NewsBiz

Understand the market in a single glance — AI-powered news insights with MongoDB Vector Search and Google Gemini.

Comment 2

Inspiration

Every day, the internet is flooded with financial news — but making sense of it is overwhelming. A simple Search throws you into a maze of links, headlines, and bloated websites filled with ads and distractions. You don’t get insight — just information.

I wanted to fix that. My goal was to create a tool that visualizes the mood of the market, lets users ask natural-language questions, and recommends high-impact news stories — instantly and beautifully.

What it does

NewsBiz is a smart, AI-powered financial news tool that gives you:

  • A visual mood graph of today's market based on the emotional tone of articles

  • A clean, summarized daily snapshot of the financial world

  • An intelligent search bar that answers your questions using Gemini + MongoDB Vector Search

  • A stock relevance bar showing most-mentioned tickers in the news

  • Each article has:

    • An impact score
    • A gauge of emotional tone (e.g. optimistic, critical, neutral)
    • A “read full article” link to the original source
    • A list of recommended related articles, based on vector similarity

Everything is designed to help you understand the market at a glance.

How i built it

The project has two folders: backend and frontend.

Backend

Built using Node.js + Express, the backend:

  • Fetches and parses financial news from Google RSS feeds

  • Summarizes content using Google Gemini

  • Generates text embeddings with Gemini’s Embedding API

  • Analyzes emotion and impact scores

  • Stores all processed data in MongoDB Atlas

  • Uses MongoDB Vector Search for:

    • Article recommendations
    • Semantic search queries

  • Implements RAG (Retrieval-Augmented Generation) to answer user questions using Gemini + top documents

Frontend

Built with React.js, the frontend:

  • Displays a Tree Map chart visualizing emotional breakdown of market news

  • Highlights the top stocks in today’s media cycle

  • Allows users to search in natural language, including with typos

  • Shows article summaries, emotional gauges, and impact scores

  • Provides recommended articles using vector similarity from MongoDB

  • Lets you download the mood graph as an image for sharing/reporting

Challenges I ran into

  • Fine-tuning emotion scores so they reflect not just mentions but urgency and tone

  • Handling question answering with RAG — ensuring the retrieval and summarization process stayed fast and relevant

  • Aligning MongoDB vector embeddings with Gemini’s — matching formats and types

  • Building everything with tight API limits on free tiers!

Accomplishments that i am proud of

  • Built a fully functional AI news assistant with zero fluff

  • Pulled off RAG-style semantic search using MongoDB and Google Gemini together

  • Created a sleek UI that even non-technical users can grasp in seconds

  • Delivered an experience that’s intuitive and simple .

What i learned

  • How to use MongoDB Vector Search to drive real-time recommendations and natural-language understanding

  • How to integrate Google Gemini’s text + embed models

  • How to compute and visualize emotions and impact metrics in a useful way

  • How to structure a fast, scalable pipeline from RSS → RAG end to end

  • How to debug deployment problems across multiple cloud platforms under pressure!

What's next for NewsViz

  • Add support for more sources (e.g. Reddit, Twitter, substack)

  • Let users follow specific companies or tickers

  • Add sentiment change tracking over time for recurring topics

  • Add alerts for stories with unusually high impact scores

Built With

Share this project:

Updates