This work presents a GenAI-based system for the creation of dynamic and functional graphical interfaces generated directly from user requests expressed in natural language. To achieve this objective, the system leverages the LangGraph framework to design a multi-agent architecture in which each agent is assigned to a specific task, defined through a dedicated prompt and supported by tool-based function calling. The project focuses specifically on the automatic generation of forms used to collect user information for producing insurance quotes, which type is specified by the prompt given by the user to start the conversation. The system identifies the requested quote type and, since the form compilation is organized into multiple phases, it dynamically retrieves the appropriate fields from the database according to the current value of phase value in the state, ensuring consistency and avoiding repetitions. The resulting system is capable of autonomously generating the entire structure of the user interface, which is then rendered by the front-end using Angular Material components. Overall, the proposed architecture represents a solid foundation for future enhancements, which include expanding the types of available insurance quotes, broadening the set of graphical elements the system can insert in the generated interface, and enriching each component with additional features, allowing for more tools to customize the interfaces and make them as complete as possible.
Questa tesi descrive un sistema basato sull'Intelligenza Artificiale Generativa per la creazione di interfacce grafiche dinamiche e funzionali generate direttamente a partire dalle richieste di un utente espresse in linguaggio naturale. Per raggiungere questo obiettivo, il sistema sfrutta il framework LangGraph per progettare un’architettura multi-agente, in cui a ciascun agente è assegnato un compito specifico, definito tramite un prompt dedicato e supportato da strumenti basati sul function calling. Il progetto si concentra in particolare sulla generazione automatica di form destinati alla raccolta delle informazioni necessarie alla creazione di preventivi assicurativi, la cui tipologia è specificata dal prompt fornito dall’utente per iniziare la conversazione. Il sistema identifica la tipologia di preventivo richiesta e, poiché la compilazione del form è organizzata in più fasi, estrae dinamicamente dal database i campi da richiedere in base al valore corrente della fase nello state, garantendo coerenza ed evitando ripetizioni. Il sistema risultante è in grado di generare autonomamente l’intera struttura dell’interfaccia utente, che viene poi renderizzata dal front-end utilizzando i componenti di Angular Material. Nel complesso, l’architettura proposta rappresenta una solida base per futuri miglioramenti, che includono l’ampliamento delle tipologie di preventivi disponibili, l’estensione dell’insieme di elementi grafici che il sistema può inserire nell’interfaccia generata e l’arricchimento di ciascun componente con ulteriori funzionalità, consentendo una maggiore personalizzazione delle interfacce e rendendole il più complete possibile.
Generazione automatica di interfacce dinamiche Angular tramite un sistema multi-agente basato su LangGraph
ILLIANO, ALESSIA
2024/2025
Abstract
This work presents a GenAI-based system for the creation of dynamic and functional graphical interfaces generated directly from user requests expressed in natural language. To achieve this objective, the system leverages the LangGraph framework to design a multi-agent architecture in which each agent is assigned to a specific task, defined through a dedicated prompt and supported by tool-based function calling. The project focuses specifically on the automatic generation of forms used to collect user information for producing insurance quotes, which type is specified by the prompt given by the user to start the conversation. The system identifies the requested quote type and, since the form compilation is organized into multiple phases, it dynamically retrieves the appropriate fields from the database according to the current value of phase value in the state, ensuring consistency and avoiding repetitions. The resulting system is capable of autonomously generating the entire structure of the user interface, which is then rendered by the front-end using Angular Material components. Overall, the proposed architecture represents a solid foundation for future enhancements, which include expanding the types of available insurance quotes, broadening the set of graphical elements the system can insert in the generated interface, and enriching each component with additional features, allowing for more tools to customize the interfaces and make them as complete as possible.| File | Dimensione | Formato | |
|---|---|---|---|
|
Illiano_Alessia.pdf
accesso aperto
Dimensione
1.93 MB
Formato
Adobe PDF
|
1.93 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/102087