AI Hub

Flexito OmniAI's AI Agents revolutionise chatbot development by providing seamless integration with leading AI platforms. Whether you're looking to enhance customer support, streamline lead generation, or automate workflows, these AI Agents offer dynamic, human-like interactions with minimal setup. Discover how this powerful feature can transform your business operations.

Flexito OmniAI has launched its AI Agents feature, simplifying the creation of AI-powered chatbots by eliminating the need for manual handling of chat completions and AI assistants.

With AI Agents, users can now effortlessly build sophisticated conversational AI bots that integrate with OpenAI and other popular GenAI platforms. These bots can perform tasks independently and engage in dynamic, human-like interactions.

Whether your goal is customer support, lead generation, or workflow automation, Flexito OmniAI’s AI Agents make AI chatbot development accessible and straightforward for businesses and developers.

You can find the template for the appointment booking AI agent created during the workshop at this link: .

Accessing AI Agents and Functions

To access AI agents and functions, navigate to your bot and click on “AI Hub” in the left toolbar.

Creating an AI Agent To create a new AI Agent, click on “+ AI Agent”.

Naming & Describing the AI Agent:

In this section, provide a name and a brief description for the AI agent. The description should give enough context about the AI Agent’s intended function.

Settings

Here, you choose the model and platform for your AI agent, and adjust various parameters that influence the agent's behaviour.

Note: For AI Agents that require function usage, it's advisable to select higher models like gpt-4-turbo-preview, as they offer greater stability and accuracy.

Flexito OmniAI supports creating agents using OpenAI, Deepseek, and Grok AI. More models, including Google Gemini and Claude, will be added soon.

You can also adjust parameters such as temperature and repetition count to further refine the agent’s behaviour.

The "Number of chat messages before auto summarize" feature helps manage lengthy conversations by summarising chat history after a set number of messages. Once the limit is reached (e.g., 10, 50, or 100 messages), the system creates a concise summary and reinserts it into the chat. This process retains key details while reducing character space usage, enhancing memory management. Summarising past exchanges helps the AI maintain important context without overwhelming the chat history, ensuring smoother interactions. Users can customise the maximum token limit for summaries, with 500 tokens suitable for general chats and 1,000 tokens recommended for complex tasks like appointment booking. This feature boosts AI performance, conversation clarity, and long-term engagement efficiency.

Finally, choose your preferred output format, either text or JSON:

AI Agent Advanced Mode

When Advanced Mode is activated, the AI agent does not directly reply to the user. Instead, it stores the response in the system field “Last AI Agent Reply”. You must select a workflow to process and handle the response before it is sent to the user.

Key Features

• Response Formatting: Break down long AI replies into multiple sections for better readability.

• Media Integration: Enhance responses by adding relevant media files (images, videos, or attachments).

• Workflow Automation: Use custom workflows to process AI-generated content, improving message delivery and presentation.

Note: Enabling Advanced Mode will disable the Auto Suggestions feature.

Agent Prompt

In this section, define the persona (how you want the AI agent to behave) and its role (or any impersonation) you want it to follow.

In the skills section, specify all the features you want your AI Agent to perform, such as collecting user information or scheduling appointments. This section also provides insights into which functions you need to set for your objectives and goals.

The Product & Service Information feature allows you to input detailed descriptions of your products and services, including specifications, booking options, pricing, and helpful references.

This ensures users can easily access relevant information without needing to ask repeatedly. Providing structured data streamlines interactions, improves customer engagement, and enhances the overall user experience.

In the constraints section, specify the behaviours you do not want your AI to perform, such as certain questions you want to avoid or words you don't want the agent to use.

For convenience, Flexito OmniAI offers a “Generate Agent Prompt” button, which creates the role, persona, skills, and constraints sections based on your description.

This feature is particularly useful in two scenarios. First, if you're inexperienced in prompt writing, it helps you quickly structure an effective prompt without needing advanced skills. Second, if you prefer not to start from scratch, it provides a predefined framework that you can modify to suit your specific needs.

Creating AI Functions Click on “+ AI Function” to create a new AI Function.

In the first section, define the name and description of the function.

Sample Description: This function needs to capture user details, including first name, last name, and email. The agent must validate the email format if the user does not provide it correctly.

In the next section, define the complete prompt for the function, detailing what you want it to do or perform.

Sample Prompt:

### Skill: capture_user_details - This skill captures essential user information: first name, last name, and email. It ensures the email provided is in a valid format.

#### Steps for Execution: 1. Prompt the user to enter their first name. 2. Request the user's last name. 3. Ask for the user's email address. 4. Validate the email format using a regex pattern (e.g., `^[\w\.-]+@[\w\.-]+\.\w{2,4}$`). 5. If the email is valid, store all details; if not, prompt the user to re-enter a valid email.

#### Constraints: - All inputs (first name, last name, and email) must not be empty. - The email must adhere to standard formatting rules.

#### Formatting Rules: - Ensure input is free of leading and trailing whitespaces.

#### Error Handling: - If validation fails, inform the user of the specific error (e.g., "Invalid email format"). - Allow a predefined number of attempts to re-enter the email.

#### Conditions: - Proceed to capture details only if the email is valid. - Request each user detail separately. - Continue with the function call only when all parameters are captured.

In the next section, define the values you want to extract from the function (like first name, last name, etc.) and describe them. Also, choose which CUF you want them saved in.

Note: Ensure you check the “Required” box to make the value mandatory for the function to collect. You can also enable the “Memory” feature, which reviews the conversation history to see if the value already exists. If it does, the system will skip asking for it again and proceed to the next parameter.

Finally, attach the flow (only workflows allowed) that should be triggered when the function is called.

This feature allows you to send captured values to another platform through native integrations (like Google Sheets) or perform API calls via an external requests node. It can also retrieve information from an external source and pass it back to the AI agent, ensuring the conversation continues smoothly.

This capability simplifies task automation, real-time information updates, and enhances AI responses with the latest data.

Note: You can now use “Send Message” nodes within workflows. This allows AI Agents to send media and other dynamic content based on received information.

Selecting the AI Function

For AI Agents to utilise AI functions, you must select them within the AI Agents modifications.

Once selected, it will appear with an overview of the function prompt.

Note: When you deselect/select a function, its prompt will appear or disappear from the overlay accordingly.

Using AI Functions

AI Functions can only be used in workflows. Select the AI Function Output node from the AI Agents tab in the action block. This data will be fed back into your AI Agent after a function is called and a workflow is processed.

Using AI Agent Create an action node and select “AI Action:”

Click on Edit Action to choose the AI Agent.

Select the primary AI Agent (the one that will trigger and fulfil first) and any secondary agents if needed. You can also set an inactivity timeout for the user. If the user stops responding during a conversation with an AI agent, this timeout will trigger, allowing you to follow up and re-engage the user.

When secondary agents are selected, the primary agent inherits any functions from those secondary agents. However, the primary AI agent’s persona, role settings, and LLM settings will be inherited by the secondary agent.

This feature is particularly beneficial when you want to enhance the primary agent with additional capabilities without changing its core functionality. By integrating secondary agents, you can expand the range of tasks the primary agent can handle while maintaining a consistent interaction flow (persona, role setting, and LLM settings).

Finally, save the information in the CUF of your choice. This step is optional and is mainly for debugging purposes.

If you choose not to save the output to any user custom field, the bot response will still be sent automatically.

To use this AI Agent, simply send the flow to the user, and the AI Agent will begin interacting with them.

You can also link multiple AI Agents (Additional AI agents) to direct the user as needed. The additional agent will receive the title and short description in the system prompt. Once the intent is identified with the additional agent, you can connect this intent with another AI agent action. This ensures your bot covers the maximum user cases while optimising prompt token usage.

The entire conversation with the user will be saved in the new System JSON field called AI Messages:

AI Agent Incomplete Timeout

The Input Incomplete Timeout feature allows the AI agent to wait a specified number of seconds to capture all user responses and process them as a single response. This ensures the AI agent processes complete user input before generating a reply.

Note: Every new input restarts the timer. For instance, if the timeout is set to 10 seconds and the user types “hi,” the countdown begins. If they send another message, like “how are you,” after 6 seconds, the timer resets to 10 seconds instead of continuing from 7.

Creating AI Tasks

In the AI Hub, click on AI Tasks and then press “+AI Task”.

In the first section, define the name and the prompt for what you want the AI task to accomplish. AI Tasks are essentially small bundles of chat completions designed to perform single tasks.

You can use one of the available presets to get an idea of how to fill in the prompt.

In the settings section, you can define the model settings and other parameters such as temperature and max tokens.

In the output field section, if your AI Task requires an output (e.g., extracting specific information from a larger text), you can set an output field where the result will be saved into a CUF.

An example output will look similar to this:

Using AI Tasks In the action block, click on AI Actions and select AI Tasks from the dropdown.

Select the AI Task you wish to perform, and in the input field, enter the content on which you want to perform the AI Task.

The end result will look like this:

This value can then be mapped to a user-field for use in other flows.

Troubleshooting AI

There are two primary ways to troubleshoot AI responses. The first is to analyse the “AI Messages” System JSON field.

Go to the bot user overview and click on the AI Message JSON to analyse what the user is asking at each step and what responses the AI generates.

You can also observe when a function is called:

The second method is to directly analyse the entire conversation from Livechat (with system messages enabled).

This allows you to see which AI Agent is being utilised in the flow.

You can also see when a function is called.

Hovering over this reveals the arguments and outputs processed by the function.

Similarly, you can see when user fields are being filled.