How to Import a GPT Template and Connect It to a Chatbot in WOZTELL

Import a Template and Set Up a GPT Assistant in WOZTELL

Import a template and configure a GPT assistant step by step

In this article, you will learn how to import a template in WOZTELL’s builder and connect it with an OpenAI and Azure Open AI assistant. With this configuration, you’ll be able to automate conversational flows using artificial intelligence, without the need for coding.

Notes
Note:
Before starting, make sure your WOZTELL account is active and that you have already created a GPT assistant connected to an OpenAI or Azure Open AI account.

Option A: Install the Open AI extension

Before starting, we must install the GPT extension:
  1. Go to the WOZTELL platform and open the " Marketplace " panel.

  1. Find the extension " ChatGPT " and click on " Learn more ".

  1. Click on " Get Extension " to start the installation.
 
  1. Select the latest version, assign a recognizable alias, and click on " Install ".

  1. A configuration window will open. To complete it, you’ll need the accessToken of the chatbot you are going to use.

Get the accessToken of the Open AI chatbot

Option 1: From advanced chatbot settings

  1. Go to the "Builder" panel and select the chatbot you want to update.

3. Open the "Chatbot Manager" and go to advanced settings. Then access the "Config Map".


4. Look for the "accessToken" field. This is the key that connects the AI assistant to the chatbot builder.
5. Copy the entire string.


Option 2: From the chatbot in OpenAi


2. In the left sidebar, select "Api Keys" 

3. In the top right corner of the screen, click on "Create new secret key"

4. Name the key and click the "Create secret key" button.


5. Copy the code
6. Return to the extension’s configuration window and paste the code into the "API Key" field.


We recommend setting the "Number of suggestions" field to 5 and the "Number of previous chat history" field to 6 for optimal performance.



7. Click on "Save" to save the changes.
 

Option B: Install the Azure Open AI extension

1. Go to the WOZTELL platform and open the "Marketplace" panel.

2. Find the "Azure Open AI" extension and click "Get Extension" to start the installation.

 
3. Select the latest version, assign a recognizable alias, and click "Install".

4. On the configuration page, you will see the option "Suggest GPT" (Suggest reply in the inbox). If you’re not using this function, you can skip this configuration.

 

Import a chatbot template

1.  In the top panel "Builder", click on "Import Chatbot".


2. In the popup window, select the "Chatbot Templates" tab.
3. Choose the most recent template (the one that appears first) and click "Import".
4. Add a recognizable name and click "Import" again.


In the case of Azure, import "Azure Chat Completion Chatbot"


5. The system will automatically open the builder with the new imported flow.

  Configure the new flow

1. Select the general node and open its settings in the right-hand panel.
2. Go to the "Pre-actions" section and click "Advanced".



3. An advanced configuration popup will open where you can select Assistant API or Azure Chat Completion in the "Resource Template" dropdown.




Fill in all required fields: API Key, Azure Open AI endpoint, and deployment. (You can find all the details in your Azure AI Foundry)





4. Select the connection and the agent that will execute the responses.
 

5. Click on "Save" to store the configuration.

 

  Activate   the chatbot

1. Go to the "Channels" panel and click "Edit" on the corresponding channel.


2. In the sidebar menu, go to "Chatbot Settings".


3. Click on "New Chatbot" and select the chatbot you imported. Note: The alias name was set in step 4.


5. Activate the chatbots you will use, deactivate those you won’t, and click "Save".

 
Done! Your chatbot is now configured.

Connection verification    (optional)

To confirm that the connection between the assistant and the chatbot is working correctly:
  1. Go to the "Platform" section from the side panel to identify the linked WhatsApp number.
  1. Send a message from another number to that channel.
  2. Verify that the chatbot responds correctly.  

Done! Your chatbot is confirmed to be properly configured.

Best practices and recommendations for using OpenAI and Azure Open AI with WOZTELL

1. API security:

  • Always keep your API Key safe and never share it publicly. Consider using additional security configurations such as two-factor authentication to protect your OpenAI or Azure Open AI account.

2. Model optimization:



  • Make sure to regularly review and adjust your GPT assistant’s configuration to improve performance. This includes personalizing the model to match the type of conversations you want to automate.



3. Constant monitoring:



  • Monitoring chatbot performance after every deployment is key. Evaluate the number of interactions and user satisfaction to adjust responses and improve the conversational flow.



4. Moderate token usage:



  • Avoid using an excessive number of tokens per interaction, as this can impact cost and service efficiency. Set a reasonable token limit per session.



5. Monitor AI behavior:



  • Review the responses generated by the chatbot to ensure the information provided is accurate and relevant for your users. While AI is powerful, it’s always advisable to keep supervision.


6. Regular model updates:

  • OpenAI and Azure Open AI technologies are constantly evolving. Make sure you're using the most updated version of the extension and adjust your settings when necessary.

7. Use of smart templates:

  • Use chatbot templates that have been tested and optimized for a more efficient workflow. Import the most relevant templates according to your business’ common interactions to reduce the need for extensive customizations.

8. Clear and concise responses:

  • Configure chatbot responses to be clear and easy to understand. The AI should provide straightforward answers that solve users' queries without unnecessary complications.

9. Fallback flow implementation:

  • Whenever possible, implement fallback flows that redirect the user to a human agent if the AI cannot handle the request correctly.

10. Constant feedback:

  • Ask users for feedback to understand how well chatbot interactions are working. This feedback is invaluable for making fine adjustments and improving the overall experience.


Frequently asked questions   

Where do I add the assistant payload, in the same node or in a new one?
You can add it in the same node or in a new one. Both options are valid as long as the logic of the flow is respected.

What inputs can the chatbot understand?
Text, images, and audio. Audio messages are supported through the transcription option in the pre-action of the first node.

What does the global node trigger?
The global node responds to any text or image message. If there are other triggers with higher priority, they will execute first.

Can I use a different flow for different triggers? 
Yes. You can work with a separate flow or integrate everything into a single tree, depending on your needs.

What happens if users send me an audio message instead of text?

If users send you a voice message, you can activate the automatic transcription option. This feature converts the audio into text so it can be sent to the AI assistant and processed as if it were a written message.

To use this feature, you need to create an additional transcription resource in your Azure AI Foundry. This involves setting up a new:

  • set of API keys

  • Azure OpenAI endpoint

  • and a specific deployment for transcription

Once configured, you’ll be able to offer a more complete and accessible experience to your users, even when they prefer to communicate by voice.