🚀 What It Does

This project uses Fetch.ai agents and REST API endpoints to create a smart, interactive AI system. It includes:

• Real-time PDF ingestion and parsing.
• Storing and embedding document content for similarity-based search.
• Retrieving relevant document chunks based on user queries.
• Generating insightful answers and interactive charts using AI agents.

Although inspired by systems like Pathway, we built our own real-time vector storage and retrieval pipeline without using Pathway directly.

🛠️ How We Built It

We designed a modular system using:

• Fetch.ai UAgents for handling all core functionalities:
  • PDF processing (upload → base64 → parsed JSON → TXT narration).
  • Embedding generation and similarity search.
  • Query interpretation and dynamic response generation.
• REST APIs to interact with agents from the frontend.
• MongoDB Atlas to store uploaded PDFs, parsed data, and processed text.
• A custom-built embedding and retrieval pipeline that mimics real-time document search.
• A Streamlit frontend for file upload, querying, and chart visualization.

⚔️ Challenges We Faced

• Understanding the documentation for Fetch.ai agents and UAgent communication.
• Designing a seamless pipeline without relying on external tools like Pathway.
• Handling real-time data flow and efficient vector search.
• Ensuring consistent communication across REST APIs, agents, and the frontend.

🏆 Accomplishments We're Proud Of

• Built a working RAG (Retrieval-Augmented Generation) pipeline end-to-end.
• Designed our own vector storage and similarity search system from scratch.
• Achieved smooth integration between multiple components — backend agents, MongoDB, and Streamlit frontend.
• Delivered a modular, extensible system ready for real-time document analysis and financial insights.

This project demonstrates how you can build a real-time, agent-driven AI system using Fetch.ai and RESTful design — without relying on external vector databases or frameworks.

Built With

• fetch.ai
• streamlit