A multi-agent RAG (Retrieval-Augmented Generation) system powered by Google Gemini, Docling, and LangGraph for intelligent document Q&A with built-in fact-checking and hallucination prevention.
Key Features:
Multi-Agent Architecture
🔍 Relevance Checker - Validates if documents contain information to answer your question
📚 Research Agent - Analyzes retrieved content and generates initial responses
✅ Verification Agent - Cross-checks responses against original documents to detect hallucinations
🔄 Self-Correction - Automatically re-runs research if contradictions or unsupported claims are found
Hybrid Retrieval System
BM25 Keyword Search - Finds exact matches and specific terminology
Vector Embeddings - Captures semantic meaning and context
Ensemble Retriever - Intelligently combines both approaches for optimal results
Advanced Document Processing
Powered by Docling - IBM’s state-of-the-art document parser
OCR Support - Handles scanned documents and images
Multiple Formats - PDF, DOCX, TXT, Markdown
Smart Caching - Avoids reprocessing unchanged documents
User-Friendly Interface
Gradio Web UI - Clean, intuitive interface
Example Documents - Pre-loaded examples to get started
Real-time Verification - See fact-checking results alongside answers

Source code: https://www.twine.net/signin

AI Icebreaker Bot project! I’ve successfully built a powerful application that leverages Retrieval-Augmented Generation (RAG) and Large Language Models to transform networking and professional connections. By integrating IBM watsonx.ai’s capabilities with the LlamaIndex framework, I’ve created a tool that can analyze LinkedIn profiles and generate personalized conversation starters bridging the gap between data and meaningful human interaction.

This project has taken me through the complete RAG workflow — from data extraction and processing to indexing, retrieval, and generation. I’ve learned how to split complex JSON data into manageable chunks, create vector embeddings, build a vector database, and construct effective prompts for various tasks. With both a command-line interface and a Gradio web application, I’ve made your innovative tool accessible to users of different technical backgrounds.
Source Code: https://www.twine.net/signin

The AI-Powered Food Recommendation System is an advanced application that combines Retrieval-Augmented Generation (RAG) with Google’s Gemini AI to provide intelligent, personalized food recommendations. The system uses vector similarity search through ChromaDB to find relevant food items and generates natural language recommendations using large language models.
You can use this chatbot for your food recommedations, you can use an advanced search option to search for the food based on your choice. You can also compare the foods which compare the total calories and matching food with your Query. I am attaching a soure code here, if you wanted to get a trail, you can clone it from github and follow the steps in the readme file.
source code: https://www.twine.net/signin