Sigma Studio Logo

🧬 Σ-SIGMA Studio

Modular AI-Native Cognitive Operating Kernel for Multi-GPU Inference, Autonomous Multi-Agent Swarms & Dynamic Extensible Ecosystem

🇬🇧 English • 🇮🇹 Italiano • 📦 GitHub Repository

📸 Platform Screenshots

Sigma Studio Bacheca & Skills Hub

1. Interactive Release Board, Dynamic Skills Showcase Slider & 4-Step Kernel Workflow

Sigma Studio Chat AI & Multi-Agent Swarm

2. Streaming Chat AI Workspace with Sub-100ms TTFT, SigmaEngine Multi-GPU & 12 MCP Servers

Sigma Studio Modelli Hub & GGUF Forge

3. Modelli Hub: Hugging Face Downloader & Tab 2 GGUF Quantization Forge

🚀 Overview

Sigma Studio is an open-source, executable AI Operating Kernel engineered around a lightweight, watertight Micro-Kernel paired with a dynamic Runtime Module Ecosystem. It combines an ultra-fast Python 3.10+ FastAPI backend with a GPU-accelerated React 19 + Vite 8 frontend.

Multi-agent teams interact through Modelfile manifests, standard Model Context Protocol (MCP) tools, and specialized lab environments installed on-demand from the official GitHub ecosystem.

+-----------------------------------------------------------------------------------------+
|                               Σ-SIGMA STUDIO COGNITIVE KERNEL                           |
+-----------------------------------------------------------------------------------------+
|  ⚡ SigmaEngine (Multi-GPU CUDA) |  ⚙️ Providers Hub (100% Interoperable) |  📜 20 Modelfiles  |
|  (C++/PyTorch FlashAttn-2 Shard) |  (OpenAI, Claude, Gemini, DeepSeek)   |  (Manifesti Hub)  |
+-----------------------------------------------------------------------------------------+
|                              🔌 12 MODEL CONTEXT PROTOCOL SERVERS                       |
+-----------------------------------------------------------------------------------------+
|  🛠️ Dev / Workspace  |  🌐 Web Search & DNS |  ✉️ Email Client |  💬 Telegram / Slack  |
|  📅 Calendar & Tasks |  🧠 Vector Memory RAG |  🏠 IoT HomeAss  |  ⚡ GPU VRAM Flush    |
+-----------------------------------------------------------------------------------------+
|                              🧩 15 MODULAR OPEN-SOURCE LABS                             |
+-----------------------------------------------------------------------------------------+
| 🎨 Creative Lab 3D/2D      | 🧠 Training Lab & SLM     | 🎙️ Voice Studio (Kokoro)      |
| 🔬 Pipelines Lab & Swarm   | ⚡ Hardware & GPU Telemetry| 🏠 Smart Domotica Assistant    |
| 📅 Roadmap & Task Audit    | 📊 Knowledge Graph D3     | 📻 Hi-Fi Audio Lounge          |
+-----------------------------------------------------------------------------------------+

🌟 Key Architecture & Capabilities

1. ⚡ SigmaEngine Native Multi-GPU Inference

Zero-Bottleneck C++/PyTorch Layer Sharding: Automatically partitions large LLMs (from 7B to 70B+) across primary and secondary GPUs (e.g. RTX 5070 Ti + RTX 5060) and system RAM.
Sub-100ms TTFT: Ultra-low Time To First Token with native FlashAttention-2 acceleration and continuous KV-cache streaming.

2. ⚡ Modelli Hub: Hugging Face Downloader & GGUF Forge

Hugging Face Downloader: Search and download any open-source model directly from Hugging Face with resilient resumable multi-stream downloading.
Tab 2 GGUF Quantization Forge: Integrated in-memory converter and quantizer (Q4_K_M, Q5_K_M, Q8_0, FP16) to tailor model weight precision to your hardware VRAM without external CLI tools.

3. ⚙️ Providers Hub (100% Interoperable Routing)

Seamlessly switch between native local SigmaEngine execution and external Cloud Providers (OpenAI GPT-4o, Anthropic Claude 3.5, Google Gemini 2.0 Flash, DeepSeek-R1, Groq, Ollama).
Intelligent local intent router for rapid ~100ms classification and autonomous multi-agent dispatching.

4. 📜 Manifesti Hub (20 Standardized Modelfiles)

Enforce strict persona contracts, ethical boundaries, and reasoning pipelines through 20 specialized Modelfile manifests (Architect, Developer, Mathematician, Medical Specialist, Legal Jurist, Security Auditor, etc.).

5. 🔌 12 Model Context Protocol (MCP) Servers

Native Kernel Servers: Developer CLI & Pytest, Web Search & DNS Diagnostics, Email Management, Messaging Webhooks, Calendar Scheduling, Inference Fallbacks.
Modular Extension Servers: Home Assistant IoT, NVIDIA NVML Hardware & VRAM Flush, Neural Voice TTS, D3 Memory Graph, QLoRA Training.
Interactive Permission Governance: Granular confirmation dialogs and access control for system-level operations.

6. 🏛️ Watertight Sandboxed Execution

Strict Path Whitelist: Confines filesystem writes to authorized directories (data/, manifesti/, scratch/, sigma_studio/, core/).
Subprocess Isolation: Static AST analysis preventing unauthorized Python code execution.

📦 Official Modules Catalog

All optional modules can be installed with a single click from the Hub Skills & Extensions:

Module ID	Name	Category	Key Features
`sigma_creative_lab`	Creative Lab 3D/2D	Multimodal & Graphics	FLUX/SDXL Text-to-Image, SAM2/rembg background removal, Hunyuan3D/TripoSR generation, PBR materials, Blender headless rendering.
`sigma_training_lab`	Training Lab & SLM	LLM Training & SLM	Unsloth QLoRA, PEFT, Gradus Functional Weight Engine (FWE), Autopilot hyperparameter search, GGUF quantization, MMLU benchmarks.
`sigma_voice_studio`	Voice Studio & Speech	Neural Voice & Audio	Kokoro 82M ultra-fast TTS (<80ms), Coqui XTTS-v2 zero-shot voice cloning, pitch/speed tuning, live waveform visualizer, Voice MCP.
`sigma_hardware_lab`	Hardware & GPU Telemetry	System & VRAM	Live VRAM allocation, GPU/CPU telemetry charts, CUDA process monitor, zombie task termination, one-click VRAM flush.
`sigma_research_lab`	Pipelines Lab & Swarm	Research & Automation	Visual DAG pipeline designer, multi-agent research loops, step-by-step execution inspector, self-healing code generator.
`sigma_knowledge`	Argomenti & Knowledge Graph	Knowledge & Memory	D3 force-directed interactive relational graph, Universal Knowledge Nodes explorer, RAG vector search, Memory MCP server.
`sigma_roadmap`	Roadmap & Task Kanban	Productivity & Tasks	Interactive Calendar, drag-and-drop Kanban task board, chronological audit trail, milestone tracker.
`sigma_domotica`	Smart Home Assistant	IoT & Home Automation	Home Assistant WebSocket/REST bridge, device control, automation triggers, climate & solar power modulation.

⚡ Installation & Quick Start

Prerequisites

OS: Windows 10/11 or Linux x86_64
Python: 3.10 or higher
Node.js: 18.0+ and npm 9.0+
CUDA Toolkit (Recommended for GPU acceleration): NVIDIA CUDA 12.0+

1. Clone the Repository

git clone https://github.com/Sigmanih/SigmaStudio.git
cd SigmaStudio

2. Automated Setup (Windows)

Double-click install_dependencies.bat or run:

.\install_dependencies.bat

3. Manual Setup (Linux / macOS)

# 1. Setup Virtual Environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
pip install -r requirements.txt

# 2. Build Frontend
cd sigma_studio
npm install
npm run build
cd ..

# 3. Start Server
python sigma_server.py

4. Launching Sigma Studio

Run the one-click launcher on Windows:

.\sigma_studio.bat

Or start manually:

python sigma_server.py

Open your browser at http://localhost:8000.

🧪 Running Automated Tests

Run the Pytest kernel test suite:

pytest tests/ -v

All kernel tests validate MCP governance, agent routing, FastAPI endpoints, security sandboxing, and chat streaming with a 100% success rate.

📜 License & Community

Sigma Studio is licensed under the Apache-2.0 License. Continuous updates and optimizations are pushed regularly. Check the official GitHub Repository for new releases.