AI Integrations
Overview
You can now integrate AI Large Language Models (LLMs) powered by OpenAI and Google Gemini into Gorilla as part of your study!
More than just a chatbot, AI components work seamlessly with the rest of the Gorilla tools, so you can customise prompts from trial to trial using the spreadsheet, automatically score participants' responses by passing data from the Store to an AI agent, and much more.
Check out our AI showcase for video examples of what you can achieve using AI features, or browse the topics in the menu to find out:
- how to get access to the AI components
- how to add an API key to your account
- where to find information about the AI components
- how to add a system prompt governing all AI interactions in your task
- what data from the AI components look like
Watch our recent AI in Gorilla: More Interactive Studies, Richer Data webinar to see the AI features in action!
Getting Access
AI components are available by request as a premium add-on to any Gorilla subscription. To get access, email subscriptions@gorilla.sc.
If you don't yet have a subscription but want to find out more about the AI tools, email info@gorilla.sc.
Adding an API Key
To use the AI components in Gorilla, you will need an API key for a supported provider (currently either OpenAI or Gemini).
To integrate the API key in your Gorilla account and keep it secure:
- Go to My Account and click API Keys.
Click Add API Key.
Enter a name for your API Key into the API Key Name field, and enter your API key itself into the API Key Secret field.
- Press Next. Your API key is now saved to your account.
To use your API Key, enter the API Key Name inside angled brackets, e.g. <OpenAI>, into the API Key setting of each AI component in your tasks.
API Credits
Using AI components in screen and Task previews or live experiments consumes credits from your OpenAI or Gemini subscription.
Some AI plans have a limit on the number of credits you can use. If you use all of your credits the AI model will stop producing any output and your Task will not work as hoped. If this happens, you will see an error message.
To fix this, please add more credits to your AI subscription directly with OpenAI or Gemini.
Please make sure you have enough credits on your subscription before launching your experiment on Gorilla!
AI Components
There are four AI components:
- AI (ChatGPT) - Access a ChatGPT model from within your task. Use in combination with the AI Chat Box or Trigger - AI components below.
- AI (Gemini) - Access a Gemini model from within your task. Use in combination with the AI Chat Box or Trigger - AI components below.
- AI Chat Box - Integrate an AI-powered chatbot into your task for participants to interact with.
- Trigger - AI - Send instructions to an AI model. For example, the AI model can process a participant's response and return another response which can be saved to your data or displayed to the participant.
System Prompt
To give core instructions or background information that should be used by every AI component in your task, you can add a System Prompt.
In your task, click Settings on the left-hand sidebar and select AI:
The prompt you enter will be used by every AI in your task by default.
To disable the system prompt for a specific AI component, open Additional Settings on the component and toggle on the 'Ignore System Prompt' setting:
AI Component Data
The AI components record the entire timestamped interaction history with the AI, whether you are presenting participants with a chatbot or using Trigger- AI in the background to score responses.
Below, you can download two example data files that show you what data from the AI components look like.
Download an example data file from a task using a chatbot
Download an example data file from a task using AI to score responses