or "· in catalog".

Have a suggestion?