ASAPI kicks off, the project for the automatic generation of Conversational Assistants from Apps Programming Interfaces

We are immersed in a process of constant digital transformation, in which conversational assistants are becoming more and more important, consolidating themselves as a key tool in multiple sectors.

This is how the ASAPI project was born, which answers to the growing demand for more natural and intuitive interfaces, integrating the latest advances in Artificial Intelligence such as large language models or LLMs.

Thus, the main objective of ASAPI is the automation of the generation of conversational assistants based on information extracted from APIs and websites, addressing two main types of assistants: 

1. Consulting assistants: They use Retrieval Augmented Generation (RAG) methods to provide accurate answers, directly accessing updated information from reliable web sources.

2. Form Completion Assistants: They are designed to interact with users using natural language, facilitating in a clear and efficient way the process of filling out online forms.

Both assistants are designed to operate automatically, eliminating the need for complex manual configurations. In addition, the system will be reinforced with monitoring technology that will ensure reliability, minimizing model hallucinations and validating that both responses and data received are consistent and truthful.

Some of ASAPI's technological challenges include automating the extraction of data models from APIs and forms, adapting language models to conversational interaction tasks, developing supervised techniques to validate and correct information, and applying RAG strategies to retrieve updated data from web sites.

ASAPI aims to develop a system capable of automatically generating functional and robust conversational assistants for forms and websites, combining the benefits of LLMs with supervised techniques, providing accurate and reliable responses.

The ASAPI project is a collaborative initiative between Zylk, Airlan, Innotu and Vicomtech, and has been funded by the Basque Government within the framework of the Hazitek 2024 program.  ASAPI aims to develop a system capable of automatically generating functional and robust conversational assistants for forms and websites, combining the benefits of LLMs with supervised techniques, providing accurate and reliable responses.

The ASAPI project is a collaborative initiative between Zylk, Airlan, Innotu and Vicomtech, and has been funded by the Basque Government within the framework of the Hazitek 2024 program.