TruthGuard AI is a multi-modal, real-time misinformation detection platform leveraging Snowflake Cortex APIs and multiple AI models (Mistral, Claude, Llama) for enterprise-grade fact verification and content safety.

• Inspiration
• What it Does
• Features
• Architecture
• Getting Started
• Usage
• Contributing
• License
• Acknowledgements

With the explosion of AI-generated content, deepfakes, and misinformation on social media, organizations struggle to verify content in real-time. TruthGuard AI was inspired by the need for automated, verifiable, and auditable misinformation detection to protect elections, public health, and scientific integrity.

TruthGuard AI processes text, images, and video to detect misinformation across six critical domains:

1. News
2. Elections
3. Climate
4. Health & Viral Content
5. Deepfakes
6. Mental Health

It runs three state-of-the-art AI models in parallel to provide a consensus-based verdict with real-time logging, audit trails, and exportable reports.

• Multi-model consensus verification (Mistral, Claude, Llama)
• Real-time streaming API with instant results
• Categorized analysis for domain-specific accuracy
• JSON/CSV output and visualization ready
• Full audit trail stored in Snowflake
• Production-ready REST API

flowchart TD

subgraph UI[User Interface]
  UI1["Streamlit Frontend - Glassmorphism UI"]
  UI2["Real-time Input - Text, Image, Video"]
end

subgraph BE[Backend Services]
  B1["Python API Service"]
  B2["Consensus Engine - Weighted Multi-Model Scoring"]
  B3["Result Formatter - JSON, CSV, Visualization"]
end

subgraph AI[Snowflake Cortex APIs]
  C1["Mistral-Large2"]
  C2["Claude-3.5-Sonnet"]
  C3["Llama3.1-70B"]
end

subgraph SF[Snowflake Data Infrastructure]
  D1["Log Table - Queries, Responses, Scores"]
  D2["Audit Trail - Timestamped Records"]
  D3["Dashboard Analytics - SQL + Visualization"]
end

UI1 -->|User enters content| UI2
UI2 -->|POST /analyze| B1
B1 -->|REST API call| C1
B1 -->|REST API call| C2
B1 -->|REST API call| C3
C1 -->|Responses| B2
C2 -->|Responses| B2
C3 -->|Responses| B2
B2 -->|Aggregated verdict| B3
B3 -->|Formatted Output| UI1
B1 -->|Store logs| D1
B2 -->|Store consensus data| D2
D1 --> D3
D2 --> D3
D3 -->|Analytics feedback| UI1

• Python 3.11+
• Streamlit
• Requests / HTTP client
• Snowflake account with Cortex API access

git clone https://github.com/yourusername/truthguard-ai.git
cd truthguard-ai
python -m venv venv
# Linux / macOS
source venv/bin/activate
# Windows
venv\Scripts\activate
pip install -r requirements.txt

1. Set your Snowflake credentials in .env:

SNOWFLAKE_ACCOUNT=<account_name>
SNOWFLAKE_USER=<user>
SNOWFLAKE_PAT= Personal access token

2. Update API keys for Mistral, Claude, Llama if needed.

Run the app locally:

streamlit run streamlit_app.py

• Open http://localhost:8501
• Enter content (text, image, video)
• Select the category (News, Election, Climate, etc.)
• View real-time consensus verdict and audit logs

We welcome contributions!

1. Fork the repo
2. Create a new branch: git checkout -b feature/my-feature
3. Commit changes: git commit -m 'Add new feature'
4. Push to branch: git push origin feature/my-feature
5. Open a Pull Request

Please follow the code of conduct and ensure all new features are tested.

This project is licensed under the MIT License - see the LICENSE file for details.