TestGenie

Generate test cases instantly 🧞‍♂️

🚀 Try it live here!

Automated, AI-powered test case generation for QA engineers and developers.

Translates feature descriptions and acceptance criteria into structured test cases using Gemini — categorizing them into Functional, Security, Boundary, and UX tests, and providing one-click exports to JSON, CSV, and Markdown.

---

The problem this solves

Writing comprehensive test cases manually is incredibly time-consuming and prone to human error — developers and QA engineers often miss critical edge cases, security vulnerabilities, or negative scenarios. TestGenie uses AI to instantly generate a robust suite of test cases (including edge cases and boundary conditions) so you can focus on building and automating tests rather than writing boilerplate documentation.

---

Architecture

Feature Description & Acceptance Criteria
      ↓
Subject App — Node.js Express server / Vercel Serverless
(validates constraints and parses configuration)
      ↓
AI Generator — Gemini 3.1 Flash Lite
(generates categorized test steps and expected results)
      ↓
Structured Test Cases (JSON)
      ↓
Interactive UI + CSV/Markdown/JSON Exports

---

Why Gemini?

Using Gemini 3.1 Flash Lite is a deliberate architectural decision. Test case generation requires a model with strong reasoning, instruction following, and the ability to strictly output structured JSON arrays. Flash Lite provides an incredibly fast time-to-first-token while maintaining high accuracy for context-heavy QA tasks, making the UI feel snappy and responsive.

---

What it generates

Each test case is categorized across four dimensions:

Category	What it checks
Functional	Does the feature work exactly as described in the happy path?
Security	Are there vulnerabilities (e.g. SQL injection, unauthorized access)?
Boundary	Does the system handle extreme limits, long strings, or edge inputs gracefully?
UX	Does the system provide clear validation messages and a good user experience?

Additionally, TestGenie supports single-case regeneration — if a specific test case isn't quite right, you can ask the AI to rewrite just that one case with more detail or a different focus.

---

Models

Role	Model	Provider	Why
Generator	gemini-3.1-flash-lite	Google AI	Fast, strong reasoning, structured JSON output

---

Sample output

When prompted to generate test cases for a "Login" feature with the "Edge Cases" toggle enabled, TestGenie consistently generates standard happy paths (e.g., successful login) but heavily focuses on boundary conditions: excessively long passwords, SQL injection attempts, empty credentials, and rate-limiting scenarios. It correctly formats the steps, expected results, and categorizes each case appropriately.

---

Stack

• Language: TypeScript / Node.js
• Frontend: Vanilla HTML5, CSS3, JavaScript (Single Page Application)
• AI Model: Gemini 3.1 Flash Lite via @google/genai
• Deployment: Vercel (Zero-Config API Routes)

---

Project structure

TestGenie/
├── api/
│   └── generate.ts            # Vercel Serverless entry point
├── public/
│   ├── index.html             # Vanilla HTML/JS frontend
│   └── loader.lottie          # Animation assets
├── src/
│   └── server.ts              # Express API & Gemini integration
├── .env
├── .gitignore
├── package.json
├── tsconfig.json
└── README.md

---

Setup

Prerequisites

• Node.js 18+
• A free Google Gemini API Key — aistudio.google.com

Installation

git clone https://github.com/puja-ui/TestGenie.git
cd TestGenie
npm install

Environment variables

Create .env and fill in your key:

GEMINI_API_KEY=your_gemini_key_here
PORT=3000

Run locally

npm start

The application will start on http://localhost:3000.

---

Deploying to Vercel

The project is architected for zero-config deployment on Vercel.

1. Push your code to GitHub.
2. Create a New Project on Vercel and import the repo.
3. Select Other as the Framework Preset.
4. Add GEMINI_API_KEY to the Environment Variables.
5. Deploy. (Vercel automatically serves the public/ directory statically and maps api/generate.ts as a serverless function).

---

Key design decisions

Why Vanilla JS / HTML instead of React? To keep the project incredibly lightweight and fast. By avoiding a heavy frontend framework, the application loads instantly and remains simple to hack on or extend without complex build steps for the client side.

Why Vercel Zero-Config? The project was restructured to decouple the Express app routing from the static file serving. By moving the static assets to the root public/ folder and the API route to api/, Vercel naturally serves the frontend on its Edge CDN while treating the backend as an isolated serverless function, completely eliminating bundler issues with the AI SDK.

---

What's next

• Implement integration testing with Playwright for the UI and pytest for API testing
• Add Jira/Linear integration to directly push generated test cases to ticket tracking
• Allow custom system prompts for different testing methodologies (e.g., BDD style "Given/When/Then")

---

Example Report