Skip to content

Getting Started

This guide walks you through your first interaction with PADAI, from launching the panel to generating your first piece of PAD code.

Prerequisites

Before you begin, make sure:

  • [x] The PADAI backend server is running (see Installation)
  • [x] PadaiPanel is installed on your workstation
  • [x] Power Automate Desktop is installed
  • [x] You have a valid API token from your IT administrator

Step 1: Open Power Automate Desktop

Open PAD and enter the Designer by opening an existing flow or creating a new one. You need to have the Designer window visible on screen.

Step 2: Launch PadaiPanel

Run PadaiPanel.exe. The panel will:

  1. Automatically detect the PAD Designer window
  2. Dock itself to the right side of the Designer
  3. Read the initial context (subflows, actions, errors)

Tip

If PadaiPanel is launched before PAD, it will wait for a Designer window to appear.

Step 3: Verify the Connection

Check the status indicator at the bottom of the panel:

  • Connected -- the backend server is responding correctly
  • Disconnected -- check the server URL in PadaiPanel settings

Step 4: Read the Context

Before asking the AI anything, make sure it knows about your current flow:

  1. Click the Refresh Context button in the toolbar
  2. The panel reads your subflows, visible actions, and any errors
  3. The context indicator shows what was detected

This step is automatic at first launch but should be repeated after you make changes in PAD.

Step 5: Generate Your First Code

Type a natural language prompt in the input field at the bottom and press Enter or click Send.

Example prompts to try:

What you want Prompt
Simple output "Display a message box saying Hello World"
File operation "Copy all .xlsx files from C:\Import to C:\Archive"
Loop "Loop through each row in the Excel file and print the value of column A"
Web automation "Open Chrome, navigate to google.com, and search for Power Automate"
Variable "Create a text variable containing today's date in YYYY-MM-DD format"

What happens behind the scenes:

  1. Your prompt is sent to the PADAI backend
  2. The AI generates a structured plan
  3. The plan is compiled through the strict DLL-based pipeline
  4. The compiled code is validated and linted
  5. The final Robin code is returned to the panel

Step 6: Apply the Code

When the AI returns generated code, you will see it in a chat bubble with action buttons:

  • Apply -- inserts the code directly into the PAD Designer via UI Automation
  • Copy -- copies the Robin code to your clipboard

Click Apply to inject the code into your active subflow.

Warning

Applying code replaces the content of the active subflow in the Designer. Make sure you are on the correct subflow before clicking Apply.

Step 7: Verify in PAD

After applying the code:

  1. Check the Designer -- the actions should appear as action cards
  2. Look for any red error indicators
  3. If errors appear, use the Fix Errors button in the toolbar for automatic correction
  4. Save your flow when satisfied

What to Try Next

Now that you have generated your first code, explore these features:

Quick Fix Buttons

The toolbar has two quick fix buttons:

  • Fix Errors -- reads errors from PAD and generates targeted fixes
  • Fix Warnings -- addresses warnings like placeholders and invalid paths

Code Patching

Instead of regenerating an entire flow, you can patch existing code:

  1. Refresh the context so the AI sees your current code
  2. Describe the change: "Add error handling around the file copy step"
  3. The AI generates a targeted patch
  4. Review and apply

Flowchart Visualization

Ask the AI to visualize your flow:

  • "Show me a flowchart of this flow"
  • "Generate a diagram of this subflow"

A dedicated diagram window opens with zoom and SVG export.

Workbench Tools

Toggle the Workbench panel for access to 27+ specialized tools:

  • Flow Health scoring
  • Variable Explorer
  • Run log replay and analysis
  • Performance metrics
  • Knowledge Base lookup

Voice Dictation

When the input field is empty, the Send button becomes a microphone. Click to start voice recording -- your speech is transcribed via OpenAI Whisper and inserted into the input field.

Tips for Better Results

  1. Be specific -- include file paths, variable names, and expected formats
  2. Context first -- always refresh the context before asking about your existing code
  3. One step at a time -- for complex flows, build incrementally rather than in one massive prompt
  4. Mention variables -- if your flow uses specific variables, include their names in the prompt
  5. Use the workbench -- the specialized tools are often more precise than free-form chat for specific tasks

Need Help?

  • Browse the full User Guide for detailed feature documentation
  • Check Troubleshooting for common issues
  • Contact support@padai.tech for assistance