FactifyAI

AI-Powered Fake News Detection for Indian Languages

A full-stack fake news detection system combining a Chrome/Edge sidebar extension with a FastAPI backend. Uses Google Gemini AI for content analysis, SerpAPI + NewsAPI for real-time source cross-referencing, and a curated credibility database of 100+ Indian news sources across 10 languages.

✨ Key Features

🧩 Browser Extension (Manifest V3)

Sidebar Panel — Tabbed interface for text analysis, source checking, and image verification
Context Menu — Right-click selected text → "Verify with FactifyAI"
Auto-Detection — Detects text selection on any webpage automatically
Real-Time Results — Confidence scores, source cards, and credibility summaries

⚙️ Backend API (FastAPI)

Gemini AI Analysis — gemini-2.5-flash-lite via Vertex AI, structured JSON output
Multi-Source Verification — SerpAPI, NewsAPI, RapidAPI, MediaStack, NewsData, GNews
Credibility Database — 100+ pre-scored Indian news domains across states
Heuristic Fallback — Clickbait/bias detection when Gemini is unavailable

🌐 Language Support

Language	Code	Regional Sources
English	`en`	International & Indian English media
Hindi	`hi`	Hindi News18, Indian Express Hindi, The Hindu Hindi
Telugu	`te`	Eenadu, Sakshi, Telangana Today
Tamil	`ta`	Dinamani, Dinakaran, Tamil The Hindu
Bengali	`bn`	Bengali News18, Bengali OneIndia
Gujarati	`gu`	Gujarati News18, Gujarati OneIndia
Marathi	`mr`	Marathi News18, Marathi OneIndia
Kannada	`kn`	Kannada News18, Kannada OneIndia
Malayalam	`ml`	Malayalam News18, Malayalam OneIndia
Punjabi	`pa`	Punjabi News18, Punjabi OneIndia

🏗 Architecture

How It Works

flowchart TD
    A([👤 User selects text on any webpage]) --> B

    subgraph EXT["🧩 Browser Extension (Manifest V3)"]
        B[Content script detects selection]
        C[Sidebar / Context Menu triggered]
        B --> C
    end

    C -->|POST /analyze_text| D

    subgraph API["⚙️ FastAPI Backend"]
        D[Request Router]

        subgraph SOURCES["📡 News Source APIs"]
            E[SerpAPI\nGoogle News]
            F[NewsAPI\nHeadlines]
            G[Optional APIs\nMediaStack · GNews\nNewsData · RapidAPI]
        end

        D --> E & F & G

        H[🔍 Relevance Filter\nRegional boost · Lang-aware\nExcludes spam domains]
        E & F & G --> H

        I[🏅 Source Credibility Engine\n100+ Indian domains · 10 languages\nDomain scoring]
        H --> I

        J{Gemini\nAvailable?}
        I --> J

        K[🤖 Gemini AI Analysis\ngemini-2.5-flash-lite\nText + Vision · Structured JSON]
        L[🔄 Heuristic Fallback\nScikit-learn TF-IDF\nClickbait & bias detection]

        J -- Yes --> K
        J -- No  --> L
    end

    K & L --> M

    subgraph RESP["📊 Structured Response"]
        M[Classification · Confidence\nExplanation · Verified sources\nCredibility summary · Language]
    end

    M --> N([✅ Results rendered in sidebar])

    style EXT fill:#e8f4fd,stroke:#4A90D9,stroke-width:1.5px,color:#1a1a2e
    style API fill:#f0fdf4,stroke:#22C55E,stroke-width:1.5px,color:#1a1a2e
    style SOURCES fill:#fef9c3,stroke:#EAB308,stroke-width:1px,color:#1a1a2e
    style RESP fill:#fdf4ff,stroke:#A855F7,stroke-width:1.5px,color:#1a1a2e

Image Verification Pipeline

flowchart LR
    A([🖼️ Image URL or base64]) --> B
    B[OCR Extraction\nExtract text from image] --> C
    C[Fact-Check Text\nRun through main AI pipeline] --> D
    D[Reverse Image Search\nGoogle Lens · Google Images] --> E
    E([📋 Authenticity report\nis_authentic · confidence\nsimilar_images])

    style A fill:#fdf4ff,stroke:#A855F7,stroke-width:1.5px
    style E fill:#f0fdf4,stroke:#22C55E,stroke-width:1.5px

Project Structure

FactifyAI/
├── 📂 backend/
│   ├── main.py                          # FastAPI app — endpoints & request models
│   ├── 📂 services/
│   │   ├── gemini_service.py            # Vertex AI Gemini (text + vision)
│   │   ├── serpapi_service.py           # SerpAPI news & reverse image search
│   │   ├── newsapi_service.py           # NewsAPI headlines & article search
│   │   ├── additional_apis_service.py   # RapidAPI, MediaStack, NewsData, GNews
│   │   ├── source_verification.py       # Domain credibility database & scoring
│   │   └── relevance_filter.py          # Regional language-aware source filtering
│   ├── 📂 utils/
│   │   ├── language_detector.py         # Unicode-based Indic language detection
│   │   └── text_processor.py            # Text cleaning, keyword extraction, bias detection
│   ├── 📂 models/
│   │   ├── fake_news_model.pkl          # Scikit-learn fallback model
│   │   └── vectorizer.pkl               # TF-IDF vectorizer for fallback
│   ├── requirements.txt
│   └── env.example                      # Environment variable template
├── 📂 extension/
│   ├── manifest.json                    # Manifest V3 config
│   ├── background.js                    # Service worker — context menus, messaging
│   ├── content.js                       # Content script — text selection detection
│   ├── sidebar.html                     # Sidebar panel UI
│   ├── sidebar.css                      # Sidebar styles
│   ├── sidebar.js                       # Sidebar logic — API calls, result rendering
│   └── 📂 icons/                        # Extension icons (16 / 48 / 128px + logo)
├── create_env.py                        # Helper to generate .env from template
├── start_backend.py                     # One-click backend startup script
└── test_final.py                        # API integration test script

🚀 Getting Started

Prerequisites

Python 3.8+
Chrome or Edge browser
API keys (see below)

1 · Clone & Install

git clone https://github.com/YOUR_USERNAME/FactifyAI.git
cd FactifyAI

# Install backend dependencies
cd backend
pip install -r requirements.txt

2 · Configure API Keys

# Option A — use the helper script (recommended)
cd ..
python create_env.py
# Then edit backend/.env with your actual keys

# Option B — copy the template manually
cp backend/env.example backend/.env
# Edit backend/.env with your keys

Required API keys:

API	Purpose	Get Key
Google Cloud (Gemini)	AI text analysis, OCR, image analysis	Google Cloud Console — Enable Vertex AI, create a Service Account, download JSON key
SerpAPI	Google News search, reverse image search	serpapi.com
NewsAPI	News article search & headlines	newsapi.org

Optional API keys (enhance accuracy with additional sources):

API	Purpose	Get Key
RapidAPI	News aggregator	rapidapi.com
MediaStack	Global news feed	mediastack.com
NewsData	News articles	newsdata.io
GNews	Google News aggregator	gnews.io

💡 Gemini Credentials: Place the downloaded service account JSON as backend/credentials.json, or set individual environment variables in .env. The service checks for the JSON file first and falls back to environment variables.

3 · Start the Backend

# Option A — startup script (recommended)
python start_backend.py

# Option B — run uvicorn directly
cd backend
uvicorn main:app --reload --host 0.0.0.0 --port 8000

API available at http://localhost:8000 · Interactive docs at /docs

4 · Load the Extension

Open chrome://extensions/ (or edge://extensions/)
Enable Developer mode
Click Load unpacked → select the extension/ folder
Pin the FactifyAI extension for easy access

📡 API Reference

`POST /analyze_text` — Analyze news text

// Request
{ "text": "News article text to analyze", "url": "https://source-url.com" }

// Response
{
  "classification": "Likely Genuine | Potentially Misleading",
  "confidence": 0.85,
  "explanation": "Detailed analysis...",
  "language": "en",
  "key_points": "Key findings...",
  "sources_found": 5,
  "verified_sources": [{ "title": "...", "domain": "...", "authenticity_score": 0.9 }],
  "credibility_summary": { "average_credibility": 0.82, "high_credibility_count": 3 }
}

`POST /check_source` — Check domain credibility

// Request
{ "url": "https://example.com/article" }

// Response
{
  "domain": "example.com",
  "credibility_score": 0.75,
  "assessment": "Moderately Credible",
  "details": "Analysis details...",
  "recommendations": "Cross-reference with other sources"
}

`POST /reverse_image` — Verify an image

// Request
{ "image_url": "https://example.com/image.jpg", "image_data": "base64..." }

// Response
{
  "status": "success",
  "analysis": "OCR + fact-check + reverse search results...",
  "extracted_text": "Text found in image...",
  "fact_check_result": { "classification": "...", "confidence": 0.8 },
  "is_authentic": true,
  "confidence": 0.85,
  "similar_images": [...]
}

Other Endpoints

Method	Endpoint	Description
`POST`	`/comprehensive_verification`	Full pipeline: text + multi-source + credibility in one request
`GET`	`/health`	Availability status of Gemini, SerpAPI, and NewsAPI
`GET`	`/languages`	List all supported languages

🛠 Tech Stack

graph LR
    A["🤖 AI Engine\nGemini 2.5 Flash Lite\nVertex AI"]
    B["⚙️ Backend\nPython · FastAPI\naiohttp · Pydantic"]
    C["📰 News APIs\nSerpAPI · NewsAPI\nRapidAPI · MediaStack\nNewsData · GNews"]
    D["🔤 NLP\nUnicode Indic detection\nRegex text processing"]
    E["🧠 ML Fallback\nscikit-learn\nTF-IDF + Classifier"]
    F["🧩 Extension\nManifest V3\nSide Panel API\nService Workers"]
    G["🔐 Auth\nGoogle Service Account\nOAuth2"]

    B --> A
    B --> C
    B --> D
    B --> E
    F --> B
    G --> A

    style A fill:#e8f4fd,stroke:#4285F4,stroke-width:1.5px
    style B fill:#f0fdf4,stroke:#22C55E,stroke-width:1.5px
    style C fill:#fef9c3,stroke:#EAB308,stroke-width:1.5px
    style D fill:#fdf4ff,stroke:#A855F7,stroke-width:1.5px
    style E fill:#fff7ed,stroke:#F97316,stroke-width:1.5px
    style F fill:#f0f9ff,stroke:#0EA5E9,stroke-width:1.5px
    style G fill:#fef2f2,stroke:#EF4444,stroke-width:1.5px

Made with ❤️ for India's multilingual internet

Navigation Menu

Repository files navigation

FactifyAI

AI-Powered Fake News Detection for Indian Languages

✨ Key Features

🧩 Browser Extension (Manifest V3)

⚙️ Backend API (FastAPI)

🌐 Language Support

🏗 Architecture

How It Works

Image Verification Pipeline

Project Structure

🚀 Getting Started

Prerequisites

1 · Clone & Install

2 · Configure API Keys

3 · Start the Backend

4 · Load the Extension

📡 API Reference

`POST /analyze_text` — Analyze news text

`POST /check_source` — Check domain credibility

`POST /reverse_image` — Verify an image

Other Endpoints

🛠 Tech Stack

About

Contributors

Languages

Name		Name	Last commit message	Last commit date
.github/workflows		.github/workflows
backend		backend
extension		extension
.gitignore		.gitignore
README.md		README.md
create_env.py		create_env.py
start_backend.py		start_backend.py
test_final.py		test_final.py

Latest commit

History

Folders and files

Repository files navigation

FactifyAI

AI-Powered Fake News Detection for Indian Languages

✨ Key Features

🧩 Browser Extension (Manifest V3)

⚙️ Backend API (FastAPI)

🌐 Language Support

🏗 Architecture

How It Works

Image Verification Pipeline

Project Structure

🚀 Getting Started

Prerequisites

1 · Clone & Install

2 · Configure API Keys

3 · Start the Backend

4 · Load the Extension

📡 API Reference

POST /analyze_text — Analyze news text

POST /check_source — Check domain credibility

POST /reverse_image — Verify an image

Other Endpoints

🛠 Tech Stack

About

Resources

Stars

Watchers

Forks

Contributors

Languages

`POST /analyze_text` — Analyze news text

`POST /check_source` — Check domain credibility

`POST /reverse_image` — Verify an image