🎭 Animetix: Next-Gen AI Sandbox & Interactive Games

Animetix is a next-generation AI-powered cognitive sandbox and game platform for Anime, Manga, Movies, and Video Games. Powered by a RAG 2.0 Knowledge Engine, Animetix fuses advanced semantic graphs, multimodal reasoning, and deep reinforcement learning to power a vast suite of interactive game modes and research-grade AI laboratories.

Built on an Atomic & Hexagonal Architecture, the platform runs with enterprise-grade cloud capabilities on Google Cloud Platform (GCP) in production, while maintaining a lightweight, zero-configuration local fallback ecosystem for local development.

🌐 Live at animetix.xyz

---

📌 Table of Contents

• 🎮 Next-Gen AI Game Suite
• 🔍 Universal Search Engine
• 🧪 The Ghost Labs (Advanced AI Sandbox)
• 💎 Rewarded Economy: The Berrix Model
• 🧠 Core RAG 2.0 & Cognitive Architecture
• ☁️ GCP Infrastructure
• 🚀 Getting Started
• 📂 Project Structure
• 📊 MLOps, Security & Observability
• 🧪 Running Tests
• 📜 License

---

🎮 Next-Gen AI Game Suite

Animetix features a rich variety of AI-driven game modes powered by customized domain services:

Game Mode	AI Engine / Mechanics	Description
Classic Game	Semantic similarity scores	Challenge your anime knowledge by searching for titles matching specific thematic descriptions.
Akinetix	Entropy-based decision trees	Think of a media or character; the AI will ask strategic questions to guess it.
Akinetix RL	Reinforcement Learning (PPO/DQN)	Play against or train a neural network agent learning optimal query strategies.
Animinator (Oracle)	CoVe Fact-Checking	Interrogate the AI Oracle. The agent dynamically audits facts before responding.
The Forge (Creative Fusion)	Generative fusion & diffusion	Merge two universes, art styles, or characters to synthesize hybrid works.
Forge VN	Interactive Visual Novel	Experience dynamic, AI-generated branching stories based on your creative fusions.
Vs Battle Arena	LLM Battle Simulation	Simulate hypothetical battles between characters with comparative capability analysis.
Emoji Decode	Translation sequence models	Decipher complex plot summaries translated into emojis by an adversarial LLM.
Blindtest	Audio recognition	Name the anime from its opening/ending theme (optionally warped by AI), with bonus points for the artist or track number.
Quiz Who? (Guess-Who Duel)	Character deduction duel	A real-time "Guess Who?" duel: narrow down the hidden character through attribute questions.
Covertest	Cover recognition	Identify the manga from its cover art alone, against the clock.
The Paradox	Temporal reasoning	Solve chronology puzzles where multiple anime timelines intertwine.
Daily Challenge	Deterministic daily seed	One shared daily target per universe — everyone races the same puzzle, tracked by streaks.
World Boss Raids	Cooperative weekly raid	A code-named boss spawns each week; the whole community deals damage by guessing the hidden work. Defeating it splits an XP reward across every participant, ranked on a contributor leaderboard.
Duel Arena	Realtime 1v1 (WebSockets)	Face another player head-to-head in a live anime-culture duel.
Multiplayer Labs	WebSockets Matchmaking	Play Undercover (find the intruder) or CodeManga (Codenames variation) in real-time.
Vision Quest	VLM Multimodal analysis	Identify visual tropes, frames, and scenes using multimodal vision queries.

---

🔍 Universal Search Engine

A single query box unifies metadata, semantics, and multi-agent reasoning through three complementary modes:

• Meta-Search: Hybrid lexical + semantic lookup across the entire catalog (anime, manga, characters, seiyuu), filterable by type and backed by the RAG 2.0 engine.
• Visual Nexus: Describe a scene in natural language and retrieve the matching temporal video segments via multimodal embeddings.
• Expert Nexus: An agentic RAG pipeline that streams its reasoning live over Server-Sent Events — a Monte Carlo Tree-of-Thoughts of specialized agents (semantic router, judge, synthesizer) that grounds every claim before delivering a synthesized answer.

---

🧪 The Ghost Labs (Advanced AI Sandbox)

The Ghost Labs are specialized playgrounds for exploring cutting-edge machine learning and cognitive architectures directly in the browser:

🎙️ Speech & Audio Labs

• Voice Lab / Zero-Shot Cloning: Synthesize voices and clone any voice from a short audio sample.
• Manga Voice Lab: Click on speech bubbles in a manga panel to generate matching voice-overs using cloned character voices.
• Speech-to-Speech Live: Converse with a real-time, bi-directional AI companion using the Gemini Multimodal Live API over raw WebSockets.

🖼️ Vision & Manga Translation

• Manga Lab: Run panel segmentation, optical character recognition (OCR), and translation pipelines on manga scans.
• Video Lab (FateZero): Perform temporal frame analysis, translations, and video style transfer using diffusion networks.

📐 Spatial Computing & 3D Labs

• Spatial Lab: Estimate depth maps from 2D images and export 3D object meshes.
• Cinematic Volumetric Reconstruction: View immersive, interactive 3D volumetric scenes rendered in real-time using Three.js and depth sensors.

🧠 Cognitive & Neuromorphic Sandbox

• Tree of Thoughts Visualizer: Watch the Monte Carlo Tree Search (MCTS) reasoning paths and decision trees of the thinking LLM.
• Swarm Consensus Arena: Spawn multiple LLM agents, watch them debate a prompt, and reach a unified semantic consensus.
• Strategy CFR Solver: Optimize game strategies using Counterfactual Regret Minimization (CFR) solvers.
• Liquid Neural Networks: Experiment with continuous-time dynamical LNN models predicting popularity trajectories.
• Quantum Fiction Lab: Interactive storytelling where narrative states exist in quantum-like superpositions.
• Neuro-Symbolic Profile: Formalize your player profile using Z3 Theorem Prover rules derived from user memories.
• Self-Evolving Compiler: Let the LLM self-improve and compile optimized backend routines on-the-fly.

---

💎 Rewarded Economy: The Berrix Model

Animetix is 100% free for all users, financed by a sustainable "attention-to-compute" economy. All advanced AI features (Creative Forge, Visual Novel Generation, Deep Graph Visualizer) are powered by Berrix (Bx).

• Passive Mining: Earn Berrix simply by playing our suite of interactive games (+20 Bx / 3 min).
• Active Injection: Watch short sponsored clips in the Power Station to instantly recharge your neuronal battery (+250 Bx / clip).
• Zero-Paywall Philosophy: No features are locked behind subscriptions. Only the developer API remains subject to strict quotas.

---

🧠 Core RAG 2.0 & Cognitive Architecture

Animetix separates domain rules from infrastructure concerns using a modular hexagonal approach:

• Local-First Inference: Minimizes latency and costs by prioritizing local SLMs (Qwen and Llama families) running on GPU Spot instances before falling back to cloud APIs.
• Hybrid Memory: Merges dense vector search (PostgreSQL + pgvector in production, SQLite fallback locally) with Neo4j logical knowledge graphs mapping relationships.
• Cross-Encoder Reranking: Fine-tuned cross-encoders score candidate matches to mitigate hallucination.
• Dependency Injection Everywhere: Views, services, and adapters are wired through dependency_injector containers — every collaborator is constructor-injected and overridable in tests.

---

☁️ GCP Infrastructure

For production scale and security, Animetix utilizes managed Google Cloud Platform services:

• Cloud Run (europe-west9): The web service and a scale-to-zero GPU "brain" service for heavy inference, deployed behind the custom domain animetix.xyz via a Cloudflare Worker proxy.
• Google Identity Platform (GCIP): Manages user registration and logins (Firebase Web SDK, custom DRF JWT validation, and OAuth providers: Google, Discord, X, MyAnimeList).
• Cloud Run Jobs & Cloud Scheduler: Serverless nightly database synchronizations and MLOps pipelines (catalog sync, drift baselines, RLHF loops), replacing standard Celery Beat infrastructure.
• Secret Manager: Every credential (database, Neo4j Aura, Hugging Face, Gemini) is mounted as a secret reference — no secrets ship in images or env files.
• Cloud KMS (CMEK): Customer-managed encryption keys for GCS buckets containing generated creative assets.
• Edge Defense (Cloud Armor): Preempts token-based DDoS, web application attacks, and prompt injections via CEL filtering.
• Managed Data: Serverless PostgreSQL + pgvector (Neon) and Neo4j Aura for the knowledge graph.

---

🚀 Getting Started

📋 Prerequisites

• Python 3.12+
• Node.js 18+ (for frontend development)
• Docker & Docker Compose
• Optional: GCP credentials & Firebase config.

🐳 Run with Docker (Development Mode)

1. Clone the Repository:

   git clone https://github.com/MissawB/Animetix.git
   cd Animetix

2. Configure Environment:

   cp .env.example .env

3. Spin up containers:

   docker-compose -f deploy/docker-compose.yml up --build

   Access the unified SPA on http://localhost:8000.

🐍 Manual Development Setup

1. Backend (Headless Django Server):

   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   pip install -r requirements.txt
   # Contributors who run the test suite also need the dev tooling:
   #   pip install -r requirements-dev.txt
   
   # Initialize SQLite database and media
   python backend/api/manage.py migrate
   python backend/api/manage.py seed_achievements
   python backend/api/manage.py sync_catalog
   python backend/api/manage.py runserver

2. Frontend (React SPA):

   cd frontend
   npm install
   npm run dev

   The React Vite server runs on http://localhost:5173 and proxies calls to the backend on port 8000.

---

📂 Project Structure

Animetix/
├── backend/                # Headless Django API & Orchestration
│   ├── core/               # Pure Domain Layer (Entities, Ports, Services)
│   │   ├── domain/         # Business domain logic (Akinetix, Swarm, LNN)
│   │   ├── ports/          # Abstract boundary interfaces (InferencePort)
│   │   └── utils/          # Model registry & shared security utilities
│   ├── adapters/           # Infrastructure Layer (Driven Adapters)
│   │   ├── persistence/    # pgvector, Vertex AI, Neo4j and Suwayomi adapters
│   │   └── inference/      # Google GenAI, Brain API and Ollama adapters
│   └── api/                # Presentation Layer, split by domain
│       └── animetix/api/   # core/, labs/, games/ packages + SSE stream views
├── frontend/               # React Single Page Application (TypeScript + Vite)
│   ├── src/
│   │   ├── pages/          # Decoupled Page Components (Games, Labs)
│   │   ├── store/          # Zustand Global Stores
│   │   └── features/       # Functional modules (Auth, Companion, Labs)
│   └── vite.config.ts
├── deploy/                 # Docker Compose, GCP templates & deployment configurations
│   └── deployments.yaml    # Single declarative source for every deploy target
├── docs/                   # Specification plans and historical architecture docs
└── tests/                  # Pytest Unit, Integration, and E2E (Playwright) suites

---

📊 MLOps, Security & Observability

• Prompt Sanitization: Ingested context is audited to prevent indirect prompt injections.
• Telemetry & Tracing: RAG decision steps and agent actions are injected into OpenTelemetry spans, which export directly to Google Cloud Trace when active.
• DPO Loop: User preferences and corrections are logged to fine-tune local models.
• Transparency Hub: Real hallucination rates, community satisfaction, and embedding-drift reports are published on a public transparency page.

---

🧪 Running Tests

Unit & Integration Suite

With the virtual environment active (a 75% coverage gate is enforced in CI):

pytest

End-to-End User Journeys (Playwright)

The e2e suite lives in the frontend and runs against a mocked backend API (no live Django/GPU needed):

cd frontend && npm run test:e2e

---

📜 License

Licensed under the Apache License 2.0.

---

Developed with ❤️ by the Animetix Team - 2026