Team Number: 41

CheckFirstLeh detecting FAKE NEWS!

Inspiration

With misinformation becoming more sophisticated and difficult to detect, it’s crucial to develop convenient and reliable fact-checking tools

What it does

CheckFirstLeh, uses Agentic Retrieval-Augmented Generation (RAG) approach, enhancing fact-checking by combining real-time information retrieval with generative AI to provide reliable, well-sourced answers. To see that it actually works, a Langchain trace is attached, where the agent will first use a web search tool, before using the querytool, which scrapes and embeds the content, before retrieving the most similar content from the vector database to the query, allowing the LLM to make an informed judgement with additonal context. It is able to verify fake news and fake images!

How we built it

We leveraged OpenAI LLMs, web scraping, and vector databases to create a seamless fact-checking experience. The system retrieves relevant documents, processes them, and provides transparent fact-checked responses.

Challenges we ran into

Latency issues – Optimizing retrieval and generation speed for real-time responses. Troubleshooting crawling of urls

Accomplishments that we're proud of

Successfully built an AI Agent, integrating LLMs for intelligent fact-checking. Developed a functional Telegram bot with real-time retrieval.

What we learned

LLM can be improved by providing real time tools, and that agentic ai is very powerful to automate tasks.

How to debug LLMs effieciently

What's next for Team Number: 41

We plan to include support for Whatsapp, while hosting it on live servers. We also want to include a section, where fake news are collated and can be reported to the police