VDSX | Documentation

Post-MCP Runtime

Overview

The Post-MCP Runtime is VDSX's proprietary execution substrate that reduces AI costs by 98% through zero-token tool execution.

Key Features

0-Token Execution: Tools run without consuming LLM context tokens
Streaming Results: Real-time output via SSE (Server-Sent Events)
Persistent State: Mumpix database for cost-effective storage
Rate: $5/hour vs traditional AI APIs ($10+ per request)

Cost Comparison

Traditional AI API: $10-50 per complex task

VDSX Runtime: ~$0.01 per task (1-10 seconds @ $5/hr)

Savings: 98-99% cost reduction

Mumpix Whitepaper

What is Mumpix?

Mumpix is a hybrid language combining MUMPS (M) persistence with Unix shell piping, optimized for AI workflows.

Architecture

LSM-Tree Storage: Log-Structured Merge Trees for fast writes
Global Arrays: Persistent hierarchical key-value store
Unix Pipes: Chain tools with | operator
Cost: $0.0002 per execution vs DynamoDB $0.25+

Example

do ^workflow('support') | grep('urgent') | ^sendEmail

Quantum SDK

Post-Quantum Cryptography

VDSX implements NIST-approved quantum-resistant algorithms for future-proof security.

Algorithms

Kyber-768: Key encapsulation for quantum-safe encryption
Dilithium-65: Digital signatures resistant to quantum attacks
Token Exchange: Zero-knowledge authentication handshake

Integration

const token = await quantum.getToken();
fetch('/api/endpoint', {
  headers: { 'X-Quantum-Token': token }
});

Workflow Guide

Building Workflows

Workflows are YAML definitions that chain tools together for autonomous execution.

Structure

name: Customer Support
description: Handle customer inquiries
steps:
  - tool: search_knowledge_base
    args: { query: "{user_input}" }
  - tool: generate_response
    args: { context: "{previous_output}" }
  - tool: send_email
    args: { body: "{previous_output}" }

Execution

Workflows run at $5/hour, billed per second. Average workflow costs $0.01-0.10 depending on complexity.

Tool Creation

0-Token Tools

Define Python functions that execute without LLM context consumption.

Example Tool

@tool(name="search_database")
def search_db(query: str) -> dict:
    """Search customer database"""
    results = db.query(query)
    return {"results": results, "count": len(results)}

# Register
tools.register(search_db)

Benefits

No context tokens consumed for execution
Streaming results via SSE
Automatic schema generation
Error handling & retries

VPS Deployment Guide

Self-Hosting VDSX

Deploy the full VDSX stack on your own infrastructure for complete control.

Requirements

Ubuntu 22.04+ or similar Linux
4GB+ RAM, 2+ CPU cores
Docker & Docker Compose
Domain with SSL certificate

Quick Start

git clone https://github.com/vdsx/runtime
cd runtime
cp .env.example .env
# Edit .env with your settings
docker-compose up -d

Cost

$0/month for self-hosted (Developer tier). Only pay for your VPS hosting.

Documentation

Core Architecture

Quantum Security

Developer Guides