Case Dashboard

The Case Dashboard is where a screened caseload turns into work: the four numbers at the top tell you how the context stands, the list shows case by case, and one click opens the report. As of the current version those four numbers are no longer decoration — each total is a filter. Clicking Anomalias lists only what has an anomaly; clicking the same card again brings everything back.

Open the dashboard from Workspace›Painel de Processos.

The totals are the filter

Four cards sit at the top of the screen. Each shows the count for a situation and, when clicked, applies that situation to the list below:

Two rules always hold:

• Clicking again clears it. The lit card is the filter in force; clicking it once more removes the filter and the list goes back to the whole caseload. You are never stuck inside one situation.
• "TOTAL" is the total, not a fourth situation. It counts the entire context — compliant, anomalies and pending added together. That is why clicking TOTAL always yields the full list, whatever the previous filter was, and why it shows as lit while no other filter is applied.

The card in force is highlighted with the platform accent colour, and each number keeps its usual colour — red for anomalies, green for compliant, amber for pending.

The four numbers are counted on the server, over the whole context — not over the page you are looking at. They add up: compliant + anomalies + pending is the TOTAL, on a caseload of any size. And the number on the card is exactly how many rows clicking it returns: clicking Anomalias with 37 on the card gives you a list of 37 cases.

With a situation filter lit the counts do not shrink to the filtered subset: the four cards keep showing the full breakdown. That is what lets you switch situation with a single click — you keep seeing how many compliant and how many pending cases exist while looking only at the anomalies.

The other filters in the bar (Empresa, Tipo, Tipo de Erro, Responsável, Pessoa, Número) do apply to the cards: with "Empresa: Aurora" in force, all four count Aurora only, with or without a situation lit.

The pagination total, right above the list, always follows the query in force, situation included — which is why it equals the lit card.

The totals follow the caseload: when a case is deleted, or when screening changes a case's verdict, the four numbers are recounted on the next load of the list — you do not have to reload the page or touch the filters to see them right.

The filter bar

Seven fields sit below the totals, and all of them really filter. What you choose goes into the query sent to the server: list, total and pagination all come out of the same slice. No field touches only the page already loaded — that would make the screen say "3 results" with 300 in the caseload.

The seven combine with each other: filling in more fields narrows the result, never widens it. "Empresa: Aurora" + "Status: Anomalia" + "Responsável: you" lists Aurora's anomalies that are in your hands.

Cards and the Status field are the same filter, not two lookalike controls:

• picking "Anomalia" in the Status field lights up the Anomalias card;
• clicking the Conformes card switches the Status field to "Conforme";
• clearing either one clears both.

In a narrow space the bar starts collapsed

The panel is the same one at any width — full screen, inside the Chat side panel, or on a phone. What changes is the presentation, and it responds to the width of the panel, not of the window: a 600px panel on a 1440px screen lays itself out as a narrow space, because that is what it is.

There the filter bar sits behind a Filtros button, closed. Seven stacked fields would push the list below the visible area, and whoever does not scroll concludes there are no cases. A dot next to the button signals that a filter is applied, so a short list is never left unexplained; the chips just below still name them. On full screen the button does not exist and the bar is always visible.

Nothing else changes: clicking a row opens the case, the cards still filter, and paging is still done by the server.

Empresa, Pessoa and Responsável: type it or pick it

Those three fields suggest values as you type. The suggestions come from what has already appeared in the loaded pages — a shortcut, not the reach of the filter: the text you type is searched across the whole caseload, even for a company that never showed up on screen. So type a fragment of the name and let the list reload.

Responsável is searched by e-mail, which is what the assignment records. Picking the person from the suggestions fills the e-mail in for you; the active filter summary shows their name.

Tipo: what the case number tells

The type is not typed in anywhere — it is read from the case number itself, in the Brazilian CNJ standard, and it is the same value the list column shows when a case has no class recorded. The options are therefore exactly the types a number can encode: Cível (CPC), Penal (CPP), Trabalhista (CLT), Eleitoral (ELE), Militar (MIL), Estatuto da Criança e do Adolescente (ECA) and Juizado Especial Federal (JEF).

Active filters and how to clear them

As soon as a filter takes effect, an Filtros ativos strip appears right below the bar: one chip per applied filter, showing the chosen value. Clicking a chip removes that filter; the Limpar tudo button at the end of the strip brings the whole caseload back at once.

It answers the question a short list always raises — "are cases missing, or are filters left over?". With the strip in sight you see the slice in force without re-reading all seven fields.

Keyboard, screen reader and mobile

Each total is a real button, so:

• Tab reaches the four cards in the order they appear; Enter and Space activate the focused one.
• The screen reader announces the label, the count and whether the filter is applied.
• The Filtros ativos chips are buttons too: the screen reader announces "Remover o filtro" with the value, and on mobile they get a comfortable touch target.
• On mobile the cards stay readable with a comfortable touch target, and the page never gains horizontal scrolling — the list becomes cards, one field per line.

The workspace decides whether there is anything to show

The dashboard reads cases, and a case only exists in a context of type Processo Judicial (judicial case). So what shows up here depends on the workspace active in the picker at the top:

• "Todos os workspaces" (all workspaces) — the dashboard uses the installation's case contexts, and the context bar offers all of them.
• A workspace that has a case context (a Processo Tributário workspace, for instance) — the dashboard queries its case contexts. If the workspace mixes domains, the contexts that hold no cases are left out: they would never have anything to show here.
• A workspace from another domain (FEBRABAN, for instance, with onboarding and credit contexts) — the dashboard shows nothing, and says why. No totals, no filters, no list: the caseload does not belong to that workspace, and showing it there would make you read those cases as belonging to the domain you have open.
• A workspace with no context associated at all — same empty screen, different message, because the way out is different: there is no point looking for a case context among zero contexts. The notice tells you to associate a context with the workspace under Sistema›Workspaces, or to go back to "Todos os workspaces". This is the most common case in practice: a workspace created and not yet linked to any data context.

Switching workspace with the dashboard open updates the screen right away — no need to leave and come back. Entering a workspace without cases clears the list and the totals along with it; going back to one that has them re-runs the query.

An older installation, whose contexts do not declare a type yet, keeps seeing the dashboard exactly as before. Without the declared type there is no way to state that the workspace has no cases, and the platform would rather show than hide on an assumption.

Workspace contexts

When the workspace you opened the dashboard from has more than one case context, a context bar appears above the totals. Switching context clears the situation filter: the totals start counting the new context's caseload and the list goes back to showing everything. This is deliberate — keeping the card lit would show one context's breakdown next to another context's list. If you want the new context's anomalies, click the card again, now over the right numbers.

Assign an owner

In a case's detail you choose the owner — the person answering for that case. The assignment lands where the case actually lives: the platform locates the case in the context it was loaded into and records the owner there, with the date they took over. A case has one owner at a time; assigning it to someone else replaces the previous one.

What happens after you assign:

• The list's Responsável column shows the name from the next reload on.
• The BPM flow instances of that case inherit the same owner, and instances created from then on are born with it — that is how the Process Monitor can filter by person.

If the number does not exist in the context, the message says which context was searched, instead of a generic "not found": most of the time it is a case not loaded yet, a mistyped number, or a different context than you assumed. If the same number exists in more than one context, the platform does not choose for you — it asks which one the assignment applies to.

Being the owner is for follow-up and filtering. It does not change who can screen, complete a phase or delete a case — that stays with your profile's permissions.

Re-running triage for one case

Every row has a ↻ button that asks the platform to triage that case again — the report is rebuilt in full, over the whole case.

Triage does not run inside the screen: it becomes a task on the server, just like batch triage. In practice:

• The row shows a circular progress indicator while the evaluation runs.
• The task shows up under Background executions (the lightning icon in the top bar), named after the case number. Clicking the item brings you back to this panel.
• You can leave the screen or reload the page. Triage keeps running, and the panel picks the task back up when you return.
• When it finishes, the row updates its status, error type and date on its own.

If triage is already running in the same context — yours or someone else's — yours is queued instead of failing. The row indicator shows fila in amber, the background executions panel says how many are ahead, and it starts on its own as soon as the slot frees up. Order is first-come, first-served.

Two triages in the same context deliberately never run together: they would share the same AI request budget and neither would finish. The queue handles that instead of handing you the problem.

Cancelling

Rest the pointer for one second over the row's progress indicator: it turns into a red ✕. Clicking cancels.

Cancelling works in both states. If triage is running, in-flight AI calls are aborted — they do not keep burning quota until they finish on their own. If it is queued, it leaves the queue and never starts. Reports already written by a cancelled triage are preserved.

Re-running and cancelling triage require the AUDIT_EXECUTE permission, the same one batch triage uses.

While it loads, when nothing matches, and when it fails

Every filter change reloads the list with a loading indicator. Three different situations get three different messages, because the way out of each is different:

• The combination matched nothing. The dashboard says no case matches the filters, reminds you that they add up, and offers Limpar filtros right there.
• The context has no cases yet. The message explains that cases show up here once they are loaded, and suggests checking the context picked in the bar above — telling you to clear a filter that does not exist would make no sense.
• The workspace has no case context. There is no list or total here to explain: the screen shows only the notice, naming the workspace. The way out is the picker at the top — another workspace, or "Todos os workspaces" — not the filters.
• The query failed. Instead of an empty list with no explanation, a failure notice appears in plain language with a Tentar novamente button. No error code and no technical message ever reaches the screen.

Where to go next

• Click any row to open the case audit detail — report, rules met and violated, decision and grounds for appeal.
• To run screening over the caseload, see batch screening.
• To remove a case loaded by mistake, see deleting a case with justification.
• To build your own dashboards alongside this one, see platform dashboards.