About FactCheck AI

An experimental AI-powered fact-checking tool designed for educational and research purposes.

Important Disclaimer

This tool is for educational and research purposes only. Results are provided "as is" without any warranties. The AI may make errors, miss context, provide incomplete analysis, return unexpected results, or misunderstand sources. Users should independently verify all information through authoritative sources and consult qualified professionals for important decisions. This tool is not intended to replace professional fact-checking, legal advice, medical advice, or other expert consultation. Do not rely on this tool for medical, legal, financial, or safety-critical decisions. Use at your own risk.

Project Overview

FactCheck AI was created as a learning project to explore AI-powered fact-checking using large language models and real-time web search. The tool combines Hugging Face's Llama 3.1 8B model with retrieval-augmented generation (RAG), powered by LangChain, web search APIs, and vector database technology.

Note: This project is fully open source and experimental. You can find the code on GitHub via the link in the footer.

How It Works

AI Content Moderation

Claims are first screened by AI for inappropriate or unsafe content

Hybrid Search

Combines web search with vector database for similar claims

AI Fact-Checking

The Llama 3.1 8B model analyzes claims against sources and detects contradictions

Result Generation

Verdict, confidence score, reasoning, and results are stored for future reference

Technical Stack

AI ModelLlama 3.1 8B Instruct
FrontendNext.js 14 + Tailwind
BackendFastAPI + LangChain
Vector DatabaseNeon PostgreSQL + pgvector
EmbeddingsSentence Transformers
Search APISerper.dev
Note: Vector search is disabled in this demo due to memory constraints.

Learning Goals

  • AI-powered fact-checking systems
  • Vector database integration and semantic search
  • Hybrid search (web + vector) and contradiction detection
  • Large language model integration with RAG
  • Full-stack AI applications with async processing