Automated Test Case Generator Agent - Capgemini

Capgemini Exceller AgentifAI Buildathon · Team Osaka Vise

---

🎯 Problem Statement

Teams lack sufficient tests, causing regressions. Build an agent that generates unit/integration test cases from code or user stories.

This agent analyzes source code repositories and code snippets to automatically generate, execute, and report on test cases — solving the test-coverage gap without manual effort.

---

✨ Features

Capability	Description
🤖 AI Test Generation	Groq LLM generates pytest-ready test cases from any Python code
🔍 Edge Case Detection	Identifies boundary conditions, null inputs, type errors, and more
📦 Repository Analysis	Scans entire repos or ZIP uploads and extracts all testable functions
▶️ Automatic Execution	Runs generated tests via pytest and captures pass/fail results
📊 Coverage Reporting	Computes and visualizes code coverage metrics
🔄 CI Integration	GitHub Actions pipeline for automated testing on every push
🌐 Interactive Dashboard	Real-time job status, test viewer, and execution logs
🌍 Multi-language Detection	Identifies languages and project structure automatically

---

🏗️ Architecture

Frontend (Next.js + Vercel)
        ↓
FastAPI Backend (Render)
        ↓
BackgroundTasks (async, no Celery/Redis)
        ↓
Groq LLM (llama-3.1-8b-instant)
        ↓
PostgreSQL (Render)

AI Pipeline

Input (Repo / Code Snippet)
        ↓
  Repository Scanner
        ↓
  Function Extraction
        ↓
  Edge Case Detection
        ↓
  LLM Test Generation
        ↓
    Pytest Execution
        ↓
  Coverage Analysis
        ↓
  Dashboard Reporting

---

🛠️ Tech Stack

Backend · FastAPI · SQLAlchemy · PostgreSQL · Groq API · Pytest · Python 3.11 · BackgroundTasks

Frontend · Next.js 16 · TypeScript · TailwindCSS · Lucide React

Deployment · Vercel (Frontend) · Render (Backend + PostgreSQL) · GitHub Actions (CI/CD)

---

📡 API Reference

POST /api/v1/analysis/start

Start analysis on a code snippet or repository.

{
  "source_type": "code_snippet",
  "source_data": "def add(a, b): return a + b"
}

POST /api/v1/analysis/upload

Upload a ZIP archive of your repository for full project analysis.

GET /api/v1/analysis/{job_id}

Poll for results — returns generated tests, execution logs, and coverage metrics.

GET /health

Health check endpoint.

---

🧪 Example Output

Given:

def divide(a, b):
    return a / b

The agent generates:

• ✅ Happy path tests
• ✅ Zero division tests
• ✅ Invalid type tests
• ✅ Boundary tests
• ✅ Negative number tests

---

📂 Project Structure

Automated_Test_Case_Generator_Agent/
├── backend/
│   └── app/
│       ├── agents/
│       │   ├── orchestrator.py
│       │   ├── llm_test_generator.py
│       │   ├── edge_case_finder.py
│       │   ├── repo_scanner.py
│       │   ├── coverage.py
│       │   ├── test_executor.py
│       │   └── code_understanding.py
│       ├── api/
│       ├── core/
│       ├── db/
│       ├── models/
│       ├── schemas/
│       └── main.py
├── frontend/
│   └── src/app/
│       ├── upload/
│       ├── dashboard/
│       ├── tests/
│       └── page.tsx
└── .github/workflows/ci.yml

---

🚀 Local Setup

Backend

git clone https://github.com/Aaronrao989/Automated_Test_Case_Generator_Agent.git
cd Automated_Test_Case_Generator_Agent/backend

python -m venv venv
source venv/bin/activate       # Windows: venv\Scripts\activate

pip install -r requirements.txt
uvicorn app.main:app --reload
# → http://localhost:8000

Frontend

cd ../frontend
npm install
npm run dev
# → http://localhost:3000

Environment Variables

Backend (.env):

DATABASE_URL=postgresql://...
GROQ_API_KEY=your_groq_key
LLM_PROVIDER=groq
GROQ_MODEL=llama-3.1-8b-instant

Frontend (.env.local):

NEXT_PUBLIC_API_URL=https://automated-test-case-generator-agent.onrender.com

---

🔄 CI/CD

GitHub Actions runs on every push:

• Backend test suite (pytest tests/ -v)
• Frontend lint + build verification
• Coverage checks

---

📈 Evaluation Criteria — How We Address Each

Criterion	Implementation
Test relevance & coverage	LLM generates context-aware tests with pytest fixtures and assertions
Correctness	Tests are executed automatically; only passing tests surface in reports
Edge case handling	Dedicated edge_case_finder.py agent identifies boundary and failure conditions
Maintainability	Generated tests follow pytest conventions with clear naming and docstrings
CI integration	GitHub Actions workflow included; API-first design supports any CI tool

---

🛣️ Roadmap

• Advanced coverage heatmaps
• Multi-file dependency analysis
• Java & Go execution support
• Mutation testing
• Real-time streaming updates
• Authentication & team workspaces
• Persistent history dashboard

---

👥 Team — Osaka Vise

Name	Role
Aaron Rao	AI Testing & Validation
Aditi Karn	System Architecture Lead
Aryan Gupta	UI/UX Developer
Nitin Chugh	Backend & API Engineer
Vidushi Srivastava	Presentation Lead

---

📄 License

MIT © Team Osaka Vise — Built for the Capgemini Exceller AgentifAI Buildathon