An autonomous, AI-driven engine and premium frontend interface that generates elite, ATS-optimized LaTeX resumes.
CVOS doesn't just format text; it acts as a Senior Technical Recruiter. It uses a dual-LLM agentic loop, Retrieval-Augmented Generation (RAG), and strict hallucination guards to engineer perfect resume bullets, infer skills from GitHub, and compile a production-ready PDF.
- 🤖 Agentic Chat Assistant: A proactive, dual-LLM architecture (Drafter + Critic). It actively interviews you, extracts measurable metrics, filters fluff, and rewrites your bullets in real-time.
- ⚡ Concurrent GitHub Integration: Enter your GitHub username, and CVOS will concurrently fetch your top public repositories, analyze the code/languages, and generate ATS-friendly project bullets in seconds.
- 🔄 Live Sync PDF Generation: A beautiful, Framer Motion-powered review dashboard. Edit your resume data and instantly re-compile a production-ready LaTeX PDF.
- 🧠 Recursive RAG & Evaluation: Uses ChromaDB to feed the AI "Golden Resume" examples. A strict Python evaluator grades the LLM's output and forces it to rewrite until the resume achieves a 95+ score.
- ✨ Premium UI: Buttery-smooth physics, spring animations, and highly responsive components built with TailwindCSS and Framer Motion.
- React (Vite / Next.js)
- TailwindCSS & Shadcn UI
- Framer Motion (Advanced Spring Physics)
- Lucide React (Iconography)
- Python 3.10+ & FastAPI
- Groq API (Llama 3.3 70B & Llama 3.1 8B)
- ChromaDB (Vector Database for RAG)
pdflatex(LaTeX to PDF Compilation)asyncio(Concurrent processing)
- Node.js (v18+)
- Python (v3.10+)
- LaTeX Distribution: You must have
pdflatexinstalled on your system to compile the PDFs.- Mac:
brew install mactexor download MacTeX. - Linux:
sudo apt-get install texlive-full - Windows: Download and install MiKTeX.
- Mac:
Navigate to the root directory and set up your Python environment:
# Create a virtual environment
python -m venv venv
source venv/bin/activate # On Windows use: venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Set up environment variables
# Create a .env file in the root directory and add:
GROQ_API_KEY=your_groq_api_key_here
# Run the FastAPI server
cd core-backend
python app.pyThe backend will run on http://localhost:8000
Open a new terminal and navigate to the frontend directory:
cd frontend
# Install dependencies
npm install
# Start the development server
npm run devThe frontend will run on http://localhost:5173 (or 3000)
- Data Intake: The user fills out basic professional information and imports relevant GitHub repositories.
- Agentic Refinement: The user chats with the CVOS Assistant to enhance bullet points. A Critic LLM oversees the process to ensure the Drafter LLM accurately applies updates directly to the global JSON state.
- The Enhancement Loop: When the user clicks Generate, the backend LLM rewrites the data using high-performing RAG (Retrieval-Augmented Generation) examples to ensure maximum impact.
- The Evaluator: An NLP-based scoring system meticulously grades the output. If the score falls below 95, it feeds the errors back to the LLM to rewrite and try again (Convergence Loop).
- Compilation: The finalized, high-scoring JSON is injected into a professional LaTeX template and compiled into a beautifully formatted PDF.
This project is open-source and available under the MIT License.