ASTREA
🚀 Problem Statement

NASA’s space biology research produces vast amounts of unstructured and semi-structured data—research papers, experimental reports, tables, and images. However:

Relevant insights are scattered across thousands of documents
Cross-paper relationships are difficult to identify
Traditional keyword search fails to capture semantic meaning

As a result, valuable connections remain hidden, slowing scientific discovery.

💡 Solution Overview

ASTREA introduces a hybrid retrieval and reasoning system that:

Scrapes and ingests bioscience documents
Extracts structured knowledge (entities & relationships)
Stores information in both:
- a Vector Database (semantic similarity)
- a Knowledge Graph (explicit relationships)
Answers user queries by reasoning over both representations

This allows ASTREA to go beyond simple document retrieval and generate context-aware, insight-driven responses.
✨ Key Features

🔍 Semantic Search across 600+ NASA bioscience studies
🧠 Hybrid RAG Architecture (Vector + Knowledge Graph)
🕸️ Entity & Relationship Extraction using LLMs
📊 Neo4j Knowledge Graph for relational exploration
📦 Pinecone Vector Store for scalable similarity search
🧩 Intelligent Chunking for high-quality retrieval
📄 PDF + Web Content Support
🚀 Multiple Interfaces: REST API, CLI, and programmatic usage

🛠️ Tech Stack

Backend: Node.js, Express.js
Web Scraping: Axios, Cheerio
LLMs & Embeddings: OpenAI / Gemini (configurable)
Vector Database: Pinecone
Knowledge Graph: Neo4j
Document Processing: Custom chunking pipeline
APIs: RESTful endpoints