1. In a nutshell
AI Agent lets you automate tasks with simple instructions using Agentic AI technology. You can create an agent by describing how the AI should behave and adding the required Deluge functions that have the scripts to perform the actions. It will then plan and execute the tasks in sequence, analyze situations, and make decisions on your behalf as you defined them to be.
2. Availability
AI agents:
- Can be created and managed in all plans of Creator, where the API invoke limit varies according to the subscribed plan
- Can only be created and managed by the super admin and admins, while all other types of users can invoke them
3. Overview
AI Agent is a specialized AI capability that can perform multiple complex tasks by making decisions on analyzing a situation and taking actions independently without human intervention. Unlike traditional workflows, AI Agents can dynamically analyze situations, choose the best course of action, and adapt their context, rather than following a rigid set of rules.
In Zoho Creator, an AI Agent allows users to define a skill for the agent using natural language, specify the required tools (such as Deluge functions), and after which an endpoint URL will be automatically generated. This URL can then be invoked to utilize the agent and perform the required tasks. AI Agent utilizes the Agentic AI model, where the AI agent acts similar to an agent performing a role fulfilling the purpose and goal that you define in a plain natural language. Each agent can be customized to fit your specific needs.
For example, say you have a Library Management App where library members can search for books and borrow them. You can create an agent called Book Suggester by instructing it to analyze a member's reading history and give personalized book recommendations. You also can ask it to send a Cliq message to the admin if any unavailable book is being searched frequently. This process involves analyzing reading history, suggesting books, and notifying the admin. Likewise, an AI agent can be created to perform multiple tasks in a logical, well-organized, and coherent manner to simplify your complex workflows
An AI Agent can be created in two simple steps:
- Defining the agent - The first step is to give an overall description of what the AI should be skilled in. This is given as simple instructions in plain natural language.
- Adding tools - The next step is to add the Deluge functions that the AI would need to perform the actions. The agent selects and executes the appropriate Deluge function based on the scenario to complete the required actions.
After creating an agent, an
endpoint URL will be generated. This endpoint URL is the URL or path to invoke the created agent wherever required. It can be called to interact with the agent and receive a response. In Creator, the agent can be called using
invokeURL task or
JS API via endpoint URL. Additionally, it can also be accessed from any external systems through API calls.
The request of this AI Agent will be a single prompt. After the agent processes all the tasks, it replies with a response. Both will be in natural language.
Sample Request
- {"data" :
- {
- "query" : "Check the feedback and update records accordingly"
- }
- }
Sample Response
- {
- "ai_response" : Feedback conveys a negative tone, added records successfully in the Feedback form. Notified in the Cliq chat
- }
Sample Deluge Script to invoke AI Agent
void invokeAISkill(string feedbackID)- {
- query = Map();
- query.put("query","Analyse the feedback, feedback_ID : " + feedbackID);
- data = Map();
- data.put("data",query);
- info data;
- resp = invokeurl
- [
- url :"https://www.zohoapis.com/creator/aiskill/zylker/feedback_analyser"
- type :POST
- parameters:data.toString()
- connection:"zoho_creator_connection"
- content-type:"application/json"
- ];
- summary = resp.get("ai_response");
- info summary ;
- }
3.1 How to access the AI Agent
AI Agent can be accessed using the endpoint URL.
What is an endpoint URL?
Endpoint URL is the unique path that allows external systems to communicate with the skill wherever required. It serves as an access point for invoking the agent and utilizing its functionalities. In Creator, this endpoint URL is automatically generated based on the agent name, with the last portion derived from the link name specified during creation. The endpoint URL will be in the below format:
Who can access this endpoint URL?
For AI Agent, the
user scope is set to
All by default. Hence, all the active users from the
Users module in the Creator account can invoke the endpoint URL. This does not include portal users.
How to invoke the endpoint URL?
- Using invokeURL deluge function
- Using JS API within widgets
Additionally, you can use it in any third-party services to invoke AI Agent.
Scope to invoke the AI Agent
Method
| Scope
|
POST
| ZohoCreator.aiskill.execute
|
3.2 Use case
Consider an Incident Handling application where users can add a list of technicians and their expertise, as well as the list of equipment for various incidents.
When unexpected incidents such as system failures, equipment malfunctions, and safety hazards occur, it requires quick and efficient responses. Delays in assessing the situation can lead to disruptions and inefficiencies. A process is needed that can swiftly manage the situation and rectify it with proper solutions.
To streamline this process, an AI agent can be created that acts as an incident manager. Whenever an issue is reported, the agent will analyze the problem, identify the expertise, assign the required equipment and technician. It can also provide a step-by-step resolution plan to address the incident efficiently. The issue can be reported in a form, and the agent will create the corresponding record in the report.
Use case implementation
For this use case, an AI Agent can be created with the below details:
1. Providing the below instructions to the AI
- You are an agent expert at providing Mitigation and containment details for an 24/7 Emergency Response & Remediation.
- Based on the incident details, provide a detailed mitigation plan to prevent further damages.
- Provide crew details, such as the Incident commander and details of the technicians.
- Also provide a list of equipment that would be needed by the responders
2. Adding Deluge functions necessary for the above tasks
- Available technicians - called for getting the list of available technicians.
- Available equipment - called for getting the list of available equipment.
- Create plan - called for assigning technicians, equipment, and developing a plan.
After creating an agent with the above details, an endpoint URL will be automatically generated through which the AI agent can be accessed.
In the live mode of the application, this endpoint URL can be invoked within a widget as shown below. By providing an incident description, the AI Agent can be triggered to execute all the instructed tasks.
3.3 Pre-requisites
Before you start, you need to
configure Zia with OpenAI. This ensures you provide the necessary permissions. AI Agent feature will be automatically enabled once you add the key for OpenAI.
Note: This feature is not yet supported in Zoho LLM.
- You also need the Deluge functions configured as per your requirement before you create an AI Agent.
3.4 Navigation guide
From the
home page of your Creator account, navigate to
Microservices and click the
+Create New button in the top-right side. Choose
Create in the
AI Models card and select the
AI Agent card under the
Build - Custom Models section. The
AI Agent builder will appear. You can create a new AI Agent as explained in
this page.
- Create and manage AI Agent
- Understand AI Modeler