Automated Bug Triage System

This project is an end-to-end machine learning application designed to automate the process of bug report triage. It uses Natural Language Processing (NLP) to read a bug report's description and metadata, and predicts both the responsible engineering team (Component) and the bug's Priority.

This repository contains two different web applications that use the same trained model:

1. A simple Flask app for single predictions via a web form.
2. An advanced Streamlit dashboard for bulk predictions and analytics.

---

Key Features

• Dual Prediction: Predicts both the responsible Component and Priority for each bug report.
• Advanced ML Pipeline: Uses a LightGBM model with a scikit-learn pipeline that combines text features (TF-IDF with n-grams) and categorical features (One-Hot Encoding).
• Data Cleaning: Includes a notebook that systematically cleans the raw data and handles the real-world problem of a highly imbalanced target variable.
• Flask Web App: A simple, user-friendly web form for submitting a single bug report and seeing the instant prediction.
• Streamlit Analytics Dashboard: A powerful tool for business insights, featuring:
  • Bulk prediction via CSV upload.
  • A flexible, interactive column-mapping interface to handle various CSV formats.
  • Plotly visualizations to analyze prediction results.
• Containerized with Docker: The entire application is containerized with a Dockerfile, allowing for easy, reproducible deployment.

---

Project Structure

├── data/
│ ├── Eclipse.csv
│ └── cleaned_bug_reports.csv
├── models/
│ └── final_multimodel.pkl
├── notebooks/
│ ├── 01-data-cleaning.ipynb
│ └── 02-model-training.ipynb
├── templates/
│ └── index.html
├── app.py
├── dashboard.py
├── Dockerfile
├── requirements.txt
└── test_api.py

---

How to Run

1. Local Setup (Without Docker)

Step 1: Clone the repository and set up the environment

git clone [https://github.com/your-username/your-repo-name.git](https://github.com/Marwakhot/bug-triage-nlp-api.git)
cd bug-triage-nlp-api

python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt 

Step 2: Generate the Model File
Before you can run the applications, you must generate the model file.
Open and run the notebooks/01-data-cleaning.ipynb notebook from top to bottom.
Then, open and run the notebooks/02-model-training.ipynb notebook. This will create the final_multimodel.pkl file.

Step 3: Run an Application You can now run either the Flask app or the Streamlit dashboard.

• To run the Streamlit Dashboard (Recommended):

  streamlit run dashboard.py

  Then open your browser to http://localhost:8501

• To run the Flask App:

  python app.py

  Then open your browser to http://localhost:8080

2. Docker Setup (Recommended)

Prerequisite: You must have Docker Desktop installed and running.
Step 1: Build the Docker image In your terminal, from the project's root directory, run:

docker build -t bug-triage-dashboard .

Step 2: Run the Docker container This will start the Streamlit dashboard.

docker run -p 8501:8501 bug-triage-dashboard

Now, open your web browser and go to http://localhost:8501