4. Workflows and Execution

4.1 Workflows List

The Workflows area shows all workflows you have created:

• Each entry includes a name, status, and last run time.
• You can open a workflow to view details or run it.

4.2 Workflow Details

In a workflow detail view you can typically:

• Inspect the steps and logic of the workflow.
• Adjust input parameters where supported.
• Start a new run of the workflow.

4.3 Monitoring a Run

When you start a run you can open the live execution or monitor view:

• Watch step-by-step progress.
• See which step is currently running.
• Stop or cancel a run if necessary.

4.4 Viewing Results

After a run finishes you can open the execution result:

• See whether the run succeeded or failed.
• Inspect logs and outputs.
• Jump back to the workflow for adjustments.

4.5 AI Changes and Script Generation Time

When you ask Ami to modify a workflow or generate one from a recording, the app may need time to:

• Call AI models to reshape or optimize the workflow definition.
• Generate helper scripts, such as scraper agents that interact with web pages.

You might see loading indicators or a "preparing" state before the workflow is ready to run. The time required depends on:

• The length and complexity of your recording or existing workflow.
• How many times you have asked the AI to refine the workflow.

Similarly, when starting a run, Ami may spend a short period preparing or updating scripts before visible steps begin to execute. This is expected; you can watch progress from the monitor or live execution views.

4.6 Using AI to Modify an Existing Workflow

In addition to generating workflows from recordings or descriptions, you can ask AI to help inspect and fix an existing workflow.

The typical pattern is:

1. Open the workflow detail view.
2. Open the AI-assisted editing or conversation panel for this workflow (if available in your version).
3. Start a short conversation where you:
   • Ask what the current workflow does: for example, "Explain the steps in this workflow".
   • Describe the expected result vs. the actual behavior: for example, "Step 2 only saves price, but I also need stock".
   • Point out where it ran incorrectly: for example, "On some pages this step fails with a missing element error".
4. Let the system propose changes. AI may:
   • Adjust one or more steps (for example, update a scraper step to capture extra fields).
   • Update data requirements, conditions, or error handling.
   • Suggest that you re-run or single-step test the updated workflow.

You can iterate in this conversational loop:

• Run or test the workflow.
• If something is still wrong, tell AI what happened and what you expected instead.
• Accept or reject the suggested edits based on your needs.