thilakx/Bankoflibreo

Bank of Libreo — Intelligent Banking RAG System

An enterprise-grade Retrieval-Augmented Generation (RAG) system for bank data analysis, document intelligence, and AI-powered query answering.

---

🖥️ Live Preview

The Oracle of Libreo — RAG-powered chat interface connected to ChromaDB vector store

---

📐 System Architecture & Flow

flowchart TD
    subgraph Frontend["🖥️ Frontend (React + Vite)"]
        UI["User Interface"]
        APIClient["API Service Layer"]
        UI --> APIClient
    end

    subgraph Backend["⚡ FastAPI Server"]
        API["FastAPI App"]
        DocProc["Document Processor"]
        RAGEngine["RAG Search & Synthesis"]
    end

    subgraph Database["🗄️ ChromaDB Vector Store"]
        Embeddings["Sentence Transformers"]
        VectorDB["Chroma Vector Database"]
    end

    APIClient -->|"1. User Query / Upload"| API
    API -->|"2. Document Ingestion"| DocProc
    DocProc -->|"3. Text Chunks & Metadata"| Embeddings
    Embeddings -->|"4. Index Vectors"| VectorDB
    API -->|"5. Search Query"| RAGEngine
    RAGEngine -->|"6. Similarity Search"| VectorDB
    VectorDB -->|"7. Top-K Context"| RAGEngine
    RAGEngine -->|"8. Synthesized Answer"| API
    API -->|"9. JSON Response"| APIClient

📊 View Full Architecture Diagram

---

✨ Features

Feature	Description
📄 Multi-Format Ingestion	PDF, CSV, Excel .xlsx, Word .docx, TXT, Markdown, JSON, HTML
🔍 Semantic Vector Search	ChromaDB + Sentence Transformers for semantically accurate retrieval
🤖 AI Answer Synthesis	Blended, paraphrased answers with cited sources
⚡ Auto Dataset Loading	Background ingestion of banking datasets on server startup
🧩 Chunk Inspector	Drill down into retrieved context chunks per answer
🎨 Premium UI	Classical-themed chat interface with glassmorphism & dark palette
☁️ Render Ready	One-click deployment to Render (backend + frontend)

---

🛠️ Technology Stack

Layer	Technology
Backend API	Python 3, FastAPI, Uvicorn
Vector Store	ChromaDB, Sentence Transformers
Document Parsing	PyPDF2, Pandas, OpenPyXL, Python-Docx
Frontend	React 19, Vite 8, Tailwind CSS v4
Deployment	Render (Web Service + Static Site)

---

🚀 API Reference

Method	Endpoint	Description
GET	/	Root health check
GET	/healthz	Platform health monitoring
GET	/api/health	Vector store health & document statistics
GET	/api/documents	List all indexed documents
POST	/api/upload	Upload & auto-chunk document files
POST	/api/query	RAG query → synthesized answer + citations
POST	/api/ingest-kaggle	Auto-download Kaggle bank dataset
DELETE	/api/documents/{filename}	Remove a document from vector store

---

💻 Local Development

Prerequisites

• Python 3.10+
• Node.js 18+

⚡ Quick Start (One Command)

# From repo root — launches both servers in separate windows
.\start_dev.ps1

Manual Setup

Terminal 1 — Backend

cd backend
pip install -r requirements.txt
uvicorn app:app --host 127.0.0.1 --port 8000 --reload

Backend runs at http://localhost:8000

Terminal 2 — Frontend

cd frontend
npm install
npm run dev

Frontend runs at http://localhost:5173

Note: Vite takes ~25s on first start while bundling dependencies. Wait for the VITE ready message.

---

☁️ Deployment Guide (Render)

🔧 Backend Service (Web Service)

Setting	Value
Environment	Python 3
Root Directory	backend
Build Command	pip install -r requirements.txt
Start Command	uvicorn app:app --host 0.0.0.0 --port $PORT

🌐 Frontend Service (Static Site)

Setting	Value
Root Directory	frontend
Build Command	npm install && npm run build
Publish Directory	dist
Environment Variable	VITE_API_BASE_URL = https://your-backend-url.onrender.com

---

Built with ⚡ by Bank of Librio — Banking with Strength. Built on Trust.