Platform UI tour

The RevoEngine workspace is an enterprise execution control plane. Its persistent navigation, global search, list-first grids, detail panels, dashboards, Chat, and AI Agents keep build, runtime, evidence, and access decisions inside one instance context.

​

Workspace anatomy

Actions are role-aware, but the API enforces access independently. A hidden button is a UI affordance, not the security boundary.

​

Filter and save a grid view

Use each column’s filter for its data type. JSON columns offer Simple property/value conditions and Advanced typed comparisons. In Simple mode, metadata.purpose addresses a nested property, items[0].code addresses an array element, and ["metadata.purpose"] addresses one literal key containing a dot. Choose the value type deliberately: the number 12 differs from the text "12". Equivalent spellings of the same path cannot be entered twice. Combine JSON conditions with All, Any or None, then apply the filter. Shared grid links retain typed paths and comparison settings, so reopen the link before using its selection for a bulk action. The grid footer joins the saved-layout selector with layout management and a Save action that appears while there are unsaved changes. Select a layout to restore it or use the management action to maintain saved layouts. These preferences change presentation, not resource access or data.

​

Confirm and follow an operation

Review the target and scope before submitting a consequential action. When a confirmation asks for text, copy or enter the displayed phrase exactly. Submission remains disabled until it matches; while the request is being accepted, the dialog prevents duplicate submission. An admission error remains visible so you can correct the issue and retry. An accepted background request is followed through its progress action or operation link. Completion and per-record failures are reported separately from acceptance. Check the retained operation before submitting the same work again after a lost response. A Database import can continue after its dialog is dismissed. Reopen it from the upload queue to recover the current draft, upload state and accepted validation reference. If an upload fails, its error remains visible; select the file again when requested instead of starting another validation blindly.

​

Build application logic

The normal authoring flow is open current Component → edit focused elements → resolve diagnostics → debug → review diff → deploy → verify the consuming runtime surface.

​

Model and inspect data

• Databases → Tables defines row-oriented application data and opens the data grid.
• Views and Materialized views expose governed query projections; materialized views can be refreshed by Schedule.
• Storage manages current file entries, folders, upload sessions, metadata, ACLs, retention, previews, and workspaces.
• Files remains the legacy explorer for existing fileId integrations.
• Cache Explorer exposes authorized cache inspection and management.

Use a Table or Storage object as authoritative state. Dashboards, search results, realtime frames, and previews are projections over that state.

​

Expose and automate

Separate definition from execution. Editing a Job template does not rewrite completed Job history; replaying an event uses current definitions but creates new lineage.

​

Observe operations

• Dashboards compose role-aware grids, charts, and Component-powered custom widgets.
• Job History owns attempts, status, timing, input/result visibility, logs, and retry lineage.
• Event History owns event payload and fan-out.
• Webhooks owns outbound attempt state.
• Activity history provides a cross-resource stream of changes and runtime events.
• Logs provides structured records; Trace correlates spans, operations, resources, and failures end to end.

Start with the user-visible failure, locate its operation, Job, or Event identifier, and move toward Trace. Do not diagnose from a dashboard count alone.

​

Govern identities and data

Secrets stores versioned sensitive values. Secret metadata can be discoverable to authorized operators while secret values stay out of normal logs, source, and documentation.

​

AI Agents and collaboration

• AI Agents → Threads provides Interactive Agent work for research, code, governed tools, approvals, and artifacts. The current thread and API object is named Assistant.
• AI Agents → Agents provides Autonomous Agent definitions, runs, sessions, Inbox, memory, plugins, and Storage workspaces.
• Chat supports direct/group conversations and a read-only system notification inbox.

These surfaces have separate persistence and access models. A Chat message is not automatically Agent memory, and an Interactive Agent draft does not become deployed Component code until the normal review and deployment path completes.

​

Example: investigate a failed order

1. Open the order-operations Dashboard and select the failed Job row.
2. Inspect attempt status, Component version, principal, and operationId.
3. Follow the originating Event History and confirm the message/filter.
4. Open Trace for the operation.
5. Inspect the database transaction, secret lookup, egress call, and Storage receipt spans.
6. Fix and debug the Component in Studio.
7. Deploy the reviewed element.
8. Retry or replay only after confirming idempotency and external state.

​

Navigation practices

• Confirm the active instance and environment before every consequential action.
• Select the instance breadcrumb to return to the instance landing dashboard; use the main navigation to choose another workspace.
• Save recurring views as favourites or dashboards.
• Keep list filters while opening details in the side panel.
• Share deep links containing the exact resource or execution ID.
• Treat Delete as a lifecycle action; prefer Restore over recreating duplicate definitions.
• Use Search for discovery and Trace/history for proof.

Continue with First production workflow for a complete balance, order, transaction, egress, and Storage example.