📈 Financial News Intelligence Ai Agent System

Welcome This is an intelligent, multi-agent financial news intelligence platform designed to help traders and investors cut through the noise. It automatically ingests, analyzes, and extracts actionable market insights from real-time news feeds.

Built with modern AI tools, this project demonstrates my ability to integrate Large Language Models (LLMs), agentic workflows, and fast APIs into a cohesive, user-friendly product.

🚀 Why I Built This

In the fast-paced financial market, information overload is a real problem. I built Tradl AI to track relevant financial news, eliminate duplicate stories, and instantly predict which stocks or sectors might be impacted—and by how much.

✨ Key Features

• Real-Time Data Ingestion: Automatically pulls and parses financial news from RSS feeds.
• Multi-Agent AI Workflow (LangGraph): Architected a robust pipeline using LangGraph's StateGraph to coordinate specialized AI agents. Key technical highlights:
  • Modular Pipeline Design: Decoupled complex LLM tasks into independent, maintainable nodes (Deduplication, Extraction, Storage), moving beyond fragile sequential prompt-chaining.
  • State-Driven Cost Optimization: Implemented a shared typed AgentState that enables conditional execution. For example, bypassing expensive LLM extraction calls if the Deduplication node flags an article as a duplicate, directly improving latency and reducing API costs.
  • Deterministic Execution of Non-Deterministic AI: Designed a reliable Directed Acyclic Graph (DAG) that perfectly orchestrates non-deterministic LLM analysis with deterministic data routing and Vector DB (ChromaDB) storage.
  • (Agents inside the graph: Deduplication Agent, Extraction Agent, Query Agent)
• Semantic Search: Utilizes ChromaDB and Sentence-Transformers for fast, vector-based information retrieval.
• RESTful API: A robust FastAPI backend that handles ingestion, stats, and search queries (/api/ingest, /api/query).
• Interactive UI: A clean, responsive dashboard built with Streamlit to visualize market intelligence.

🛠️ Tech Stack

• Backend/API: Python, FastAPI, Uvicorn, Pydantic
• AI & NLP: LangChain, LangGraph, Mistral AI, SpaCy, Sentence-Transformers
• Vector Database: ChromaDB
• Frontend: Streamlit
• Data Gathering: Requests, BeautifulSoup4, Feedparser

💻 Quick Start

1. Clone & Install

git clone <your-repo-url>
cd AI-Powered-Financial-News-Intelligence-System-1
pip install -r requirements.txt

2. Set Environment Variables

Create a .env file in the root directory and add your required LLM API keys (e.g., Mistral API key):

MISTRAL_API_KEY=your_api_key_here

3. Run the Backend (FastAPI)

Start the backend server:

python app/main.py

The API will be available at http://localhost:8000.

4. Run the Frontend (Streamlit)

In a new terminal, start the UI:

streamlit run app/ui.py

You can now interact with the dashboard to view stats, ingest news, and run context-aware queries!

5. (Optional) Run the Demo Script

To see it work directly from the CLI without the UI, run the demo script:

python demo.py

📈 Future Roadmap

• Integration with real-time stock price APIs (e.g., Yahoo Finance/Alpaca) to track actual impact vs. predicted impact.
• User authentication and personalized stock watchlists.
• Automated deployment pipeline (CI/CD) to AWS/GCP.

🤝 Let's Connect!

I am actively looking for software engineering, AI/ML, and roles. If you're a recruiter, hiring manager, or fellow engineer who finds this project interesting, I'd love to chat!