MyCatalyst: Settings

The MyCatalyst Settings area allows administrators to manage the AI Chat tool settings.

Site Administrators can access the MyCatalyst settings area via the ‘Site Administration' side menu.

General

See the settings available under the General tab below.

Open AI API Key

The Open AI API Key section allows administrators to add the Open AI API key required to use the chat service.

To obtain your Open AI key, you will need to follow the steps below:

1. Create an OpenAI account:

• Head to the OpenAI website found here.
• Sign up for an account using email.

1. Verify Email Address
2. Login and Access the API Dashboard.
3. Create a New API Key

• Via the dashboard, click on “Personal” or the account username in the top right corner and select ‘API Keys’ from the dropdown menu.
• Click on ‘Create new secret key’ to generate a new API key.

Language Model Configuration

The Temperature slider controls the creativity of the chat output. Either more precise or more creative.

A lower temperature setting would be used for tasks that require factual or highly reliable information, such as answering a technical question or generating formal text.

A higher temperature can be used when creative writing or brainstorming is acceptable when searching.

A higher temperature would be recommended if you want the AI to present more variety in the search output.

 

Language Model

OpenAI uses various language models, each having varying costs.

See the Open AI language model pricing for details here. 

Administrators can use the drop-down menu to change which model the MyCatalyst tool is using. Customers incur the costs.

 

See a brief overview of each language model below:

GPT-4o: Strikes a balance between performance, cost, and usability, making it a versatile choice for businesses and developers looking to implement AI solutions.

GPT-4o mini: GPT-4o Mini combines efficiency, speed, and versatility, making it an excellent choice for lightweight applications requiring AI capabilities.

GPT-4: Multimodal (text + image) inputs, enhanced accuracy, better reasoning, and multilingual support.

GPT-3.5: GPT-3.5 Turbo balances speed, quality, and cost for various AI-driven tasks.

Prompt Engineering

See the settings available under the prompt engineering tab below.

Extend & customise the default prompt

MyCatalyst handles default prompt engineering in the background.

However, you may wish to extend and customise the default prompt, for example, to return responses in a particular style.

"Use custom expressions to enhance the prompt, such as:

"Don't mention the name, title or source of the document in response to the user."

"Only use information contained in the following pieces of retrieved context to answer the question - do not use your knowledge or any other source such as the internet."

"If you need to refer to the retrieved context in your response, refer to it as "our research".

Relevance

This slider sets the relevance of the content used in the returned responses.

 

Rerank Algorithm

The rerank algorithm dropdown sets the age of content that should be used in the returned responses. See the dropdown options below.

None

The most relevant content is used regardless of content age.

Half-life

More recent content is preferred. Only documents with a relevance score above min relevance are considered. The relevance score is then reduced over time, with the half-life being the number of days taken for the score to be reduced by half.

Content Similarity 

Older content that is superseded by more recent, relevant content is not used. Only documents with a relevance score above min relevance are considered. The older document is not used if the similarity between any two documents is greater than the maximum similarity.

Disclaimer management 

See the settings available under the Disclaimer Management tab below.

Response Inline disclaimer 

Administrators can set a required inline disclaimer after each response is generated.

A small note that reminds users to verify the information can also be included here. 

When a user copies any part of the AI's response, the same disclaimer note will be automatically appended to the copied content. 

Disclaimer reference link

Set an optional 'Full terms' link that will be applied across multiple components, including inline disclaimers, initial disclaimer dialogs, reminder disclaimer alerts, and any AI response content copied to the clipboard. 

A link can also be set to the original reference page using the toggle below. 

Initial disclaimer dialog

Users can be prompted to accept the disclaimer when first using the tool if the initial disclaimer is set using the toggle below. 

Once the 'Enable initial disclaimer' toggle is switched on, A Dialog modal title and Dialog modal content will be available. See below.

Once the disclaimer settings have been set, the disclaimer will appear for users, as shown below. 

Re-Prompt Users

Once the initial disclaimer content has been saved and changes have been made, a further Re-Prompt Users option will be available.

This allows administrators to Notify all users about changes to the Terms and & Conditions. The last prompted date will also be visible under this option.

Disclaimer acceptance logs

Administrators can export a CSV format list showing the users that have accepted the initial disclaimer and when. 

Exported Data

The exported CSV displays the following data: 

username: The user's username.

disclaimer: The text of the disclaimer which has been accepted by the user. 

URL: The URL of the disclaimer reference page if this has been enabled. See the Disclaimer Reference Link above. 

accepted: The date that the user accepted this disclaimer text. 

See an example data export below

Remember to save any changes made in the settings via the ‘Save’ Icon at the top of the page.

 

Administrators who switch to another settings tab will be prompted to Save or Leave & Discard the changes. 

If you have any further questions about the 'MyCatalyst' settings page, please contact the support team.