By configuring a ChatBot, you have the option of having artificial intelligence support your agents in creating texts for ticket creation and email processing.

As of 2.2025 you can use ChatGPT/OpenAI  and DeepSeek.

ChatGPT

To access ChatGPT you need an account with OpenAI and a valid API key, see here

Open the "System Parameters" list and select "New Parameter" to save a new system parameter.

Enter “ChatGPT” as the name. Please pay attention to upper and lower case letters.

Enter the required values ​​to access the ChatGPT API. Use the following text as a template and adapt it to your requirements.

Please insert your own API key into the apiKey line:

In addition to the user request, ChatGPT always receives a system command with instructions on how the ChatBot should respond. 

todo4teams can use different commands depending on the context:

1. when creating new tickets: creation_system_command
2. when replying to emails:  reply_system_command
3. in all other contexts:  neutral_system_command

The system commands should always contain the instruction "Format the response as an HTML fragment." included so that the AI's answer can be adopted in the correct format.

In the reply_system_command above, it is also required that the AI ​​uses the appropriate tone towards a customer.

You can also have a short command precede and follow the user request. To do this, use the parameters reply_prefix, creation_prefix, etc., depending on the context.

Example ChatGPT configuration:

Then save the settings. The ChatBot function can then be used immediately.

DeepSeek

Open the "System Parameters" list and select "New Parameter" to save a new system parameter.

Enter “DeepSeek” as the name. Please pay attention to upper and lower case letters.

The configuration of DeepSeek is very similar to that for OpenAI (see above). 

Use the following template for configuration:

Here, too, you need to register as an API user with DeepSeek and an API key.

In addition to the user request, DeepSeek always receives a system command with instructions on how the ChatBot should respond. 

todo4teams can use different commands depending on the context:

1. when creating new tickets: creation_system_command
2. when replying to emails:  reply_system_command
3. in all other contexts:  neutral_system_command

The system commands should always contain the instruction "Format the response as an HTML fragment." included so that the AI's answer can be adopted in the correct format.

In the reply_system_command above, it is also required that the AI ​​uses the appropriate tone towards a customer.

You can also have a short command precede and follow the user request. To do this, use the parameters reply_prefix, creation_prefix, etc., depending on the context.

Empolis Buddy

The Empolis chatbot "Buddy" uses RAG ("Retrieval-Augmented Generation") to generate responses, meaning it is designed to be enriched with company-specific knowledge.

  Example of an Empolis Buddy configuration:

Create an API user in Empolis and configure the username, password, etc. settings to match your Empolis instance.

Also, configure the docBaseUrl setting to display the Empolis documents used for the response.

Basic Facts

To provide the AI ​​bot with some facts about your company, your products, etc., save a system parameter called ChatBaseFacts:

Enter facts in plain text that the AI ​​agent should use in its responses. The amount of text is unlimited – but keep in mind that the chatbot's processing time will increase significantly if you pass more than a few screens here.

Example:

Company: Margarete Steiff GmbH / Steiff Beteiligungsgesellschaft mbH
Founded: 1880 (Official Founding Date)
Headquarters: Giengen an der Brenz, Baden-Württemberg, Germany
Industry: Toy manufacturer (high-quality teddy bears and plush toys, including children's fashion)
Trademark: Button in Ear
Number of employees: Approx. 1,300 - 1,700 (depending on source and reference year)
Revenue: €96.7 million (Steiff Beteiligungsgesellschaft mbH, 2022) / €112 million (Steiff Group, other source)
Owner: Traditionally family-owned
Steiff is the global market leader in the high-quality teddy bears and plush toy segment.

Analysis of Images and Attached Documents

Optionally—and depending on the selected AI model—you can include received image and document attachments in the analysis:

Analysis of Image Files

By setting image_analysis=true, you enable the transmission of attached image files to the AI. This allows for the automatic answering of customer questions that relate to image content.

Analysis of Attached Documents

To enable the analysis of text within attached PDF, DOCX, or ODT files, add the following lines to your ChatBot configuration:

read_attachments=true
max_chars_per_attachment=2000

The max_chars_per_attachment setting determines the maximum number of text characters included in the analysis per attachment. 2,000 characters correspond to approximately one full A4 page of text.
Please note that setting a high value here will increase the number of tokens required by your AI, which may subsequently impact your costs!

One-Click Prompt

In the chatbot dialog, you can request an answer without typing a question yourself:

Simply click on .

The prompt that is then sent to the chat agent can be configured in the system setting AUTO_ANSWER_PROMPT, e.g., "Answer the customer's request."

Customize this prompt as needed.