Skip to content

ASHUTOSH-A-49/Shard

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

86 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

💎 Shard | AI-Driven Invoice Intelligence

Shard is a high-performance, "Database-First" invoice processing engine that transforms unstructured documents into structured, actionable financial data. Built for speed and reliability, Shard solves the "Trust Gap" in AI by combining LLM-based semantic extraction with an Intelligent Review Queue and Computer Vision pre-validation.

DEPLOY LINK - https://shard-9qyi.onrender.com/

🚀 Key Features

  • Semantic AI Extraction: Powered by Groq/Grok, Shard understands the context of financial documents, making it layout-agnostic.
  • Image Quality Guardrail: Uses Laplacian variance and histogram analysis to detect blur or low-contrast uploads before they hit the AI.
  • Human-in-the-Loop (HITL): A dedicated Review Queue for low-confidence extractions ensures 100% data integrity.
  • Database-First Architecture: Strict synchronization between React/Zustand and MongoDB, moving away from stale local caching.
  • Real-Time Analytics: Visualize spend trends, vendor distributions, and AI accuracy metrics instantly.

🛠️ Tech Stack

  • Frontend: React, Vite, Tailwind CSS, Framer Motion, Zustand
  • Backend: Flask (Python), OpenCV (Image Processing)
  • AI/LLM: Groq LPU™ Inference / Grok
  • Database: MongoDB
  • Cloud Readiness: Architected via Kiro for AWS (S3, DocumentDB, Cognito)

🏗️ Architecture

  1. Ingestion: User uploads PDF/Image.
  2. Validation: Backend performs blur/contrast checks.
  3. Extraction: AI identifies vendor, line items, taxes, and totals.
  4. Persistence: Data is saved to MongoDB with a unique userId identity token.
  5. Review: Low-confidence scores (< 85%) are quarantined for manual approval.

🏁 Getting Started

Prerequisites

  • Node.js (v18+)
  • Python 3.9+
  • MongoDB instance

Installation

  1. Clone the Repo
    git clone [https://github.com/ASHUTOSH-A-49/Shard.git](https://github.com/ASHUTOSH-A-49/Shard.git)
    cd Shard
  2. Backend Setup
    cd backend
    pip install -r requirements.txt
    Create a .env file with your MONGO_URI and GROQ_API_KEY
    python app.py

`` 3. Frontend Setup

cd frontend
npm install
npm run dev

Identity & Security A great README.md is often the difference between a project that gets ignored and one that wins a hackathon. Since you are positioning this as an innovative, production-ready AI tool, the README needs to be clean, technical, and visually organized.

Create a file named README.md in your root directory and paste the following:

Markdown

💎 Shard | AI-Driven Invoice Intelligence

Shard is a high-performance, "Database-First" invoice processing engine that transforms unstructured documents into structured, actionable financial data. Built for speed and reliability, Shard solves the "Trust Gap" in AI by combining LLM-based semantic extraction with an Intelligent Review Queue and Computer Vision pre-validation.

🚀 Key Features

  • Semantic AI Extraction: Powered by Groq/Grok, Shard understands the context of financial documents, making it layout-agnostic.
  • Image Quality Guardrail: Uses Laplacian variance and histogram analysis to detect blur or low-contrast uploads before they hit the AI.
  • Human-in-the-Loop (HITL): A dedicated Review Queue for low-confidence extractions ensures 100% data integrity.
  • Database-First Architecture: Strict synchronization between React/Zustand and MongoDB, moving away from stale local caching.
  • Real-Time Analytics: Visualize spend trends, vendor distributions, and AI accuracy metrics instantly.

🛠️ Tech Stack

  • Frontend: React, Vite, Tailwind CSS, Framer Motion, Zustand
  • Backend: Flask (Python), OpenCV (Image Processing)
  • AI/LLM: Groq LPU™ Inference / Grok
  • Database: MongoDB
  • Cloud Readiness: Architected via Kiro for AWS (S3, DocumentDB, Cognito)

🏗️ Architecture

  1. Ingestion: User uploads PDF/Image.
  2. Validation: Backend performs blur/contrast checks.
  3. Extraction: AI identifies vendor, line items, taxes, and totals.
  4. Persistence: Data is saved to MongoDB with a unique userId identity token.
  5. Review: Low-confidence scores (< 85%) are quarantined for manual approval.

🏁 Getting Started

Prerequisites

  • Node.js (v18+)
  • Python 3.9+
  • MongoDB instance

Installation

  1. Clone the Repo
    git clone [https://github.com/ASHUTOSH-A-49/Shard.git](https://github.com/ASHUTOSH-A-49/Shard.git)
    cd Shard

Backend Setup

Bash

cd backend pip install -r requirements.txt

Create a .env file with your MONGO_URI and GROQ_API_KEY

python app.py Frontend Setup

Bash

cd frontend npm install npm run dev 🛡️ Identity & Security Shard uses a standardized JWT Identity Payload stringified into the Authorization header. This ensures that every API request is partitioned by user identity, preventing data leaks in a multi-tenant environment.

Future Roadmap

  • Direct AWS S3 integration for document archiving.
  • Adding a Chatbot to ask questions directly from dashboard and analytics insights
  • Export to QuickBooks/Xero API.
  • Multi-currency conversion via real-time exchange rate APIs.

Built with ❤️ by team JETT-2-HOLIDAY for Hackxios contributors and team members

About

Shard: Fast, sharp, and cuts through data.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors