🐛 AI Auto Bug Tester

An intelligent, end-to-end bug reproduction assistant that reads bug reports and automatically executes UI test steps using Playwright—enhanced with AI-driven fallback selectors and DOM understanding.

---

🚀 Features

• ✅ Natural language bug report parsing via Phi-3 mini instruct (LLM)
• ✅ Converts bug descriptions into structured JSON test steps
• ✅ Executes steps with Playwright
• ✅ Uses AI fallback for dynamic DOM-based selector recovery
• ✅ Streamlit UI for easy interaction
• ✅ Generates detailed markdown reports with screenshots
• ✅ Browser runs in headed mode for live inspection

---

🧠 Powered by AI

• Uses Phi-3 Mini (Locally running via LM Studio) for:
  • Parsing bug descriptions
  • Suggesting fallback selectors dynamically by analyzing live DOM
  • Recovering from broken test steps

---

🔧 Tech Stack

Layer	Technology
UI	Streamlit
Backend	FastAPI
Test Execution	Playwright (Python)
AI Integration	Phi-3 via LM Studio (GGUF LLM)
Markdown + Screenshot	Python Imaging, Base64

---

🗂️ Project Structure

ai-auto-bug-tester/
├── backend/
│   ├── main.py              # FastAPI entry point
│   ├── executor.py          # Core test executor with Playwright
│   ├── gpt_parser.py        # LLM integration & bug parsing
│   └── outputs/             # Screenshots + Reports
├── frontend/
│   └── app.py               # Streamlit interface

---

⚙️ Setup Instructions

1. Clone the repository
   git clone https://github.com/guptasoumya26/ai-auto-bug-tester.git

2. Install dependencies
   pip install -r requirements.txt

3. Install Playwright browsers
   playwright install

4. Run LM Studio locally (phi-3-mini with instruct GGUF)

5. Launch backend
   cd backend && uvicorn main:app --reload --port 8000

6. Launch frontend
   cd frontend && streamlit run app.py

---

📤 Output

After running, the following is generated:

• Markdown bug reproduction report
• Step-wise screenshots
• Stored inside backend/outputs/

---

📌 Example Use Case

Feed this:

“Steps to Reproduce: Open login page, type wrong password, click login. Expect error message 'Invalid credentials'. No message shown.”

🧠 AI will:

• Convert into JSON steps
• Execute UI actions
• Capture errors + screenshots
• Generate a detailed report

---

🤝 Contribute

Want to add GitHub issue parsing? LLM selector scoring? Multi-browser support?

Open a PR or connect on LinkedIn!