Intent Classifier with LLM tool calls

https://www.pexels.com/photo/flashing-sparks-coming-from-the-angle-grinder-9665360/

Intent Classification Problem, ICP has been around for many years. This solution is to categorize user inputs into predefined intents, which represent the user's goal or desired action. It's a crucial process in any Multi Agent Systems MAS. There are many approaches to solve this problem with high accuracy (above 0.9). Here we are using LLM with tool calls to solve ICP.

The source code is available here.

Here is the predefined intents. I am using a restaurant for the use case. 

user intention can be making appointment, ordering food via AI Agent, product inquiry, etc. These intents are converted in LLM tools definition (ChatCompletionToolParam)  and send it along with the request to generative AI.

 Here is a demo (no audio). Please watch in full screen mode.

Good to read

1. How to Build an AI Agent With Semantic Router and LLM Tools

2. Intent recognition and entity extraction options for your intelligent application workloads

3. Intent detection and management with chatGpt API (Discuss around using function calls)

4. Agentic AI — III : Understanding LLM Parallelization and Routing,Tool Calling and Function Calling