https://github.com/ericshantos/fakecheck_api
API that analyzes news content and verifies its veracity using AI.
https://github.com/ericshantos/fakecheck_api
deep-learning fake-news nestjs nodejs tensorflow
Last synced: 3 months ago
JSON representation
API that analyzes news content and verifies its veracity using AI.
- Host: GitHub
- URL: https://github.com/ericshantos/fakecheck_api
- Owner: ericshantos
- Created: 2025-04-03T20:15:40.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2026-01-17T23:27:40.000Z (6 months ago)
- Last Synced: 2026-01-18T10:38:48.537Z (6 months ago)
- Topics: deep-learning, fake-news, nestjs, nodejs, tensorflow
- Language: TypeScript
- Homepage:
- Size: 32.6 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
[🇧🇷] [Lê em português](README.pt.md)
# FakeCheck API v4.1.1
A high-performance API for detecting fake news in Portuguese using deep learning, built with NestJS and Python microservices.
## Table of Contents
- [FakeCheck API v4.1.1](#fakecheck-api-v411)
- [Table of Contents](#table-of-contents)
- [Problem Statement](#problem-statement)
- [Solution](#solution)
- [Features](#features)
- [Core Functionality](#core-functionality)
- [Technical Features](#technical-features)
- [Technical Architecture](#technical-architecture)
- [Model Details](#model-details)
- [Architecture](#architecture)
- [Performance](#performance)
- [Key Features](#key-features)
- [API Documentation](#api-documentation)
- [Endpoints](#endpoints)
- [Installation](#installation)
- [Prerequisites](#prerequisites)
- [Setup](#setup)
- [Contributing](#contributing)
- [License](#license)
- [Acknowledgments](#acknowledgments)
## Problem Statement
Fake news has become a significant social problem, spreading misinformation rapidly through digital channels. The lack of reliable tools to automatically verify news authenticity in Portuguese exacerbates this issue, particularly in Brazilian media ecosystems.
## Solution
FakeCheck provides an API that:
1. Extracts text from news articles via URL using advanced scraping
2. Processes content with optimized NLP pipelines
3. Classifies articles using our custom LSTM model (95% accuracy)
4. Returns veracity with confidence scoring
5. Provides real-time system health monitoring
## Features
### Core Functionality
- **System Health**: Comprehensive monitoring with rate limiting
- **Model Metadata**: Version and architecture details
- **Project Information**: Technology stack and credits
### Technical Features
- **Microservices Architecture**: Node.js + Python services
- **Containerized**: Docker Compose ready (Node, Python)
- **Rate Limiting**: Protection against abuse
- **Detailed Logging**: Request/error tracking
- **Swagger Docs**: Interactive API documentation
- **Validation**: Input sanitization and validation
- **Type Safety**: TypeScript interfaces throughout
## Technical Architecture
```
├── API Gateway (NestJS)
│ ├── Modules
│ │ ├── Check - News verification
│ │ ├── Health - System monitoring
│ │ ├── Info - Model metadata
│ │ └── Shared - Common utilities
│ ├── Controllers
│ ├── Services
│ ├── Middlewares
│ └── Pipes/Interceptors
│
└── Predictor Service (Python)
├── LSTM Model (TensorFlow/Keras)
├── Text Preprocessing (spaCy)
└── Socket Server (Threaded)
```
## Model Details
### Architecture
Input Layer → Embedding Layer (300D) → Bidirectional LSTM (128 units) →
Dense Layer (64 units, ReLU) → Output Layer (1 unit, Sigmoid)
### Performance
| Metric | Value |
|--------------|-------|
| Accuracy | 95% |
| Precision | 96% |
| Recall | 94% |
| F1-Score | 95% |
| ROC AUC | 96% |
### Key Features
- Optimized for Brazilian Portuguese
- Journalistic vocabulary handling
- Confidence thresholding (0.7 default)
- Hugging Face Hub integration
- Efficient text preprocessing
## API Documentation
Interactive documentation available at `/docs` when running locally.
### Endpoints
| Method | Endpoint | Description | Parameters | Rate Limit |
|--------|----------|-------------|------------|------------|
| POST | /check | Verify news | `{ "url": "string" }` | 50/15min |
| GET | /info | Model info | - | 100/15min |
| GET | /health | System health | - | 10/1min |
**Example Request:**
```bash
curl -X POST http://localhost:3000/check \
-H "Content-Type: application/json" \
-d '{"url":"https://example.com/news-article"}'
```
**Example Response:**
```json
{
"veracity": "real",
"confidence": 0.92,
"threshold": 0.7,
"extracted_at": "2025-04-18T12:34:56.789Z"
}
```
## Installation
### Prerequisites
- Docker 20.10+
- Docker Compose 2.0+
### Setup
```bash
git clone https://github.com/ericshantos/fakeCheck_API.git
cd fakeCheck_API
docker-compose up --build
```
Access API at `http://localhost:3000`
## Contributing
1. Fork the repository
2. Create feature branch (`git checkout -b feature/your-feature`)
3. Commit changes (`git commit -am 'Add feature'`)
4. Push branch (`git push origin feature/your-feature`)
5. Open Pull Request
## License
This project is licensed under the MIT License - see the [LICENSE](./LICENSE) file for details.
## Acknowledgments
- [Fake.Br Corpus](https://github.com/roneysco/Fake.br-Corpus) for training data
- Programadores do Futuro for support
- TensorFlow/Keras team
- NestJS framework