The four layers of an Extend application
An enterprise Workday Extend application is rarely just a page. A production application spans at least four distinct structural layers, each with its own runtime behavior, configuration syntax, and failure modes:
- Presentation Model Definition (PMD): The user interface declarations, page layouts, component trees, data binding expressions, and client-side micro-scripts.
- App Model Definition (AMD): The underlying data architecture, Custom Business Objects, relationships, custom event definitions, and data types.
- Orchestration for Integration (O4I): Multi-step backend integrations running transformations, REST/SOAP outbound calls, routing logic, and error handlers.
- Tenant Security and Data (WQL): Domain Security Policies, Security Group memberships, and Workday Query Language data sources that feed information to the UI.
When an employee or manager encounters a failure in the browser, finding which of these four layers caused the breakdown is historically one of the most frustrating tasks in Workday development.
The debugging disconnect
In a conventional developer setup, debugging requires jumping across disconnected tools, downloading multi-megabyte log files, and manually correlating timestamps:
- Automatic Log Fetch: Queries tenant execution history via Workday APIs without downloading zip archives.
- Cross-Layer Trace: Maps the UI button click to the Orchestration trigger and downstream WQL query.
- Plain-Language Verdict: Isolates the exact step and parameter mismatch; no raw Java stack traces.
- Verified One-Click Fix: Edits the PMD or orchestration script, checks Workday compiler, and redeploys.
The LEAK rule: clarity over stack dumps
In many development environments, an error log is a 500-line Java stack trace ending with a generic message:
com.workday.orchestration.runtime.exceptions.StepExecutionException:
at com.workday.orchestration.steps.RestCallStep.execute(RestCallStep.java:184)
at com.workday.orchestration.runtime.PipelineRunner.run(PipelineRunner.java:92)
Caused by: com.workday.framework.core.ValidationException: Field format invalid.
[error_code: VALIDATION_ERROR_DATE_MISMATCH] This raw output confuses functional leads and forces developers into manual archaeology. kiweely operates under the LEAK rule: the assistant never passes unformatted technical stack traces directly to the person. It reads the trace, checks the step context, cross-references tenant field configurations, and delivers a clear verdict:
"The orchestration failed at Step 3 (Invoke_Worker_REST). The transformation script passed the worker'shireDateinMM/DD/YYYYformat, but Workday's REST API requiresYYYY-MM-DD. The preceding WQL query returned an unformatted date string because the query omitted the date formatting function."
An end-to-end diagnosis walk-through
Consider an Extend app that submits remote work stipend requests:
- The Observed Symptom: A manager selects an employee, inputs an allowance of $500, and clicks Submit. The browser displays a toast: "Submission failed. Check application logs."
- Consultant Ask in kiweely: "Why did the stipend submission fail for employee 10482 in the Dev tenant?"
- Agent Inspection:
-
kiweely reads the latest execution record for the
Stipend_Request_Pipelineorchestration. - It verifies that Step 1 (Trigger) and Step 2 (Get_Worker_Eligibility) succeeded.
- It discovers that Step 4 (Create_One_Time_Payment) failed with an HTTP 400.
-
It inspects the employee's compensation plan in the tenant: the worker is based in
Toronto with currency
CAD, while the orchestration hard-codedUSDin the request body.
-
kiweely reads the latest execution record for the
- Proposed Solution: kiweely proposes modifying the orchestration transformation step to dynamically pull the worker's primary currency code from the WQL data source.
- Verification: Upon approval, the agent updates the orchestration manifest, runs Workday Extend validation, deploys the package, and initiates a test transaction to confirm the fix.
Why this transforms team productivity
When debugging is isolated to individual surfaces, resolving an orchestration error takes multiple handoffs between functional analysts and technical developers. By correlating UI events, data models, orchestration logs, and live tenant security rules into a single continuous conversation, kiweely eliminates the guesswork.
The result is that functional consultants who understand the business logic can confidently identify, explain, and repair Extend application issues without getting lost in raw log archives.