Set up an LLM chatbot

Before you start

Here's what you need to know before you start:

Step one: Set up your chatbot

  1. Go to AI Hub > AI actions > LLM chatbots.

  2. Click Create new LLM chatbot.

  3. Choose from one of the available bot templates. We'll start you off with some default prompts to speed up the creation process. Or, you can choose to create a custom personality.

  4. Click Get started.

Step two: Select your LLM Connector and specify the LLM Model

In the creation process, you will be prompted to select your LLM connector.

Choose the OpenAI Connector that you installed earlier. Additionally, specify the LLM model you want to use. This selection will determine the language capabilities and behavior of your LLM Chatbot.

Step three: Build your LLM chatbot prompt

Building your LLM Chatbot prompt is a crucial step in defining its behavior and responses. You will be guided through several sections to complete:

  1. Purpose of the Bot: Specify the main goal or purpose of your chatbot. Understanding the intent helps in crafting relevant responses.

  2. Key Features and Tasks: Identify the specific tasks or functions that your chatbot should perform. This will guide the development and design process.

  3. Target Audience: Define the target audience for your chatbot. Knowing the audience helps in tailoring the language, tone, and content to suit their needs.

  4. Tone of Voice: Specify the tone of voice for your chatbot. Is it friendly, professional, or humorous? Consider the level of formality and technicality that aligns with your brand and audience.

  5. Limitations/Boundaries: Set the limitations or boundaries for your chatbot. Specify what it cannot do or should not ask in conversation. This includes scenarios where it might need to hand over to a human agent or lacks the necessary knowledge to resolve an issue.

  6. Exit Strategy and Handover Protocol: Define the conditions under which your chatbot should gracefully exit the conversation. For instance, if it encounters a query beyond its capabilities, it could explain its limitation and offer to escalate the conversation to a human agent or provide relevant resources for further assistance.

  7. Custom Instructions: Provide any further instructions for your chatbot. You can be as detailed as you want, ensuring that your chatbot understands your specific requirements.

Step four: Define knowledge base access

Specify which knowledge base folders your chatbot can access. This will determine the information it can retrieve to provide accurate responses.

Step five: Save and start testing

Once you have completed building your LLM Chatbot prompt and defined knowledge base access, save your settings. You can now start testing your chatbot's capabilities. Refine the prompt and knowledge base content as needed to ensure optimal performance.

Step six: Enable in Flows

To fully leverage the capabilities of your LLM Chatbot, you need to enable it in Flows. Flows is a powerful tool that allows you to create conversational flows and automate interactions with users across various channels.

Here's how you can enable your LLM Chatbot in Flows:

  1. Go to the Flows section of the Bird Engagement platform.

  2. Set up a conversational or message-based trigger in Flows on any channel of your choice. This trigger will initiate the conversation with your users.

  3. Add the 'LLM Chatbot' action as the next step in your flow. This action will invoke your LLM Chatbot and allow it to provide responses based on the user's input.

By incorporating the LLM Chatbot action into your flows, you can seamlessly integrate it into your conversational experiences. Users will be able to interact with your chatbot and receive accurate and personalized responses.

Step seven: Review the logs

To enhance your chatbot's performance, regularly review its logs – a detailed record of its decision-making process. In these logs, you'll find information about the bot's analysis (or it’s thoughts and observations), and the documents it used to respond. By understanding these details, you can refine your chatbot's prompts and knowledge base for better outcomes.

Congratulations! You have successfully set up your LLM Chatbot and are ready to provide accurate and personalized responses to your users. Remember to continuously refine and optimize your chatbot's prompt and knowledge base content to enhance its performance.

Last updated