AirOps
Guide for using Ionic with AirOps
Introduction
AirOps is one of our favorite frameworks with a user-friendly platform for creating AI agents. This application allows for easy integration with tools like the Ionic API, enabling users to build functional AI assistants without complex coding. It’s particularly known for its simple “what you see is what you get” editor and diverse customization options.The project’s objective is to create an interactive agent for holiday gift shopping assistance using various tools, including AirOps, itegration to an LLM (in this case GPT-3 Turbo), and Ionic API’s OpenAI Plugin.
Video tutorial
A video tutorial of this guide is embedded below and can also be found HERE
Starting the project
The project begins by accessing the AirOps template selection page and choosing the blank agent template. Initially, users are presented with a guide, but for simplicity, this guide will be skipped.
Click on Create New App
Click on Use Template
Configuring the Agent
In this phase, GPT-3 Turbo is selected as the model. The agent’s role and functionalities are defined through the system prompt setup, focusing on creating a granular configuration for precise task handling. The configuration interface will be displayed in an image.
Click on the “Configure” button in the Agent box
Agent configuration
The agent has many configuration options. The most important is the System message or System prompt window. You can be as granular as you would like but you should include some combination of a) what the agent should do and b) let it know how and when to use the Ionic API plugin:
Other configuration options include setting the temperature, specifying the opening remarks, tool visibility and memory settings, and giving specific training examples to improve output accuracy. When configured as desired click on “Save Changes”
Integrating Ionic
The next step is to integrate Ionic. This will be done via an OpenAI plugin. The Ionic integration is what empowers our agent to find and suggest gifts. We’ll go through a simple configuration process to align it with our needs:
Click OpenAI Plugin on the side bar menu to add into the visual editor
Ensure the OpenAI Plugin box is linked to the Agent box and click on the “Configure” bar when ready
In the “Manifest URL” enter the Ionic OpenAI plugin API. You can copy and paste from the section below:
Describe how the tool should be used. Feel free to experiment for an optimized output but something to the effect of the text below should be utilized:
Click on Save Changes
Testing and refinement
Perform a test to visualize the functionality and output. Our test scenario involves finding a gift for a three-year-old interested in school buses.
Click on Test All
Issue a query
In our case, “Can you help me find a gift for my nephew? He is 3 years old and loves school busses.”
Ionic will return results that match the query from a variety of vendors
Finalizing and Sharing
Save and publish the agent
Click on Save Draft
Name the agent
Publish App
The app is shareable via a URL
Intermediate features
Monitoring performance and logs
Bot logs and history are can be analyzed to understand the performance, and this aspect is represented in the History tab
Integrations options
Various integration options, like widgets and webhooks, are available, and these options are located in the Integration tab
Advanced: Attribution via Ionic API Key
Follow our Guide to using Ionic API Keys