---
license: mit
tags:
- code-quality
- tensorflow
- machine-learning
- code-review
- agentic-ai
---

# AutoReview Agent - Code Quality Scorer

A TensorFlow neural network trained to predict code quality scores (0-10).

## Model Details

- **Framework**: TensorFlow/Keras
- **Input**: 10 code features
- **Output**: Quality score (0-1)
- **Validation Loss**: 0.0006
- **Precision**: 100%

## Training

- Dataset: 1000 code samples
- Training samples: 800
- Validation samples: 200
- Hardware: GPU (Tesla T4) on Kaggle

## Usage

```python
import tensorflow as tf
import numpy as np

# Load model
model = tf.keras.models.load_model('code_quality_model.keras')

# Extract features from code
features = np.array([[200, 15, 1, 1, 5, 2, 0, 1, 3, 1]])

# Predict
prediction = model.predict(features)
quality_score = prediction[0][0] * 10
print(f"Code Quality: {quality_score:.1f}/10")
```

## Project

Part of AutoReview Agent - Autonomous Code Reviewer

Technologies:
- TensorFlow: Quality detection
- Hugging Face: Model hosting
- LangChain: Agentic reasoning
- OpenRouter 70B: Complex analysis

GitHub: https://github.com/aviral199/autoreview-agent

---

Trained on Kaggle with GPU acceleration.