Skip to content

AxelS27/Phylaxify

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

33 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

πŸ›‘οΈ Phylaxify: The Quiet Guardian of Digital Spaces

A high-performance AI-powered filtering system designed to protect content creators and streamers from donation spam, illegal gambling promotions, predatory loans, and malicious bots. Built with a Triple-Layer Shield architecture and a real-time OBS Overlay Integration.

πŸŽ₯ Demo

Click the image below to watch the demo video:

Phylaxify Demo Video

πŸ“Έ Preview




πŸ›οΈ System Architecture

Phylaxify moves away from traditional monolithic models in favor of a Distributed approach (leveraging specialized cloud services) to ensure high scalability and performance.



graph LR
    subgraph Client_Tier [Client]
        A["React Dashboard"]
        B["OBS Live Overlay"]
    end

    subgraph Infrastructure_Tier [Infrastructure & Logic Tier]
        subgraph Supabase_Infra [Supabase Cloud]
            C["Auth Service (GoTrue)"]
            D["PostgreSQL Database"]
            E["Realtime Engine (WebSockets)"]
        end

        subgraph Vercel_Edge [Vercel]
            F["Webhook Receiver API"]
            G["Protection Orchestrator"]
        end

        subgraph AI_Microservice [Hugging Face]
            H["Shield BERT"]
        end
    end

    %% Communication Flows
    A --> C
    A --> D
    B --> E
    D --> E
    F --> G
    G --> D
    G --> H

    %% Styling to remove yellow and keep boxes consistent
    style Client_Tier fill:none
    style Supabase_Infra fill:none
    style Vercel_Edge fill:none
    style AI_Microservice fill:none
    style Infrastructure_Tier fill:none,stroke:none
Loading

Phylaxify's distributed model ensures that each component can scale independently:

  • Cloud Infrastructure: Supabase handles Auth and DB management, providing a secure, auto-scaling data layer.
  • Microservice Isolation: The AI Inference Engine (Shield BERT) is isolated as a dedicated microservice, allowing it to scale independently on GPU-optimized hardware.
  • Event-Driven Webhooks: High-performance Vercel functions orchestrate the "Shield Logic" only when a donation event is triggered.

πŸ›οΈ System Communication Flow

graph TD
    A[Incoming Donation: Webhook/API] --> B[Inference Gate]
    B --> C{Shield Logic Consensus}
    C --> D[Layer 1: Smart Blocklist]
    C --> E[Layer 2: Shield BERT Neural Net]
    C --> F[Layer 3: Linguistic Heuristics]
    D & E & F --> G[Consensus Verdict]
    G --> H[Supabase Realtime Engine]
    H --> I[OBS Live Overlay]
    H --> J[Guardian Dashboard]
Loading

🏁 Quick Start

1. Guardian Dashboard (Web UI)

# Clone the repository
git clone https://github.com/AxelS27/Phylaxify.git
cd Phylaxify

# Install dependencies
npm install

# Launch development server
npm run dev

2. Environment Configuration

Create a .env file in the root directory with your Supabase credentials:

VITE_SUPABASE_URL=your_supabase_url
VITE_SUPABASE_ANON_KEY=your_anon_key

πŸš€ Key Features

  • Triple-Layer Defense: Combines deterministic Regex/Keyword matching with probabilistic Neural Analysis (BERT) and linguistic heuristics.
  • Sub-second Latency: Engineered for live environments, ensuring protection decisions are made in under 200ms to maintain stream flow.
  • OBS Real-time Overlay: Dynamic overlay system that broadcasts "Passed" or "Blocked" status instantly to your production software.
  • Smart Analytics Hub: Comprehensive dashboard to monitor protection logs, confidence scores, and spam trends.
  • Proactive Blocklist: Pre-configured filters specifically tuned for regional "Judol" (gambling) and predatory loan patterns.

πŸ—οΈ Full-Stack Architecture & Infrastructure

πŸ› οΈ Technology Stack Detail

1. Frontend: The Guardian Interface

  • Framework: React 19 with TypeScript and Vite.
  • Styling: Vanilla CSS : a custom framework focused on cinematic aesthetics, glassmorphism, and responsive layouts.
  • Motion Engine: Framer Motion for micro-animations and fluid section transitions.

2. Backend: Data & Realtime Layer

  • Infrastructure: Supabase handles all persistent data, including protection logs, user profiles, and custom blocklists.
  • Auth Service: Managed GoTrue via Supabase for secure, encrypted user authentication.

3. AI Engine: Shield BERT

  • Model Architecture: Based on the BERT (Bidirectional Encoder Representations from Transformers) model.
  • Fine-tuning: Specifically trained on a custom dataset of regional spam patterns, "leetspeak" gambling promotions, and predatory loan messages.
  • Inference: Optimized for sub-second responses using ONNX Runtime or specialized serverless workers.

4. Deployment & DevOps

  • Dashboard Hosting: Vercel (Global Edge Network) for maximum frontend performance.
  • API Layer: Vercel Serverless Functions (Node.js/Python) for handling webhooks and shield logic.
  • Inference Hosting: Hugging Face Spaces (Dockerized) for heavy-duty AI processing or Vercel Edge Functions for lightweight heuristics.
  • Database: Supabase Cloud (AWS Infrastructure).

πŸ› οΈ Technology Stack

Core Intelligence & Security

  • Shield BERT: Context-aware neural network for detecting "hidden" spam patterns.
  • Smart Filter Engine: High-performance Regex and Keyword processing.
  • Consensus Logic: Weighted voting mechanism between AI and rule-based layers.

Guardian Interface

  • Frontend: React 19, TypeScript, Vite.
  • Styling & Motion: Custom CSS + Framer Motion for rich, cinematic aesthetics.
  • Database & Realtime: Supabase.

🧠 The 'Triple-Layer' Protection Engine

Our protection framework is designed to eliminate both obvious spam and sophisticated, context-dependent malicious messages.

1. Analytical Pipeline

To ensure absolute security, every incoming message passes through a 3-stage validation gate:

  • Smart Blocklist (Layer 1): Immediate rejection of known blacklisted terms, domains, and regex patterns.
  • Shield BERT (Layer 2): Analyzes the semantic intent of the message. This catches spam that uses "leetspeak" or creative spacing to bypass traditional filters.
  • Linguistic Heuristics (Layer 3): Examines character distributions, URL density, and suspicious patterns common in automated bot attacks.

2. Shield Verdict Comparison

Layer Type Detects Latency
Smart Filter Deterministic Known Keywords / Domains < 10ms
Shield BERT Neural Contextual & Hidden Spam ~150ms
Heuristics Statistical Bot Patterns & Repetitions < 30ms

3. Performance Integrity

Phylaxify is optimized for high-traffic environments, maintaining a 99.7% Accuracy Rate in identifying malicious donation payloads while ensuring legitimate viewer interaction remains uninterrupted.

πŸ“‚ Repository Roadmap

  • src/pages/: Core functional views (Dashboard, Test Lab, Settings).
  • src/components/: Reusable UI components built with "Rich Aesthetics" principles.
  • src/lib/: Integration logic for Supabase, Auth, and Profile management.
  • src/overlay/: (Planned) Dedicated logic for OBS browser source integration.

Β© 2026 AxelS27 | Guardian of Digital Integrity

About

AI powered livestream donation filter & alert protection system. Built with a Triple Layer Shield (BERT Neural Net, Regex, and Heuristics) and real-time OBS Overlay integration to protect streamers from donation spam, gambling promotions, and malicious bots.

Topics

Resources

Stars

Watchers

Forks

Contributors