Voreenth

AI Runtime Security Gateway for Large Language Models

Created by Jibril Anifowoshe

Voreenth is an open-source AI Runtime Security Gateway that inspects prompts before they reach a Large Language Model (LLM). Rather than relying on the model itself to determine whether a request is safe, Voreenth evaluates prompts against security policies, assigns a risk score, and enforces allow/block decisions before inference occurs.

This project demonstrates a practical enterprise security pattern that can be applied to local and cloud-hosted AI systems.

---

Links

🚀 Live Application https://voreenth.streamlit.app

🎥 YouTube Walkthrough
https://youtu.be/kFEC75G0gJ8

🌐 Website
https://jibrilctrl.com

💼 LinkedIn
https://www.linkedin.com/in/jibril-anifowoshe

---

Features

• Runtime Prompt Inspection
• Prompt Injection Detection
• Environment Reconnaissance Detection
• System Prompt Extraction Detection
• Secret & API Key Detection
• Sensitive Data (DLP-like) Detection
• Risk Scoring & Severity Classification
• Allow / Block Policy Enforcement
• Local SQLite Security Telemetry
• Interactive Streamlit Dashboard
• Local Ollama Integration

---

Architecture

flowchart TD
    User[User Prompt] --> UI[Streamlit UI]
    UI --> Engine[Voreenth Policy Engine]
    Engine --> Detection[Security Detection Engine]
    Detection --> Risk[Risk Scoring]
    Risk --> Decision{Allow / Block}
    Decision -->|Block| Audit[SQLite Telemetry]
    Decision -->|Allow| Ollama[Ollama LLM]
    Ollama --> Response[LLM Response]
    Response --> Audit
    Audit --> Dashboard[Security Dashboard]

Loading

---

Detection Capabilities

Voreenth currently detects:

• Prompt Injection
• System Prompt Extraction
• Environment Reconnaissance
• Secret / API Key Exposure
• Sensitive Data (PII)
• Credential Extraction Attempts
• Oversized Prompt Inspection

---

Technology Stack

• Python
• Streamlit
• SQLite
• Ollama
• Qwen 3

---

Repository Structure

app.py
policy_engine.py
database.py
ollama_client.py
README.md
LICENSE
requirements.txt

---

Running Locally

python3 -m venv .venv

source .venv/bin/activate

pip install -r requirements.txt

ollama pull qwen3:1.7b

ollama serve

streamlit run app.py

---

Enterprise Mapping

Although this project runs locally using Ollama, the architecture is model agnostic and can be be positioned in front of:

• Azure OpenAI
• Azure AI Foundry
• Amazon Bedrock
• Google Vertex AI
• Anthropic Claude
• OpenAI Enterprise
• Internal AI Platforms
• Agentic AI Workflows

The same runtime inspection, risk scoring, policy enforcement, and telemetry workflow can operate independently of the underlying language model.

---

Roadmap

Future enhancements include:

• Policy-as-Code
• RBAC Integration
• Microsoft Entra ID Authentication
• Azure Key Vault Integration
• Microsoft Sentinel Integration
• Adaptive Risk Scoring
• Configurable Detection Policies
• Agent Runtime Protection
• MCP Security Controls
• RAG Inspection

---

Disclaimer

Voreenth is an educational proof-of-concept intended to demonstrate runtime AI security concepts. It is not intended to replace enterprise security products.

---

Created by Jibril Anifowoshe

© 2026 Jibril Anifowoshe • Released under the MIT License.