ADFMentor 0.4.0

Parse Azure Data Factory project files and evaluate them using AI models.

ADFMentor

A Python package that evaluates student Azure Data Factory (ADF) submissions using Google Gemini AI.

Features

• 🤖 AI-Powered Evaluation — Gemini analyzes ADF JSON files and scores them against lecture tasks
• 🎯 Auto Task Identification — AI identifies which lecture task each submitted file belongs to
• 📦 ZIP Support — Automatically extracts .zip submissions
• 📄 Single JSON — Also accepts individual .json files
• 📊 Per-Task Scoring — Returns overall score + per-task breakdown with feedback

Installation

pip install ADFMentor

Or from source:

git clone https://github.com/yourusername/ADFMentor.git
cd ADFMentor
pip install -e .

Quick Start

from ADFmentor import ADFMentor

mentor = ADFMentor(api_key="your-gemini-api-key")

questions = """
Part 1 — Create a pipeline named API_Customers_To_SQL that copies CSV data from HTTP API to Azure SQL.
Part 2 — Create a dynamic pipeline that processes multiple CSV files from an API folder.
"""

prompt = """Evaluate the ADF pipeline for correctness, best practices, and requirements alignment.
Provide a score out of 100 and concise feedback."""

result = mentor.evaluate(
    answer_path="submission.json",   # or "submission.zip"
    questions=questions,
    prompt=prompt,
)

print(f"Score: {result['score']}/100")
print(f"Feedback: {result['feedback']}")

for task in result.get("task_mapping", []):
    print(f"\n  File: {task['file']}")
    print(f"  Task: {task['identified_task']}")
    print(f"  Score: {task['score']}/100")
    print(f"  Feedback: {task['feedback']}")

Response Format

{
  "score": 85,
  "feedback": "Overall summary...",
  "task_mapping": [
    {
      "file": "taskA.json",
      "identified_task": "Part 1 — Static API CSV Pipeline",
      "score": 90,
      "feedback": "Pipeline correctly implements..."
    },
    {
      "file": "taskB.json",
      "identified_task": "Part 2 — Dynamic Pipeline",
      "score": 80,
      "feedback": "Good structure but missing..."
    }
  ]
}

Package Structure

ADFmentor/
├── __init__.py        # Package entry point
├── core.py            # ADFMentor class with evaluate() method
├── utils.py           # File handling, ZIP extraction, JSON reading
└── models/
    ├── __init__.py
    ├── model.py       # Abstract base model
    └── gemini.py      # Google Gemini implementation

API Reference

ADFMentor(api_key, model_name="gemini-2.0-flash-exp")

Initialize the mentor with your Gemini API key.

mentor.evaluate(answer_path, questions, prompt) -> dict

Parameter	Type	Description
answer_path	str	Path to a .json file or .zip archive
questions	str	Full lecture task descriptions (all tasks)
prompt	str	Evaluation criteria for the AI

Returns a dict with score, feedback, and task_mapping.

Configuration

Create a .env file:

API_KEY=your_gemini_api_key_here

from dotenv import load_dotenv
import os

load_dotenv()
mentor = ADFMentor(api_key=os.getenv("API_KEY"))

Requirements

• Python 3.9+
• Google Gemini API key (get one here)

License

MIT License — see LICENSE file for details.