AGENTIA

Multi-Agent RAG-Based Financial Analysis and Prediction System

---

Team Members

• Tavishi Gupta
• Lakshitha V
• Risu Kumari Nayak

---

Overview

AGENTIA is an AI-powered multi-agent financial analysis system designed to analyze company reports, retrieve relevant financial information, detect risks, generate predictions, and provide intelligent recommendations using Generative AI and Retrieval-Augmented Generation (RAG).

The project combines multiple specialized AI agents that collaboratively process financial documents and validate outputs to improve reliability and reduce hallucinations.

---

Features

• Multi-Agent AI Architecture
• Retrieval-Augmented Generation (RAG)
• Financial Report Analysis
• Risk Detection and Assessment
• Revenue and Profitability Analysis
• Future Trend Prediction
• Recommendation Generation
• Conflict Resolution Between Agents
• Reviewer-Based Output Validation
• PDF Document Processing
• Vector Database Retrieval

---

Architecture

User Input
   ↓
Planner Agent
   ↓
Retriever Agent (RAG)
   ↓
Worker / Analysis Agents
   ↓
Reviewer Agent
   ↓
Final Recommendation Engine

---

Agents Used

1. Planner Agent

Determines:

• Required analysis tasks
• Execution sequence
• Agent coordination

---

2. Retriever Agent

Responsible for:

• Retrieving relevant document chunks
• Searching vector database
• Providing contextual information

Technologies:

• ChromaDB / FAISS
• Sentence Transformers
• Embeddings

---

3. Financial Analysis Agent

Analyzes:

• Revenue
• Expenses
• Profit/Loss
• Debt
• Cash Flow

---

4. Risk Analysis Agent

Detects:

• Operational risks
• Market risks
• Regulatory risks
• Financial instability

---

5. Prediction Agent

Predicts:

• Growth potential
• Business stability
• Future opportunities
• Investment outlook

---

6. Reviewer Agent

Validates:

• Agent outputs
• Logical consistency
• Evidence-based reasoning
• Conflict resolution

---

Technologies Used

AI / LLM Frameworks

• LangChain
• LangGraph
• OpenAI GPT
• HuggingFace Transformers

Vector Databases

• ChromaDB
• FAISS

Backend

• Python
• FastAPI / Flask

Frontend

• Streamlit

Data Processing

• Pandas
• NumPy

---

RAG Pipeline

The project uses Retrieval-Augmented Generation (RAG):

1. Financial documents are uploaded
2. Documents are chunked
3. Chunks are converted into embeddings
4. Embeddings are stored in vector database
5. Retriever agent fetches relevant context
6. LLM generates grounded responses

This reduces hallucination and improves factual accuracy.

---

Validation Methodology

The system validates outputs using:

• Multi-agent cross verification
• Context-grounded retrieval
• Reviewer agent validation
• Evidence-supported reasoning

---

Installation

Clone Repository

git clone https://github.com/TavishiGupta057/agentia.git
cd AGENTIA

---

Create Virtual Environment

Windows

python -m venv venv
venv\Scripts\activate

Linux / Mac

python3 -m venv venv
source venv/bin/activate

---

Install Dependencies

pip install -r requirements.txt

---

Environment Variables

Create a .env file:

OPENAI_API_KEY=your_api_key

---

Run the Project

Build Vector Database

python ingest.py

OR

python build_vector_store.py

---

Start Application

Streamlit

streamlit run app.py

Flask

python app.py

FastAPI

uvicorn main:app --reload

---

Example Workflow

Upload Financial Report
        ↓
Retriever Fetches Context
        ↓
Agents Analyze Data
        ↓
Reviewer Validates Output
        ↓
Final Insights + Predictions

---

Example Output

• Revenue increased by 18%
• Debt risk is moderate
• Future growth probability is high
• Investment outlook is positive

---

Applications

• Financial Analysis
• Investment Research
• Business Intelligence
• Risk Assessment
• AI-Powered Decision Support

---

Academic Domain

• Generative AI
• Multi-Agent Systems
• Retrieval-Augmented Generation (RAG)
• Financial AI
• Natural Language Processing

---

Future Improvements

• Real-time stock integration
• Advanced forecasting models
• Multi-document comparative analysis
• Dashboard analytics
• Autonomous agent orchestration

---

License

This project is developed for academic and research purposes.

---

One-Line Summary

AGENTIA is a multi-agent RAG-based AI system that retrieves, analyzes, validates, and predicts financial insights from company reports using collaborative intelligent agents.