📈 TrendTrackr: An Agentic AI-Powered Dashboard

A sophisticated, AI-powered dashboard that demonstrates agent-to-agent communication and multi-component pipeline (MCP) design. The application analyzes sentiment from real-time news and Amazon product reviews, and includes interactive research modules on advanced tech topics.

---

🤖 Agentic AI Architecture

This application is built using a modular, agentic design pattern. The main Home.py script acts as an Orchestrator, which calls a series of specialized Tools (agents) to perform specific tasks. This approach, known as a Multi-Component Pipeline (MCP), makes the system more robust, maintainable, and scalable.

Agent-to-Agent (A2A) Workflow

When a user initiates an analysis, the Orchestrator executes a pipeline, with the output of one agent becoming the input for the next. This flow is made visible in the UI using st.status to show which agent is currently active.

Here is the workflow for an Amazon Review Analysis:

graph TD
    A[User Input: ASIN] --> B{🤖 Orchestrator};
    B --> C[🔍 API Agent: tool_fetch_amazon_product];
    C --> D[🧠 Sentiment Agent: tool_analyze_sentiment];
    D --> E[☁️ Visualization Agent: tool_generate_wordcloud];
    E --> F[📊 UI: Display Results];

Loading

1. User Input: The user provides an Amazon ASIN.
2. Orchestrator (Home.py): Receives the request and begins the pipeline.
3. API Agent (tool_fetch_amazon_product): Called by the orchestrator to fetch the product title and reviews from an external API.
4. Sentiment Agent (tool_analyze_sentiment): Receives the reviews from the API Agent and performs sentiment analysis on the text.
5. Visualization Agent (tool_generate_wordcloud): Receives the analyzed data and generates a word cloud image.
6. UI (Home.py): The orchestrator displays the final results (dataframes, charts, images) in the Streamlit interface.

---

✨ Features

• Visible Agentic Workflow: The UI uses an animated status box to show the active agent at each step of the analysis pipeline (e.g., API Agent, Sentiment Agent).
• Multi-Component Pipeline (MCP) Design: The code is refactored into a clear "Orchestrator" and "Tools" structure, demonstrating a robust and scalable software pattern.
• Real-Time News Sentiment Analysis: Fetches and analyzes the latest news headlines for any topic.
• Amazon Product Review Analysis: Fetches and analyzes user reviews for any Amazon product ASIN.
• Interactive Research Modules:
  • RDMA Simulation: An interactive module that visually contrasts traditional TCP/IP data transfers with high-speed RDMA, demonstrating the concepts of kernel bypass and zero-copy for AI training.
  • CDN Paywall Simulation: A simulation of Cloudflare's "Pay Per Crawl" mechanism, allowing users to experiment with different AI crawler bids.
• Advanced Sentiment Analysis: Employs the VADER model to classify text as Positive, Negative, or Neutral.
• Rich Visualizations: Generates interactive charts and word clouds to display results.
• Efficient Caching: Uses Streamlit's caching to optimize performance and avoid redundant API calls.

---

🛠️ Tech Stack

• Framework: Streamlit
• Architecture: Agent-to-Agent (A2A), Multi-Component Pipeline (MCP)
• Data Sources: NewsAPI.org, RapidAPI (for Amazon data)
• Sentiment Analysis: VADER
• Visualizations: Plotly Express, WordCloud, Matplotlib
• Core Libraries: Pandas, Requests

---

🚀 Local Setup and Execution

1. Clone the Repository

   git clone <your-repo-url>
   cd sentiment-analysis-agent

2. Create and Activate a Virtual Environment

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

3. Install Dependencies

   pip install -r requirements.txt

4. Set Up Your API Keys

   • Create a file at .streamlit/secrets.toml.
   • Add your API keys to this file:

     NEWS_API_KEY = "your_news_api_key"
     RAPIDAPI_KEY = "your_rapidapi_key"

5. Run the Application

   streamlit run Home.py

The application will now be running and accessible at http://localhost:8501.

---

👨‍💻 About the Creator

This project was developed by Shivakant Dubey.

Feel free to connect with me and follow my work: 🔗 LinkedIn Profile

---