CONTENTS

Sl. No Content Page

Number
i List of figures I
ii List of tables II

1 Chapter 1-Introduction
1.1 Overview 1
1.2 Problem statement 1
1.3 Objectives 2
1.4 Sope of the project 2
1.5 Methodology overview 3
1.6 Project significance 3

2 Chapter 2 – Literature Survey

2.1 Design and Implementation of Artificial Intelligence 4
2.2
AI Agent Marketplace 4
2.3
Cooperative and Competitive Multi-Agent Systems 4
2.4
Confluence of Evolutionary Computation multiagent
2.5 The 4
Large Language Model Enhanced Multi-Agent Systems 5
2.6
Literature survey summary table 6

3 Chapter 3 – Overview of Proposed Methodology

3.1 Overview of Proposed System 7
3.2 System Architecture 8
3.3 Project modules description 8
3.4 Technologies and Tools used 9
3.5 Advantages of Proposed System 10

4 Chapter 4 – Requirement Specifications

4.1 Functional requirements 11

4.2 Non-functional requirements 12
4.3 Software requirements 12
4.4 hardware requirements 12
4.5 Other requirements 13
Sl. No

Page
Content
Number

5 Chapter 5 – Project Design

5.1 SystemDesign
5.1.1
User Interface Design 13
5.1.2
Data Flow diagram 14
5.1.3
Database design diagram 15
5.1.4
UML diagrams 16

5.2 Detailed Design

5.2.1User Interface (UI)Module 17
5.2.2
Qurey classification Module 18

6 Chapter 5- Implementation
6.1 Module-1: frontend implementation 19
6.2 Module-2: query classification module 20
6.3 Module-3: backend service implementation 21
6.4 Module-4: API intergration 22
6.5 Module-5: database implemation 23

7 Chapter 7- Testing
7.1 Objectives of Testing 24
7.2 Testing methodologies 25
7.3 Test cases and Results 26
7.4 Performance Evaluation 27

8 Chapter 8 – Results and Discussions

8.1 Results 28
8.2 Discussions 29
8.3 limitation
30
9 Chapter 9- Conclusion & Future scope
9.1 Conclusion 31
9.2 Future Scope 32
10 Chapter 10- Applications 33
34
11 References
 LIST OF FIGURES
Figure. No Figure Name Page. No
Figure 3.2 System architecture 8
Figure 5.1.1 User interface diagram 14
Figure 5.1.2 Dataflow diagram of Project 15
Figure 5.1.3 ER diagram of Database 16
Figure 5.1.4 UML diagram of project 17
Figure 5.2.1 UI interface 19
Figure 5.2.2 Query classification model 20
Figure 6.1 frontend 21
Figure 6.2 Query model 22
Figure 6.3 Backend implementation 24
Figure 6.5 Database implementation 25
Figure 8.1.a Dashboard 28
Figure 8.1.b Image Output
28
Figure 8.1.c Sample Output 29

I
 LIST OF TABLES
Table. No Table Tittle / Caption Page. No
Table 2.1 Literature survey summary 5

II
 AI AGGREGATOR

CHAPTER 1

INTRODUCTION
1.1. OVERVIEW
Artificial Intelligence has rapidly grown into a transformative force across almost every
industry. Today, AI systems assist with analyzing data, generating text and images, automating
business workflows, predicting financial trends, and supporting real-time decision-making. A
wide range of AI tools and services are available to users, including natural language models
like ChatGPT, image generation platforms such as Stability AI and DALL-E, and financial
analytics systems powered by market data [Link] their impressive power, these AI
services exist in isolation. A user who wants to perform different tasks — writing a document,
generating a graphic, analyzing stock data, or converting text into summarized insights — must
switch between multiple platforms, logins, interfaces, and data formats. This constant switching
interrupts workflow, creates inefficiency, and reduces productivity. Each platform follows its
own structure and terminology, making it difficult for users to combine results or compare
outputs effectively This growing fragmentation in the AI ecosystem highlights the need for a
unified environment where users can access multiple capabilities under one roof. A centralized
solution could intelligently route user requests to the right AI tool without requiring them to
choose manually or understand technical differences. The development of such systems marks
the next stage of AI usability, aiming to make intelligent services more convenient, accessible,
and seamless.

1.2. PROBLEM STATEMENT

The digital world is filled with specialized AI platforms, each designed for a specific type of
task. For example, text-based assistants excel at writing and reasoning, image models generate
visual assets, and finance APIs handle real-time market analysis. However, the problem arises
when a user needs to combine these capabilities within one workflow. The problem becomes
more evident when tasks involve multiple steps, such as writing content, producing visual
graphics, and analyzing business insights simultaneously. Existing AI platforms work well
when serving a single purpose, but they lack the ability to collaborate or integrate results
automatically. Users are required to manually gather outputs from separate services, convert
formats, interpret different response structures, and merge results on their own, which is time-
consuming and inefficient. The lack of a unified system also creates barriers for beginners, who
may not understand how to choose the most suitable tool for their requirements. As AI adoption

Department of Computer Science & Engineering, GMIT Page 1

 AI AGGREGATOR

increases, the need for an integrated solution that simplifies access to multiple AI services in
one place becomes essential.
Users often struggle when they need to complete multi-step tasks that require more than one AI
model. For instance, someone might need a written project description, an illustrative image,
and a budget analysis table. With existing systems, they must manually gather each piece
separately and attempt to assemble them themselves. This restricts creativity, slows down
delivery time, and increases frustration.

1.3. OBJECTIVES
Users face significant challenges when performing tasks that require multiple AI services,
because existing platforms operate independently without communication between them.
There is currently no unified interface that enables users to submit a single query and
automatically receive structured results generated by the most suitable model. To address
this challenge, the AI Aggregator project proposes the development of a unified platform
that brings together multiple AI models and services under a single, user-friendly interface.
Instead of forcing users to operate separate applications, the system intelligently interprets
the user’s query, identifies whether it relates to text processing, image generation, financial
analysis, or another category, and then routes the request to the most appropriate AI model.
This automated classification process reduces decision-making effort and ensures that the
best-suited tool is selected without requiring technical knowledge from the user. The
platform not only integrates multiple services but also standardizes and formats the
responses so that results are easy to understand, compare, and reuse. By delivering a
consistent structure for output presentation, the system improves clarity and usability. The
lack of integration among domain-specific AI services makes interactions inefficient and
inconvenient. There is a need for a centralized system that can: Allow users to access
multiple AI tools in one place automatically classify the type of request route the request to
the appropriate AI model deliver consistent, well-formatted, and easy-to-use results.

1.4. SCOPE OF THE PROJECT

The scope of the AI Aggregator project includes designing and developing a unified web-based
platform that integrates multiple Artificial Intelligence services into a single accessible
interface. The system focuses on enabling users to perform tasks such as text processing, image
generation, and financial data analysis without switching between different applications. It
supports intelligent query classification, automatic routing to suitable AI models, organized

Department of Computer Science & Engineering, GMIT Page 2

 AI AGGREGATOR

result presentation, and storage of past outputs for easy retrieval. The project also includes the
implementation of user authentication, feedback collection, and secure API communication.
The scope covers only the integration and orchestration of existing AI models rather than
building new AI algorithms. The system operates online and depends on real-time API
responses

1.5. METHODOLOGY OVERVIEW

The scope includes: The methodology followed in developing the AI Aggregator system is
based on a structured and sequential workflow that ensures smooth processing from user
input to final output delivery. The process begins when a user submits a query through the
web interface. The backend interprets the input and applies classification logic to determine
whether the request is related to text generation, image creation, or financial analysis. Once
classified, the request is routed to the appropriate external AI model using secure API calls.
After receiving the response, the system formats the output into a clear and standardized
structure that enhances readability and usability. The result is then displayed on the interface
in real time. Each interaction is stored in the database for history access, and users can
provide feedback to improve routing accuracy. This iterative approach enables better
performance and system refinement over time.

1.6. PROJECT SIGNIFICANCE

The AI Aggregator project holds significant importance in addressing the growing need for a
centralized and efficient workspace for accessing multiple AI services. With the increasing
availability of specialized AI tools across domains such as text processing, visual content
creation, and financial analytics, users often face difficulties switching between different
platforms to complete a single task. This system eliminates that inconvenience by offering a
unified interface that intelligently selects the most suitable model based on the user’s
requirement. By offering a unified environment, the platform transforms AI usage from a tool-
collection experience into a smooth, intelligent workflow, supporting students, researchers,
designers, businesses, and everyday users. This report is organized into several chapters that
discuss different aspects of the project in detail. The introduction provides background
information about the need for a unified AI platform and explains the motivation behind the
system. The literature survey reviews existing research studies and tools relevant to AI
integration. The proposed methodology describes the system workflow, architecture, and
module functionalities.

Department of Computer Science & Engineering, GMIT Page 3

 AI AGGREGATOR

CHAPTER 2
LITERATURE REVIEW
2.1 . Design and Implementation of Artificial Intelligence Platform
Dedicated to Power System (IEEE, 9664153)
This paper explores the need for an intelligent, unified platform capable of managing real-time
monitoring and decision-making within power systems. The authors highlight that traditional
power infrastructures operate in isolation, which limits efficiency and situational awareness. To
address this gap, they propose a fully integrated AI-based environment that brings multiple
intelligent modules together to improve operational accuracy. Although their solution
significantly enhances system visibility and control, it still lacks broader adaptability across
diverse energy networks. In the context of the proposed methodology, this limitation is
overcome by implementing a more scalable unified AI platform that can support complex
decision-making under varying system conditions.

2.2. AI Agent Marketplace (TPJES, 2025 – Paper ID 149)

This study discusses the challenge posed by AI tools functioning independently, forcing users to
switch between multiple platforms to complete different tasks. The authors propose the concept
of an open marketplace where AI agents can be published, discovered, reused, and
interconnected. Their framework enables collaborative service usage, making AI capabilities
more accessible. However, the approach falls short when it comes to real-time coordination and
seamless integration across diverse services. In the proposed methodology, this limitation is
addressed by enabling a unified environment where multiple AI tools work together through a
collaborative platform, ensuring easier access and interaction.

2.3. Cooperative and Competitive Multi-Agent Systems: From Optimization

to Games (IEEE/CAA Journal of Automatica Sinica, Vol. 9, No. 5, 2022)
This research analysis the complexity involved in coordinating multiple autonomous agents
operating within dynamic environments. The authors apply optimization and game-theoretic
models to support cooperative, negotiative, and competitive behaviour among agents. Their
approach provides structured interaction strategies for distributed problem solving. Despite
these strengths, the study offers limited practical validation and scalability under real-world
uncertainty. In the proposed work, this gap is addressed by introducing improved distributed

Department of Computer Science & Engineering, GMIT Page 4

 AI AGGREGATOR

decision-making models that enable agents to collaborate effectively in complex and changing
environments.

2.4. The Confluence of Evolutionary Computation and Multi-Agent

Systems: A Survey (IEEE/CAA Journal of Automatica Sinica, Vol. 12, No.
11, 2025)
This survey focuses on the need for adaptive learning methods in large-scale intelligent
systems. The authors examine hybrid approaches combining evolutionary algorithms with
multi-agent frameworks to enhance cooperation, coordination, and scalability. Although the
work provides deep theoretical insight, it lacks a clearly defined implementation model for
practical deployment. The proposed methodology resolves this limitation by applying
evolutionary enhancement techniques to multi-agent learning structures, enabling more
adaptable and scalable system behaviour.

2.5. Large Language Model Enhanced Multi-Agent Systems for 6G

Communications (IEEE Wireless Communications, 2024)
This paper addresses the difficulty large language models face when handling complex
communication tasks in emerging 6G environments. The authors introduce CommLLM, an
architecture where multiple specialized agents support planning, retrieval, and validation to
improve communication quality and performance. While the approach significantly enhances
task handling, it focuses more on conceptual design than full-scale operational deployment. The
proposed methodology extends this by offering a more robust intelligent communication
framework capable of supporting real-time decision-making in 6G systems.

Department of Computer Science & Engineering, GMIT Page 5

 AI AGGREGATOR

2.6. LITERATURE SURVEY SUMMARY TABLE:

Shortfalls in the
Sl
Publica author’s approach
. Problem addressed Authors
Tittle of the tion and solution provided
no in the paper approach
paper details in the proposed
methodology

A unified AI platform
Design and Developme
Traditional power combines multiple
Implementatio nt of a fully
systems lack a intelligent modules to
n of Artificial integrated
IEEE, unified intelligent improve monitoring
1 Intelligence AI
9664153 platform for real- accuracy and
Platform environmen
time monitoring and operational decision-
Dedicated to t for power
decision support. making in energy
Power System systems.
networks.

Introduces an ecosystem
Concept where AI agents can be
AI tools work
TPJES, framework published, shared, and
separately, making it
AI Agent 2025 for an open interconnected so users
2 difficult to discover
Marketplace (Paper marketplac can access services from
or use multiple AI
ID 149) e of AI one collaborative
services together.
agents. platform.

IEEE/C
Analysis of
AA
optimizatio
Cooperative Journal Coordinating Provides models
n and
and of multiple autonomous supporting negotiation,
game-
Competitive Automat agents becomes coordination, and
3 theory
Multi-Agent ica complex in dynamic competition among
strategies
Systems: From Sinica, problem-solving agents for effective
for multi-
Optimization Vol. 9, scenarios. distributed decision-
agent
to Games No. 5, making.
interaction.
2022

Department of Computer Science & Engineering, GMIT Page 6

 AI AGGREGATOR

IEEE/C Shows how evolutionary

The Confluence Large-scale Survey of
AA algorithms enhance
of Evolutionary intelligent systems hybrid
Journal agent learning and
Computation need adaptive evolutionar
4 of cooperation, enabling
and Multi-Agent algorithms to y and
Automat more adaptive and
Systems: A improve learning and multi-agent
ica scalable intelligent
Survey coordination. methods.
Sinica systems.

Introduces CommLLM,
Large Language Multi-agent where multiple
IEEE LLMs alone struggle
Model Enhanced enhanceme cooperative agents
Wireless with complex and
Multi-Agent nt handle planning,
5 Commu domain-specific
Systems for 6G architecture retrieval, and validation
nication communication
Communications around to improve
s, 2024 tasks.
LLMs. communication-related
performance in 6G.

Table 2.1 : Literature Survey Summary

Department of Computer Science & Engineering, GMIT Page 7

 AI AGGREGATOR

CHAPTER 3
OVERVIEW OF PROPOSED METHODOLOGY
3.1. OVERVIEW OF THE PROPOSED SYSTEM
System analysis is a critical phase in software development where the problem domain is
examined thoroughly in order to define the system requirements, evaluate existing limitations,
and formulate an optimized solution structure. For the AI Aggregator project, system analysis
helps identify the key challenges faced by users who interact with multiple domain-specific AI
services separately, such as natural language processing models, image generators, and
financial analysis tools. This chapter provides a detailed understanding of how the proposed
system bridges the gap between existing disjointed services and the need for an intelligent,
unified platform.
The analysis further explains the comparison between existing systems and the proposed
approach, highlights feasibility considerations, and discusses design strategies that ensure
technical, operational, and financial practicality.

3.2. SYSTEM ARCHITECTURE

In the current technological landscape, AI services are widespread but fragmented. Each service
excels at a particular task—text generation, code processing, image creation, financial
forecasting, or automated reasoning—but they remain isolated tools. For instance, a user might
depend on a text-based AI model for writing tasks, switch to a different application for image
generation, and use separate financial APIs for stock or market data. These tools do not
communicate or share common output standards.
Limitations of the Existing System

• Users must manually switch between multiple platforms to complete related tasks.
• Output formats vary significantly and require manual interpretation.
• No automatic decision-making or model selection exists based on user intent.
• Combining multiple AI outputs for research, creativity, or analytics becomes time-
consuming.
• Inconsistent interface experience reduces accessibility and increases cognitive load.
• Due to these drawbacks, productivity drops, workflows fragment, and advanced
automation becomes underutilized.

Department of Computer Science & Engineering, GMIT Page 8

 AI AGGREGATOR

Fig 3.2 System architecture of AI AGGERATOR

3.3. MODULES DESCRIPTION
Module Description with Explanation of Major Components
The AI Aggregator system is divided into several modules, each responsible for a specific task
within the workflow. These modules work together to provide a seamless experience, starting
from accepting user input to delivering processed AI-generated results. The modular structure
enhances scalability, maintainability, and smooth integration with multiple external AI services.

Major Modules and Their Explanation

1. User Interface Module
This module handles direct interaction with users through a web-based interface. It provides an input
area for entering prompts, displays generated results, and manages navigation to features such as
history and feedback. It ensures responsive performance and clear presentation of output in the
form of text, images, charts, or financial summaries. The UI communicates with the backend
using REST API calls and converts user actions into structured requests for processing.

2. Query Classification Module

This module analyzes the user’s input and identifies the type of request being made. It determines
whether the query belongs to text generation, image creation, or financial analysis. By using
pattern recognition and keyword-based classification, it assigns a category label to the request,
enabling correct routing. This component ensures that users do not need technical knowledge of
AI model selection.

Department of Computer Science & Engineering, GMIT Page 9

 AI AGGREGATOR

3. Routing and Orchestration Module

This module acts as the control centre of the system. Based on the classification result, it selects
the appropriate AI service and forwards the request for processing. It manages workflow
sequencing, handles request structure, and ensures proper coordination between modules. It also
handles error response cases, retries, and timeout conditions to maintain smooth operation.

4. API Integration Module

This component manages communication with external AI services such as OpenAI, Stability
AI, or financial APIs. It securely sends formatted requests and retrieves responses. It manages
API authentication, headers, tokens, error codes, bandwidth limits, and response conversions.
This abstraction prevents the rest of the system from needing to understand individual API
structures.

5. Output Formatting and Rendering Module

AI responses are often raw and unstructured. This module formats output into readable
structures, applying templates for consistency. It converts text into formatted paragraphs or
bullet structures, transforms data into tables or charts, and displays image results. This ensures
clarity and improves usability while maintaining a standardized output format regardless of the
service used.

6. Database Management Module

This module stores user information, query history, processed outputs, and feedback in a cloud
database such as firestore. It enables persistent storage and quick retrieval of previous results. It
also ensures authentication-based access control to protect user data. This allows users to track
previous tasks and improves system reliability.

7. Feedback Management Module

This module collects user ratings and comments on generated responses. The feedback is stored
in the database and analysed to refine routing rules and improve service accuracy over time. It
supports continuous learning and personalization by identifying which models produce the best
results for specific query types.

Department of Computer Science & Engineering, GMIT Page 10

 AI AGGREGATOR

Summary
These integrated modules work collaboratively to process user queries efficiently, route them
correctly, and deliver meaningful results in an easy-to-understand format. The modular
approach improves scalability, allowing additional AI services to be added in future
development stages.

3.4. TECHNOLOGIES AND TOOLS USED

The AI Aggregator project utilizes a combination of modern development technologies,
frameworks, and external services to deliver a responsive and scalable system. The frontend is
developed using [Link], supported by HTML, CSS, and JavaScript, providing a dynamic and
user-friendly interface. The backend is implemented using FastAPI with Python, enabling high-
performance request handling and efficient integration with third-party services. Communication
between the frontend and backend is achieved through RESTful APIs. The system integrates
external AI tools such as OpenAI for text processing, Stability AI for image generation, and
financial market data APIs for analytical output. A cloud-based database like Firebase Firestore
is used for storing user history, feedback, and authentication details. Version control is managed
through Git and GitHub, while Postman is used for testing API endpoints. Hosting and
deployment utilize platforms such as Netlify, and Render for seamless cloud deployment.

3.5. ADVANTAGES OF THE PROPOSED SYSTEM

The proposed AI Aggregator system offers several advantages over traditional isolated AI
tools. One of the major benefits is the convenience of accessing multiple AI services from a
single unified platform, eliminating the need to switch between different applications for
text, image, or financial tasks. This improves productivity and significantly reduces time
consumption. The system’s intelligent routing mechanism automatically selects the most
suitable AI model for each request, ensuring higher accuracy and eliminating the need for
technical expertise. Standardized output formatting enhances readability and provides a
consistent user experience across different services. The ability to store previous queries and
results adds value by allowing users to revisit and reuse information. Additionally, the built-
in feedback mechanism supports continuous learning and refinement, making the system
more efficient over time. Overall, the system enhances usability, scalability, and
performance while providing a user-friendly solution for diverse AI needs.

Department of Computer Science & Engineering, GMIT Page 11

 AI AGGREGATOR

CHAPTER 4
REQUIREMENTS SPECIFICATION
4.1. FUNCTIONAL REQUIREMENTS:
The system shall classify user queries into categories such as text-based tasks, image
generation, or financial data analysis.
•The system must route the request to the correct AI model based on classification.
•The system shall integrate external API services like OpenAI, Stability AI, and financial
APIs.
•The system must process API responses and convert them into standardized output
format.
•The system shall display results in structured form such as formatted text, tables,
images, or charts.
•The system must allow users to provide rating-based feedback for response quality
improvement.
•The system shall store query history and responses for user reference.
•The system must support multi-user access with session management.
•The system shall handle errors and provide meaningful error messages for failed
requests

4.2. NON-FUNCTIONAL REQUIREMENTS:

• Performance: The system should generate responses within acceptable time limits,
ideally between 5–20 seconds depending on service load.
• Scalability: The architecture must support additional AI models and increasing user
traffic without redesign.
• Reliability: The system must operate consistently and ensure dependable output even
under high demand.
• Usability: Interface should remain simple, intuitive, and accessible to both technical
and non-technical users.
• Security: All API credentials and user data must be protected using
encrypted connections and access control.
• Availability: System should maintain high uptime and operate continuously except
during planned maintenance.

Department of Computer Science & Engineering, GMIT Page 12

 AI AGGREGATOR

• Compatibility: The platform should support major browsers and function on desktop
and mobile screens.

4.3. SOFTWARE REQUIREMENTS:

• ProgrammingLanguages:
Python, JavaScript
• Frameworks: FastAPI, [Link]
• Database: Firebase / Firestore
• Tools & Libraries: Axios, Postman, GitHub, [Link] packages
• OperatingSystems: Windows / Linux / macOS supported for deployment
• Version Control: Git and GitHub for code management

4.4. HARDWARE REQUIREMENTS:

Development Machine:
• Processor: Minimum Intel i5 / Ryzen 5
• RAM: 8 GB or higher
• Storage: Minimum 256 GB SSD recommended
Server Requirements:
• Cloud Hosting for backend and database services
• Stable internet connection for API calls
Client System:
• Any device capable of running a modern browser
• Interface Responsiveness: support for desktop, laptop, and tablet devices
• Supported browsers: Chrome, Edge, Firefox, Safari

4.5. OTHER REQUIREMENTS:

• Users must have basic familiarity with using web-based platforms.
• Users should have a stable internet connection for accessing services.
• Users may need email-based or token-based authentication.
• No advanced technical skills required to operate the system.
• Dependent on external APIs for generating responses.
• Internet connectivity required for all operations.
• System performance influenced by third-party response time.
• Cost increases with API usage during high-volume operations.

Department of Computer Science & Engineering, GMIT Page 13

 AI AGGREGATOR

CHAPTER 5

PROJECT DESIGN
5.1. SYSTEM DESIGN:
5.1.1. User interface design

Fig 5.1.1 User interface diagram

Department of Computer Science & Engineering, GMIT Page 14

 AI AGGREGATOR

5.1.2. Data flow diagram

Fig: 5.1.2: data flow diagram of project

The workflow begins when the user submits a query to the system through the interface. The
system first receives and processes the input to check for validity and completeness. Once the
request is captured, it moves to the classification stage, where the system identifies whether
the query belongs to text generation, image creation, or financial analysis. Based on this
classification, the query is routed to the most suitable external AI service for processing. After
the AI model generates the response, the system formats and structures the output and stores it
for future reference. The formatted result is then displayed to the user. Additionally, the
system collects feedback regarding response quality, which is saved and later used to improve
decision- making and routing accuracy within the platform.

Department of Computer Science & Engineering, GMIT Page 15

 AI AGGREGATOR

5.1.3. Database design diagram (ER-diagram, schema etc.)

Fig 5.1.3: ER diagram of Database

Department of Computer Science & Engineering, GMIT Page 16

 AI AGGREGATOR

5.1.4. UML diagrams (use case, class, activity, sequence diagrams, etc.)

Fig 5.1.4: UML diagram of project

Explanation of Major Components
• Web Interface
• Handles interaction between users and the system.
• Receives raw input from users and forwards structured requests to backend
components.
• Displays processed responses and collects feedback after delivery.
• User Profile
• Stores user-specific information such as unique ID and preference settings.
• Helps personalize system behaviour and access stored history.
• Request
• Represents a user-generated query containing request ID, content, and metadata.
• Used as the main input model shared across system components.
• Query Classifier
• Determines the category of user input, such as text, image, or finance-related
tasks.
• Supports extraction of intent to help routing decisions.

Department of Computer Science & Engineering, GMIT Page 17

 AI AGGREGATOR

• Routing System
• Acts as the core decision-making unit.
• Routes incoming requests to the appropriate domain-based service.
• Maintains routing rules and manages service registration.
• Domain Service (Interface)
• Defines a standard structure for all domain-level AI services.
• Contains a method to process requests and return formatted responses.
• Text Gen Service, Image Gen Service, Finance Analysis Service
• Specific implementations of AI processing services for generating text,
generating images, and performing financial analytics.
• Each service uses the external AI API that matches its function.
• API Manager
• Manages external API communication including authentication keys and request
payload handling.
• Ensures correct formatting of outgoing API calls and response retrieval.
• External Response
• Represents unprocessed raw response data received from external AI services.
• Stores the raw model output and source identification.
• Response Standardizer
• Converts raw AI responses into a structured and readable format.
• Also supports merging responses when multiple services are involved.
• Response
• Final formatted result returned to the user.
• Contains output content and metadata.
• Data Base Log
• Responsible for saving request activity, generated responses, and operational
logs.
• Supports searching stored records for analytics.
• User Feedback Integration
• Manages user ratings and comments.
• Feeds feedback into routing
algorithms to optimize model
improvement performance over time.

Department of Computer Science & Engineering, GMIT Page 18

 AI AGGREGATOR

5.2. DETAILED DESIGN:

This section explains the internal design of the major modules in the AI Aggregator system.
Each module is illustrated with the help of design diagrams and followed by an explanation of
its functionality, workflow, and interactions with other components. The goal is to provide an
in-depth understanding of how the system processes user requests and delivers structured
outputs.

5.2.1 Module-1: User Interface (UI) Module

Design Diagram – UI Module

Fig 5.2.1 UI interface

The User Interface module is responsible for handling user interaction and providing an
accessible platform for entering requests. Users submit queries through a web-based interface
designed using React/HTML/CSS/JavaScript. This module sends input data to the backend
through REST APIs and displays the processed responses returned from AI models. It includes
screens for query input, output presentation, history view, and feedback submission. The UI
supports responsive design and works across multiple devices to ensure smooth accessibility.

Department of Computer Science & Engineering, GMIT Page 19

 AI AGGREGATOR

5.2.2 Module-2: Query Classification Module

Design Diagram – Query Classification

Module

Fig 5.2.2 Query classification model

Explanation:
The Query Classification module analyzes the text entered by the user to determine the
appropriate processing route. It identifies whether the request is related to text generation, image
creation, or financial data analysis using predefined rules and pattern recognition. The module
outputs a category label which is passed to the routing engine. This ensures users receive
relevant results without needing to know which AI model should be used. Accurate
classification helps improve performance and response accuracy.

Department of Computer Science & Engineering, GMIT Page 20

 AI AGGREGATOR

CHAPTER 6
PROJECT IMPLEMENTATION
6.1. MODULE-1: FRONTEND

Fig 6.1: Frontend

• Developed using [Link] for a responsive and interactive interface.
• UI components created for:
• User input area / prompt box

• Result display section

• History and feedback screens

• API response visualization (text, image, charts, tables)

• Axios/Fetch is used to send requests to the backend.

Department of Computer Science & Engineering, GMIT Page 21

 AI AGGREGATOR

6.2.
MODULE-2: QUERY CLASSIFICATION MODULE

FIG 6.2: QUERY CLASSIFICATION MODULE

Explanation:
• The function analyzes keywords to identify query type.
• Confidence score indicates reliability of classification.
• classify endpoint communicates results to both UI and routing engines.
• This achieves the system requirement of automated service selection.
• An Enum used to represent the three supported categories: TEXT, IMAGE, and
FINANCE.
• Helps the routing module easily understand what kind of service should be used next.
• Functional requirement achieved: “System shall classify the query into appropriate
service type.”

Department of Computer Science & Engineering, GMIT Page 22

 AI AGGREGATOR

Query Request and Query Response

• Pydantic models defining the structure of input and output for this API.
• Query Request holds the raw query from the user interface.
• Query Response returns the detected type and a basic confidence value.
• Objective supported: Standardized input and output formats between frontend and
backend.
• Classify query function
• Contains the core logic for query classification.
• Uses simple keyword checks to identify whether the query is financial, image-
related, or general text.
• In a real deployment, this can be replaced by an ML or NLP model without
changing the API interface.
• Functional requirement achieved: Automatic identification of task domain.
• Classify endpoint

• Exposed through FastAPI so the frontend can send user queries via HTTP POST.
• Calls classify query and return a structured response.
•
Objective achieved: Provides a clean service that other modules (like
routing/orchestration) can depend on.

Department of Computer Science & Engineering, GMIT Page 23

 AI AGGREGATOR

6.3 BACKEND SERVICE IMPLEMENTATION

Fig 6.3 backend implementation

The backend is developed in FastAPI due to its asynchronous support and high performance.
It handles routing, integration logic, and request management.

6.4 API Integration:

API integration plays a crucial role in the AI Aggregator system, enabling communication
between the backend and external AI service providers. The system uses secure REST API
calls to connect with multiple AI engines such as OpenAI for text processing, Stability AI
for image generation, and financial market APIs for real-time business and stock analytics.
Each service requires a unique request format and authentication process using encrypted
API keys stored securely in environment variables to prevent unauthorized access. The
backend formats user inputs and sends them as requests to the relevant AI model based on
query classification. Returned responses are processed asynchronously to improve
performance and reduce waiting time. Error-handling mechanisms, fallback responses, and
timeout controls are implemented to ensure system reliability even if external services are
busy or unreachable.

Department of Computer Science & Engineering, GMIT Page 24

 AI AGGREGATOR

6.5 Database Implementation:

Fig 6.5 Database implementation

The AI Aggregator system uses a cloud-based NoSQL database, such as Firebase Firestore, to
store and manage application data efficiently. The database maintains records including user
profiles, login authentication details, submitted queries, generated responses, and system
feedback. A structured collection-based schema is used to improve data retrieval time and
organization. Data is stored with unique identifiers such as user id, request id, and response id
to support secure relational mapping. The database enables users to access their previous
results through history features while maintaining privacy through role-based access control.
Automatic timestamps track activity history for analytical purposes. Indexing is applied to
improve search and filtering performance. Backup and synchronization capabilities ensure data
consistency across devices. This implementation supports scalability and future expansion
without major architectural changes.

Department of Computer Science & Engineering, GMIT Page 25

 AI AGGREGATOR
AI AGGREGATOR

CHAPTER 7
TESTING
7.1. OBJECTIVES OF TESTING:
▪ Verify that system functionality works as expected across all modules.
Ensure proper integration between frontend, backend, and external AI
▪ services.
▪ Validate performance efficiency and accuracy of
output. Detect and correct faults before final
deployment.
▪
7.2. TESTING METHODOLOGIES:
Unit Testing

• Individual functions tested, such as query classifier, routing engine, and API handlers.

Integration Testing

• Checked communication among UI, backend, database, and external APIs.

System Testing

• Verified full workflow from user input to AI response rendering.

Functional Testing

• Confirmed that features perform according to requirements.

User Acceptance Testing

• Conducted with sample real users to measure usability and satisfaction.

Performance Testing

• Tested response time, concurrency handling, and throughput limits.

7.3. TEST CASES AND RESULTS:

• Postman – API endpoint validation

• Jest / React Testing Library – Frontend unit tests

• Py Test – Backend logic testing

• Firebase Emulator – Database testing environment

Department of Computer Science & Engineering, GMIT Page 26

 AI AGGREGATOR
AI AGGREGATOR

7.4. PERFORMANCE EVALUATION:

• Response time remains within acceptable range for most tasks.

• Asynchronous request pipeline improves efficiency.

• Database responses retrieved with low latency.

• Minimal performance degradation under multiple active users.

Department of Computer Science & Engineering, GMIT Page 27

 AI AGGREGATOR
AI AGGREGATOR

CHAPTER 8

RESULTS AN DISCUSSION
8.1. RESULTS: D

FIG 8.1(A): DASHBOARD

FIG 8.1(B): IMAGE OUTPUT

Department of Computer Science & Engineering, GMIT Page 28

 AI AGGREGATOR
AI AGGREGATOR

FIG 8.1(C): SAMPLE OUTPUT

8.2. DISCUSSIONS:
• The results indicate that integrating multiple AI tools within one platform enhances
efficiency and productivity for different types of users.

• Compared to traditional standalone AI applications, the unified approach reduces

complexity and eliminates repetitive login and switching processes.

• The intelligent routing system ensures that users do not require technical knowledge
about model selection, making AI more accessible.

• Standardized output formatting increased clarity and facilitated easier understanding

and usage of generated results.

• The feedback mechanism strengthens the continuous improvement cycle and supports
future personalization.

• Limitations found include dependency on third-party API responsiveness and internet

connectivity constraints.

• The performance and experience of the system showed strong potential for scaling with
additional advanced AI services.

• The AI Aggregator system successfully integrated multiple AI services such as text

generation, image creation, and financial analysis into a single unified platform.
Department of Computer Science & Engineering, GMIT Page 29
 AI AGGREGATOR
AI AGGREGATOR

• The query classification module accurately identified and categorized user requests with
high precision.
• The routing engine correctly directed requests to relevant AI models without requiring
manual model selection by the user.
• The system produced standardized and structured outputs that were easy to read and
understand.
• Performance tests demonstrated smooth and stable operations with acceptable response
times.
• Users reported improved workflow efficiency when compared to working with multiple
separate tools.

8.3 Limitations

• Heavily dependent on external API performance and reliability.

• Requires continuous internet connectivity.

• API usage may involve operational cost for large-scale usage.

• System currently supports limited domains (text, image, financial); future expansion
needed.
Response times may vary based on external server load.

Department of Computer Science & Engineering, GMIT Page 30

 AI AGGREGATOR
AI AGGREGATOR

CHAPTER 9
CONCLUSION& FUTURE SCOPE

9.1. CONCLUSION:
• The AI Aggregator project successfully demonstrated the implementation of a unified
AI platform that combines multiple AI models into a single efficient interface.
• The system allows users to perform diverse tasks without switching between various
tools or understanding internal model details.
• The architectural approach using modular integration, intelligent routing, and structured
formatting proved effective in delivering consistent results.
• The project offers a functional prototype that supports automation, real-time processing,
and improved user productivity.
• Although certain limitations exist related to API dependency and performance
variations, the system provides a strong foundation for future scaling and enhancement.

9.2. FUTURE SCOPE:

• The Implement support for additional AI services such as speech-to-text, video
generation, and advanced data analytics.
• Enable real-time collaboration features for multi-user project work.
• Add offline features and local caching to reduce dependency on uninterrupted internet.
• Integrate secure payment gateways and subscription systems for commercial use.
• Expand to mobile applications for Android and iOS.
• Include AI model comparison features and custom model selection options for power
users.
• Enhance personalization using advanced preference tracking and recommendation
engines.
• Incorporate model performance tracking algorithms to improve routing decisions
automatically.

Department of Computer Science & Engineering, GMIT Page 31

 AI AGGREGATOR

CHAPTER10
APPLICATIONS

Education : MULTIMEDIA AGENT

• To converttext tospeech/video generation, to improve personalize learning and academic outcomes,
to get a learning resource and consolidating data on student performance.

Finance: Fraud Detection and Risk Assessmen

• Monitor complex, real-time transactions to flag suspicious activity quickly and accurately.

Healthcare: Automated Diagnostics and Triage

• Provide rapid, accurate initial assessment of medical images or patient records.

E-commerce: Dynamic Pricing and Forecasting

• Optimize product pricing in real-time to maximize profit margins.

Manufacturing and Industry:

• In the manufacturing sector, AI aggregators are transforming operations by consolidating data
from various production lines, sensors, supply chains, and maintenance logs to enhance
efficiency and reduce costs.

Agriculture:
• AI aggregators are revolutionizing agriculture by collecting and analyzing data from a variety
of sources, including sensors, satellite imagery, weather forecasts, and crop health data.

Department of Computer Science & Engineering, GMIT Page 32

 AI AGGREGATOR

REFERENCES

[1] Z. Wang et al., “Design and Implementation of Artificial Intelligence Platform

Dedicated to Power System,” IEEE, doi: 10.1109/9664153.

[2] M. Praveen, “AI Agent Marketplace,” TechPioneer Journal of Engineering & Science
(TPJES), vol. 4, no. 1, 2025, Article ID 149.

[3] J. Wang, Z. Li, and L. Li, “Cooperative and Competitive Multi-Agent Systems: From
Optimization to Games,” IEEE/CAA Journal of Automatica Sinica, vol. 9, no. 5, pp. 763–
783, May 2022, doi: 10.1109/JAS.2022.105506.

[4] H. Chen et al., “The Confluence of Evolutionary Computation and Multi-Agent

Systems: A Survey,” IEEE/CAA Journal of Automatica Sinica, vol. 12, no. 11, pp. 2175–
2193, Nov. 2025, doi: 10.1109/JAS.2025.125246.

[5] D. Wu et al., “Large Language Model Enhanced Multi-Agent Systems for 6G

Communications,” IEEE Wireless Communications, 2024, doi: 10.1109/MWC.016.2300600.

[6] R. Verma and S. Chen, "Stateful Orchestration of LLM Agents for Complex Query
Routing," in Proc. IEEE International Conference on AI and Robotics (ICAIR), New
Delhi, India, 2024, pp. 210-217.

[7] A. Khan and D. Jones, " Engineering for Multi-Domain Reasoning in AI Routing Systems,"
IEEE Access, vol. 12, pp. 11345-11358, Feb., 2024.

Department of Computer Science & Engineering, GMIT Page 33