https://github.com/yashdesai023/chatbot-ai-project

An intelligent Retrieval-Augmented Generation (RAG) chatbot that leverages Google’s Gemini 2.0 Flash model and Pinecone Vector Database to provide accurate, context-aware answers for UPSC exam preparation.
https://github.com/yashdesai023/chatbot-ai-project

chatbot flask-application gemini langchain llm pineconedb python rag upsc-chatbot

Last synced: 2 months ago
JSON representation

An intelligent Retrieval-Augmented Generation (RAG) chatbot that leverages Google’s Gemini 2.0 Flash model and Pinecone Vector Database to provide accurate, context-aware answers for UPSC exam preparation.

• Host: GitHub
• URL: https://github.com/yashdesai023/chatbot-ai-project
• Owner: yashdesai023
• Created: 2025-10-06T14:59:14.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2025-10-07T18:22:36.000Z (9 months ago)
• Last Synced: 2025-10-07T20:33:19.035Z (9 months ago)
• Topics: chatbot, flask-application, gemini, langchain, llm, pineconedb, python, rag, upsc-chatbot
• Language: Jupyter Notebook
• Homepage:
• Size: 17.6 KB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# 🦅 Enterprise Knowledge Graph & RAG Pipeline

**High-Accuracy Contextual Retrieval Engine for Complex Knowledge Domains**

[![Python](https://img.shields.io/badge/Python-3.10%2B-blue)](https://github.com/yashdesai023/chatbot-ai-project)

[![GenAI](https://img.shields.io/badge/GenAI-Gemini%202.0%20Flash-8A2BE2)](https://github.com/yashdesai023/chatbot-ai-project)

[![VectorDB](https://img.shields.io/badge/VectorDB-Pinecone-green)](https://github.com/yashdesai023/chatbot-ai-project)

## 📌 Executive Summary

This project is a production-grade **Retrieval-Augmented Generation (RAG)** system designed to handle high-stakes knowledge retrieval (originally optimized for UPSC exam data). It demonstrates a 28% improvement in contextual relevance over base LLM models and a 45% reduction in retrieval latency.

## 🚀 Engineering Highlights

- **Optimized Latency:** Reduced response time from 1.7s to **0.9s** via intelligent caching and context-window tuning.

- **Precision Retrieval:** Integrated **MiniLM embeddings** and semantic reranking to ensure high-accuracy responses for complex queries.

- **Scalable Backend:** Built with a clean Flask architecture designed for horizontal scalability and high-concurrency handling.

## 🏗️ Architecture

```mermaid

graph LR

    A[Raw Document] --> B(PDF Parser)

    B --> C(Recursive Chunker)

    C --> D[Embeddings Model]

    D --> E[(Pinecone Vector DB)]

    F[User Query] --> G(Semantic Search)

    G --> E

    E --> H(Contextual Reranker)

    H --> I[Gemini 2.0 Flash]

    I --> J[Actionable Answer]

```

## 🛠 Tech Stack

- **LLM:** Google Gemini 2.0 Flash

- **Vector Store:** Pinecone

- **Backend:** Flask, Python

- **Embeddings:** MiniLM-L6-v2

- **Tools:** LangChain, Postman, Docker

---

**Author:** Yash Desai  

**License:** MIT