Research Intelligence Platform (Enterprise Edition)

An enterprise-grade research intelligence platform that transforms unstructured documents into actionable insights. Powered by Hybrid Semantic Search, it combines vector embeddings (Qdrant) with keyword search (Elasticsearch) to deliver precise results from thousands of PDFs, reports, and financial statements.

---

🚀 Key Features

🧠 Intelligent Search

• Hybrid Search Engine: Fuses vector proximity (semantic meaning) with keyword matching (exact terms) using Reciprocal Rank Fusion (RRF).
• Deep Semantic Understanding: Uses sentence-transformers/all-MiniLM-L6-v2 for state-of-the-art embedding generation.
• Smart Filtering: Drill down by company entity, document type, or date range.

💼 Corporate UI / UX

• Professional Design System: A "Deep Navy & Vibrant Blue" aesthetic tailored for financial and research environments.
• Glassmorphism Interface: Modern, sticky headers and tactile interactive elements.
• Data-Grid Repository: Enterprise-class document management view with status badges and metadata analysis.

⚡ Robust Architecture

• Asynchronous Processing: Celery + Redis pipeline for non-blocking document ingestion.
• Scalable Storage: MinIO (S3-compatible) for object storage and PostgreSQL for structured metadata.
• Fault Tolerance: Comprehensive error handling and retry mechanisms.

---

🏗️ Architecture

graph TD
    User[Web Client] -->|Next.js App| Frontend
    Frontend -->|REST API| API[FastAPI Gateway]
    
    subgraph Data Processing
        API -->|Task| Queue[Redis Task Queue]
        Queue -->|Consume| Worker[Celery Worker]
        Worker -->|Extract Text| Python[PDF Parser]
        Worker -->|Generate| Model[Embedding Model]
    end
    
    subgraph Storage
        Worker -->|Store Vectors| VectorDB[(Qdrant)]
        Worker -->|Index Text| SearchEngine[(Elasticsearch)]
        Worker -->|Save File| ObjectStore[(MinIO S3)]
        Worker -->|Update Meta| RelationalDB[(PostgreSQL)]
    end
    
    API -->|Query| VectorDB
    API -->|Query| SearchEngine
    API -->|Read| RelationalDB

Loading

---

🛠️ Quick Start

Prerequisites

• Docker & Docker Compose
• Python 3.11+
• Node.js 18+

1. Launch Infrastructure

Start the database, vector store, and object storage services:

docker-compose up -d

Note: Wait ~30 seconds for all services to become healthy.

2. Backend Setup

cd backend
python -m venv venv
# Windows:
venv\Scripts\activate
# Mac/Linux:
# source venv/bin/activate

pip install -r requirements.txt
python main.py

API available at http://localhost:8000

3. Frontend Setup

cd frontend
npm install
npm run dev

UI available at http://localhost:3000

---

🧪 Demo Mode (No Docker Required)

Want to try the UI without setting up the full database stack? You can run the backend in Demo Mode, which uses in-memory storage.

# In backend directory
python demo_main.py

Note: Uploaded documents will not persist after server restart in demo mode.

---

📚 API Documentation

Once the backend is running, access the interactive Swagger UI:

• Docs: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

Core Endpoints

• POST /api/upload: Asynchronous PDF ingestion.
• POST /api/search: Hybrid search query with filters.
• GET /api/documents: List managed documents.

---

🔮 Roadmap

• Phase 1: MVP & Core Search (Completed)
• Phase 1.5: Corporate UI Redesign (Completed)
• Phase 2: Advanced Analytics (Time-series data, Entity linking)
• Phase 3: Multi-User Collaboration (Shared workspaces, annotations)

---

📄 License

MIT © 2026 Research Platform Team