Mockly: Agentic AI Mock Interviewer ⚡

Mockly is an AI-powered Mock Interview platform designed to help developers ace their technical interviews. By leveraging an agentic workflow, real-time web research, and resume parsing, Mockly generates tailored, context-aware interview questions and provides detailed, structured feedback.

---

🚀 Key Features

• 📄 Intelligent Resume Parsing: Extract skills, experience level, projects, and education directly from uploaded PDF and Word (.docx) documents to personalize the interview experience.
• 🔍 Web-Augmented Research: Performs live web searches (via Tavily) to fetch the latest industry-relevant technical questions and trends.
• ⚡ Smart-Fast LLM Routing: Dynamic model routing optimizes latency and costs:
  • Fast Model: Summarization and search query generation.
  • Smart Model: Complex question generation and detailed answer evaluations.
• 💾 Query Caching: In-memory caching for web research results to save Tavily API credits and improve API response times.
• 📝 Interactive Chat Sandbox: Responsive, sticky chat interface with isolated scroll areas for questions and evaluations.
• 🔐 NextAuth Integration: Out-of-the-box support for Google & GitHub social logins.
• 📊 Database Integration: Prisma Client connected to a PostgreSQL database for session management.

---

🛠️ Tech Stack

Frontend

• Framework: Next.js (App Router, React 19)
• Styling: Tailwind CSS, Shadcn UI
• Icons: Phosphor Icons
• ORM: Prisma with PostgreSQL

Backend

• Framework: FastAPI (Python 3.10+)
• Agentic Orchestration: LangGraph & LangChain
• Models: NVIDIA NIMs (Llama 3.1 8B/70B Instruct)
• Search Engine: Tavily Search API
• Text Extraction: PyMuPDF (fitz) & python-docx
• Parsing Resiliency: json-repair

---

📂 Project Structure

Mockly/
├── backend/                # FastAPI backend & LangGraph agents
│   ├── app/
│   │   ├── agents/         # Graph definitions, nodes, and prompts
│   │   ├── api/            # API endpoints (resume parsing, interview)
│   │   ├── tools/          # Web search cache, resume text extraction
│   │   └── utils/          # State serialization helpers
│   ├── main.py             # FastAPI entrypoint
│   └── requirements.txt
├── frontend/               # Next.js web application
│   ├── app/                # App router pages (dashboard, sandbox)
│   ├── components/         # React components (interview panel, inputs)
│   ├── prisma/             # Schema definitions and migrations
│   └── package.json
└── README.md               # Project documentation (this file)

---

⚙️ Getting Started

Prerequisites

• Node.js 18+ & pnpm (or npm/yarn)
• Python 3.10+
• PostgreSQL database (or Prisma Postgres/Accelerate connection string)

---

1. Backend Setup

1. Navigate to the backend directory:

   cd backend

2. Create and activate a virtual environment:

   python3 -m venv .venv
   source .venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Configure environment variables in a .env file inside the backend/ directory:

   NVIDIA_BASE_URL="https://integrate.api.nvidia.com/v1"
   NVIDIA_API_KEY="your-nvidia-api-key"
   NVIDIA_FAST_MODEL="meta/llama-3.1-8b-instruct"
   NVIDIA_SMART_MODEL="meta/llama-3.1-70b-instruct"
   TAVILY_API_KEY="your-tavily-api-key"
   CORS_ORIGINS="http://localhost:3000"

5. Start the FastAPI server:

   fastapi dev main.py
   # or
   uvicorn main:app --reload --port 8000

---

2. Frontend Setup

1. Navigate to the frontend directory:

   cd frontend

2. Install packages:

   pnpm install
   # or npm install

3. Configure environment variables in a .env file inside the frontend/ directory:

   DATABASE_URL="postgresql://username:password@localhost:5432/mockly"
   AUTH_SECRET="your-generated-secret"
   
   # Social Login OAuth Credentials
   AUTH_GOOGLE_ID="google-client-id"
   AUTH_GOOGLE_SECRET="google-client-secret"
   AUTH_GITHUB_ID="github-client-id"
   AUTH_GITHUB_SECRET="github-client-secret"
   
   NEXT_PUBLIC_API_URL="http://localhost:8000"

4. Run Prisma migrations & database setup:

   pnpx prisma db push
   # optionally run the studio to view your tables:
   pnpx prisma studio

5. Start the development server:

   pnpm dev
   # or npm run dev

Open http://localhost:3000 to start preparing with Mockly!