Processing Workloads

This page explains how to run, monitor, and review workload processing in Simple Agent Designer.

Starting Processing

Processing a Single Workload

From the Workload Card

Open the workload you want to process
Choose Process from the action bar
The orchestration engine starts processing

From the Workload List

Select the workload in the list
Choose Process from the action bar

Processing All Workloads

Open the Workload List
Choose Process All from the action bar
All workloads with status Not Started are processed in sequence
The system commits after each workload, so a failure in one does not affect others

Processing from a Workload Type

For DataBased workload types:

Open the Workload Type Card
Choose Process from the action bar
A workload is created from the type configuration and immediately processed
The Table Splitter creates child workloads from matching records

What Happens During Processing

The Orchestration Flow

When you choose Process, the orchestration engine performs these steps:

License Check — validates your Simple Agent Designer license
Reprocess Check — if the workload was previously processed, asks for confirmation
Load Configuration — reads the Workload Type settings
Splitting Decision:
- DataBased: Calls Table Splitter, creates child workloads, then automatically processes them (see below), marks parent as Success, exits
- PayloadBased with Excel Splitter: Calls Excel Splitter, creates child workloads, then automatically processes them (see below), marks parent as Success, exits
- PayloadBased without splitter: Proceeds to agent processing

When a parent workload is split, the orchestration engine does not stop at creating the children. It automatically iterates over all child workloads with status Not Started and runs each one through the agent pipeline end to end. You do not need to process the children manually — splitting and processing happen in a single Process action. Individual children can still be reprocessed later if needed.

File Attachment — if the workload has Additional Data, attaches it to the AI conversation
Agent Initialization — creates the starting worker with full context
First Message — sends the "PleaseStart" system prompt to the agent
Agent Processing Loop:
- Agent calls the AI provider
- If tool calls are returned, executes tools and loops back
- If other agents are referenced, queues messages via orchestration
- Process continues until the agent completes or fails
Status Update — marks the workload as Success or Fail with result text

Tool Execution Loop

When an agent's AI response includes tool calls:

AI Response (with tool calls)
    ↓
Execute tool call 1 → Get result
Execute tool call 2 → Get result
    ↓
Send tool results back to AI
    ↓
AI Response (with more tool calls or final answer)
    ↓
Repeat until no more tool calls

The system continues this loop until the AI provides a response without tool calls.

Monitoring Processing

Debug Run (Real-Time)

For detailed, real-time monitoring:

Open the Workload Card
Choose Debug Run from the action bar
The Process Chat page opens
Processing begins, and you see each activity as it happens

The Process Chat shows:

Column	Description
Worker	Which agent performed the action
Date	Timestamp of the activity
Description	First 2024 characters of the AI text

Actions in Process Chat:

Next Step — manually trigger the next worker in the chain (useful for step-by-step debugging)
View AI Stack — shows all currently active workers and their state

Viewing History After Processing

View Last Run

Open the Workload Card
Choose View Last Run from the action bar
The ChatViewer page opens with the complete conversation history
Review the full AI text for each activity

View Last Run Data

Open the Workload Card
Choose View Last Run Data from the action bar
The Process History page opens with detailed activity records

Process Activity Records

Every AI interaction during processing is logged as a Process Activity record:

Field	Description
Line No.	Auto-incremented unique identifier
Work ID	References the workload being processed
Date	Timestamp of the activity
Worker	Agent ID that performed the action
Description	First 2024 characters of the AI response
AI Text	Full AI response stored as a blob (for large responses)
Tool Used	Names of any tools executed during this activity

These records provide a complete audit trail of what each agent did during processing.

Reviewing Results

Workload Status

After processing, check the workload status:

Status	Meaning
Success	The agent completed the task successfully
Fail	The agent encountered an error or could not complete the task
Not Started	The workload has not been processed yet
Started	Processing is currently in progress

Processed Result

The Processed Result field on the Workload Card contains the agent's summary text — typically the final response from the primary agent.

Documents Created

If the agent created Business Central documents:

Open the Workload Card
Choose Documents Created from the action bar
The Workload Links page shows all linked documents

Link Type	Description
General Journal	General journal entry created
Sales Quote	Sales quote created
Sales Order	Sales order created
Sales Invoice	Sales invoice created
Sales Credit Memo	Sales credit memo created
Purchase Quote	Purchase quote created
Purchase Order	Purchase order created
Purchase Invoice	Purchase invoice created
Purchase Credit Memo	Purchase credit memo created
Item Journal	Item journal entry created
Data	A referenced data record

Click on a document number to drill down to the actual Business Central document.

Background Processing

Simple Agent Designer supports background processing for workloads. When a workload is processed in the background:

A background session is created
The orchestration engine's CompleteProcess procedure runs in that session
When complete, an Agent Work entry is created to track the work performed
The workload record is refreshed with the final status

Background processing is useful for long-running workloads that should not block the user's session.

Reprocessing

To reprocess a workload that has already been processed:

Open the workload (or select it in the list)
Choose Process
The system detects it has been processed before and asks for confirmation
If confirmed, the workload is reset and processed again
Previous activity records are retained for reference

Error Handling

Common Processing Errors

Error	Cause	Resolution
License validation failed	Invalid or expired license	Activate or renew your license
AI provider error	Invalid endpoint, key, or model	Check your AI Provider configuration
Timeout	AI provider took too long	Retry the workload; consider a faster model
Tool execution error	Tool failed during execution	Check the process activity for details
Rate limit (429)	Too many AI requests	The system retries automatically (up to 10 times)

Diagnosing Issues

Open the failed workload
Use View Last Run to see the conversation history
Look for error messages in the activity log
Check the Tool Used field for any tool that may have failed
Review the AI response for unexpected behavior

Tips and Best Practices

Use Debug Run for new workload types until you are confident they work correctly
Always review the first few results when setting up a new workflow
For bulk processing, process a small sample first before running Process All on hundreds of workloads
Monitor AI token usage in the activity log to track costs
Set up incremental processing for recurring data-based workloads
Review the Documents Created links to verify agents created the correct documents
Keep the Process Chat open during development to see real-time agent behavior