International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 04 | Apr 2025
p-ISSN: 2395-0072
www.irjet.net
EASY GO VENTURE: A Multi-Agent AI System for Destination Insights Using Modular Taskflow AI Pipelines Yashika, Suman Pattnaik, Suresh D, Surya koundinya C 1Yashika, Student, Jain University, Karnataka, India
2Suman Pattnaik, Student, Jain University, Karnataka, India 2Suresh D, Student, Jain University, Karnataka, India
2Surya Koundinya C, Student, Jain University, Karnataka, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The emergence of intelligent travel solutions
The coordination amongst agents—each driven by language and supplemented by external tools and APIs—is the fundamental breakthrough. For instance, the Travel Agent retrieves real-time flight information from the Amadeus API, while the Web Research Agent uses Wikipedia and SerperSearch to collect contextual data. After that, a Reporter Agent compiles the results into an approachable format. Additionally, the architecture adheres to the concepts of Retrieval-Augmented Generation (RAG), which enables agents to make informed judgements beyond the static knowledge of the model by referencing real-time data.
has increased the demand for comprehensive and real-time trip planning systems. TripPlanner, a production-ready, modular system driven by Multi-AI Agents utilizing the Taskflow AI framework, is presented in this paper. Web Research, Travel, and Reporter Agents are specialist agents that use external technologies such as SerperSearch, Wikipedia APIs, Amadeus Flights API, and Weather.com to collect and aggregate data in order to process a single question. GPT-3.5 Turbo balances memory and API limitations while enabling reasoning. The system employs Retrieval-Augmented Generation (RAG) to boost accuracy and real-time awareness. Analysis reveals that TripPlanner provides incredibly pertinent and eye-catching trip information. It opens the door for more extensive uses outside of travel by showcasing the usefulness and scalability of agent coordination.
The following are the contributions made by this work: • A scalable and modular architecture that uses AI agents to generate itineraries. • The incorporation of external tools and APIs for pertinent, real-time data.
Key Words:
Multi-Agent Systems, Trip Planning, Retrieval-Augmented Generation, Taskflow AI, GPT-3.5, Real-Time Insights, Modular AI, Travel Itinerary
• The Taskflow framework facilitates effective task decomposition and agent collaboration.
1.INTRODUCTION
• An economical implementation that takes memory and API limitations into account.
Artificial intelligence (AI) has changed how people use digital platforms in recent years, particularly in areas like research, travel, and tailored suggestions. Conventional travel planning is looking through numerous websites, evaluating flight alternatives, examining weather predictions, and reading a tonne of blogs or articles. In addition to being time-consuming, this procedure lacks a cohesive, intelligent interface that can accommodate a wide range of changing user preferences.
This paper's remaining sections are arranged as follows: Related work is discussed in Section II. The system architecture is shown in Section III, and the roles and toolchains of each agent are explained in Section IV. Use cases and implementation are described in Section V. Evaluation and results are presented in Section VI, and the paper's conclusion with recommendations for the future is covered in Section VII.
In order to overcome these constraints, this paper presents TripPlanner, an intelligent itinerary creation system that uses several independent AI agents and is based on the Taskflow AI framework. The TripPlanner system uses a modular and multi-agent strategy that breaks down user enquiries into smaller tasks that are performed by specialised agents, in contrast to monolithic AI models that try to answer general questions in isolation. Every agent is made to concentrate on a particular task, such gathering travel data, assessing the weather, or summarising content.
© 2025, IRJET
|
Impact Factor value: 8.315
2. RELATED WORK Intelligent trip planning research has progressed from early rule-based systems to contemporary designs that integrate multi-agent coordination and retrievalaugmented generation (RAG). The main goal of early travel recommender systems was to tailor suggestions based on static databases and past user information. Recent studies on recommender systems for sustainable tourism have pointed out that these systems frequently
|
ISO 9001:2008 Certified Journal
|
Page 1073