🚀 AI-Powered Resume Analyzer

An intelligent application that leverages Generative AI to analyze and score resumes against job descriptions in real-time. This project uses a FastAPI backend hosted on Hugging Face Spaces and a reactive Next.js frontend deployed on Vercel.

✨ Live Demo

• Frontend (Vercel): https://nebula-resume.vercel.app/

• Backend API Docs (Hugging Face): https://letapreemas-nebula-resume-api.hf.space/docs

🌟 Key Features

• Concurrent Batch Processing: Upload and analyze multiple resumes against a job description simultaneously.

• Real-Time Progress Streaming: See the status of each analysis—from text extraction to final scoring—live in the UI.

• Multi-Layered AI Analysis:

  1. Hard Keyword Matching: Fuzzy search for technical keywords.

  2. Semantic Similarity Scoring: Uses Google's Gemini embeddings for a strict, requirement-focused semantic match.

  3. LLM-Powered Verdict: A final verdict and actionable suggestions are generated by a Gemini language model.

• Persistent Job Descriptions: Save, manage, and reuse job descriptions via an integrated database.

• Decoupled Architecture: A robust FastAPI backend and a sleek, modern Next.js frontend for a scalable and maintainable solution.

🏗️ Architecture

The application is built with a decoupled frontend and backend architecture:

+--------------------------+          +--------------------------------------+
|   Next.js Frontend       |          |      FastAPI Backend on HF Spaces    |
|   (Hosted on Vercel)     |          |                                      |
|                          |          |   +--------------------------------+ |
| - File Uploads           |  HTTP    |   |    API Endpoints (/analyze)    | |
| - Real-time SSE Display  | <------> |   +--------------------------------+ |
| - Job Management UI      |          |   +--------------------------------+ |
|                          |          |   | LangGraph Workflow Engine      | |
+--------------------------+          |   |  - Text Extraction             | |
                                      |   |  - Normalization               | |
                                      |   |  - AI Comparisons (Gemini)     | |
                                      |   +--------------------------------+ |
                                      +--------------------------------------+

🛠️ Tech Stack

Backend (Hugging Face Spaces):

• Framework: FastAPI

• Orchestration: LangGraph

• AI / LLMs: LangChain with Google Gemini (LLM & Embeddings)

• Text Processing: NLTK, thefuzz

• Database: SQLite

Frontend (Vercel):

• Framework: Next.js

• Language: TypeScript

• UI: React, Shadcn/ui, Tailwind CSS

• Real-time: @microsoft/fetch-event-source for POST request streaming (SSE)

🚀 Getting Started (Local Setup)

Follow these instructions to set up and run the project locally.

Prerequisites

• Node.js (v18.0 or later)

• Python (v3.9 or later)

• A Google AI API Key (for Gemini).

1. Backend Setup (FastAPI)

You can set up the backend using a traditional virtual environment + pip, or using Poetry (recommended if you use Poetry in the project).

Option A — Virtualenv + pip

# 1. Clone the repository
git clone https://github.com/urffsamhunt/nebula.git
cd nebula/backend

# 2. Create and activate a virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# 3. Install Python dependencies
pip install -r requirements.txt

# 4. Set up environment variables
# Create a .env file in the `backend` directory and add your API key:
echo "GOOGLE_API_KEY='your_google_api_key_here'" > .env

# 5. Download NLTK data
python -c "import nltk; nltk.download('punkt'); nltk.download('stopwords'); nltk.download('wordnet')"

# 6. Run the backend server
uvicorn main:app --reload --port 8000

Option B — Poetry (recommended)

# 1. Clone the repository
git clone https://github.com/urffsamhunt/nebula.git
cd nebula/backend

# 2. Install Poetry if you don't have it (see https://python-poetry.org/docs/)
# Example (Linux/macOS):
# curl -sSL https://install.python-poetry.org | python3 -

# 3. Install dependencies (creates a virtualenv automatically)
poetry install

# 4. Activate the Poetry shell (optional)
poetry shell

# 5. Set up environment variables
# Create a .env file in the `backend` directory and add your API key:
# echo "GOOGLE_API_KEY='your_google_api_key_here'" > .env

# 6. Download NLTK data
poetry run python -c "import nltk; nltk.download('punkt'); nltk.download('stopwords'); nltk.download('wordnet')"

# 7. Run the backend server
poetry run uvicorn main:app --reload --port 8000

The API server will be running at https://${API_BASE}.

2. Frontend Setup (Next.js)

# 1. From the repository root (this project contains the Next.js app at the root)
npm install

# 2. Set up environment variables (Next.js expects .env.local at the project root)
# Point the client to the backend API (use NEXT_PUBLIC_ prefix so it's exposed to the browser):
echo "NEXT_PUBLIC_API_BASE=https://${API_BASE}" > .env.local

# 3. Run the frontend development server
npm run dev