Skip to content

Using the Agent

The AI Agent is your coding assistant integrated directly into ScriptO Studio.

Overview

AI Agent Panel

The AI Agent panel appears on the right side of the editor. It provides: - Natural language code generation - Error explanation and fixes - Code optimization suggestions - MicroPython and device-specific help

Opening the Agent

Click AI Agent in the toolbar to toggle the Agent panel.

How It Works

  1. Describe what you want in the chat input
  2. Agent generates a response with code and explanation
  3. Configure or use the code via the action buttons

Example Interaction

You: "Emulate a Tesla model 3 brake pedal on the CAN bus"

Agent: Here is the complete ScriptO code to emulate a Tesla Model 3 brake pedal on the CAN bus.

The Agent provides: - Complete working code - Explanation of what the code does - Configuration options

Action Buttons

Each code response includes buttons:

Button Action
Configure Open configuration dialog for the code
Open in new tab Create a new editor tab with the code

Code Generation

The Agent can generate code for:

Hardware Control

"Blink the LED on GPIO2 every 500ms"
"Read temperature from the BME280 sensor on I2C"
"Send a CAN message with ID 0x123"

Protocol Implementation

"Create a MQTT client that publishes sensor data"
"Implement a simple web server"
"Parse UDS diagnostic responses"

Device Configuration

"Set up CAN bus at 500kbps"
"Configure WiFi credentials"
"Initialize the SPI display"

Error Assistance

When you encounter errors, paste them into the Agent:

You: 

I'm getting this error:
TypeError: 'bytearray' object 'dtyarray' isn't a tuple or list

Agent: The error indicates that the CAN.send() method expects a standard Python 'list' or 'tuple' of integers for the data payload, rather than a 'bytearray'.

The Agent explains: - What the error means - Why it occurred - How to fix it

Context Awareness

The Agent can see: - Current editor content - Reference your code directly - Terminal output - See recent execution results - Device status - Know if you're connected

Use this in your prompts: 

"Look at my current code and optimize it for memory"
"Why did the last run fail?"
"Modify this to use asyncio instead"

ScriptO Configuration

When the Agent generates a ScriptO (configurable script), clicking Configure opens a configuration dialog:

ScriptO Configuration

The dialog shows: - Script name and version - Description of what it does - Configuration options (dropdowns, inputs) - Generate Code button

Example configuration options: - CAN Bus: CAN 0 (default) - CAN Mode: LOOPBACK - Internal test (no hardware) - Bitrate (bps): 500000 - 500 kbps (CANopen standard)

Click Generate Code to create the configured script.

Iterative Development

The Agent remembers your conversation, so you can iterate:

  1. "Create a temperature logger"
  2. "Add MQTT publishing"
  3. "Make it retry on connection failure"
  4. "Add a watchdog timer"

Each request builds on the previous context.

Tips for Better Results

Be Specific

❌ "Make a sensor script" ✅ "Read the BMP280 sensor on I2C address 0x76, log temperature and pressure every 10 seconds"

Include Constraints

❌ "Create a web server" ✅ "Create a minimal web server that serves a single status page and uses under 20KB RAM"

Reference Your Code

❌ "This doesn't work" ✅ "The CAN message in my current code isn't being received - can you check the bus configuration?"

Limitations

  • Code must be copied/run manually (Agent cannot execute directly)
  • Large codebases may exceed token limits
  • Requires internet connection to LLM provider
  • May need adjustment for specific hardware variants