Page cover

Welcome

Welcome to InfraMind

Overview

InfraMind is a decentralized compute mesh built for high-performance AI model deployment. It provides an open, fault-tolerant, and latency-aware infrastructure for developers, node operators, and autonomous agents to deploy and run containerized AI models across a globally distributed network of independent servers.

InfraMind is not a hosted platform. It’s an execution layer. Every endpoint, every job, every node interaction happens without centralized orchestration. Intelligence should not be bottled up in a data center. InfraMind allows it to move.

Features

  • Decentralized Runtime Mesh Models are served from nodes distributed globally, selected based on latency, capacity, and reliability.

  • Containerized AI Serving Models are packaged in portable OCI-compliant containers, versioned, and distributed with integrity proofs.

  • Low-Latency Scheduling Job routing is dynamic and locality-aware. Closest capable node receives the inference call.

  • Permissionless Node Contribution Any machine can become a compute node by running a lightweight agent. No KYC. No centralized approval.

  • Token-Incentivized Execution Jobs are cryptographically signed, performance is measured, and work is rewarded in $INFRA tokens.

  • Support for gRPC & REST APIs Endpoints support modern communication protocols and stateless or session-aware inference payloads.

Supported Use Cases

  • On-demand model inference with REST/gRPC endpoints

  • Distributed training across multi-node mesh clusters

  • Vision/audio model deployment to edge environments

  • Quantized model serving in low-resource regions

  • Swarm-based AI agent orchestration

  • Private model execution using TEE/WASM/FHE (in progress)

System Requirements

Minimum Requirements for Node Operators:

  • Linux/macOS (x86_64 or ARM64)

  • 2+ vCPU, 4GB+ RAM

  • Stable public IP and bandwidth

  • Docker installed (>= 20.10.x)

  • Optional: NVIDIA GPU with CUDA >= 11.0 (for high-capacity workloads)

Required Ports:

  • 9000 TCP (default inference port)

  • 4369 TCP (heartbeat + peer index)

  • 9100 TCP (metrics, optional)

Quick Start

curl -sL https://inframind.host/install.sh | bash

This command performs the following:

  • Installs Docker if not present

  • Pulls the official InfraMind Node container

  • Registers the node using a cryptographic identity

  • Connects to the InfraMind scheduler mesh

  • Starts listening for jobs in the background

Check status:

docker logs -f inframind-node

Or via the CLI:

infra status

Node Roles

Node Type
Requirements
Capabilities

CPU-only Node

Standard VM or bare-metal

Basic inference, control routing

GPU-enabled Node

CUDA-capable + drivers

LLMs, stable diffusion, quantized chains

Edge Node

ARM SBC or browser-based WASM

Low-power jobs, privacy-aware processing

TEE Node

Trusted enclave / SGX

Confidential workloads (FHE/zkML support)

What InfraMind Is Not

  • Not a cloud replacement — it's compute infrastructure that operates outside cloud vendors.

  • Not a model hub — it doesn’t host weights, only executions.

  • Not a blockchain — it's not tied to any single chain, though it uses on-chain proofs and payments.

  • Not speculative — rewards are based on completed work, not token emissions.

Ecosystem Compatibility

  • Container Format: OCI / Docker

  • Model Runtimes: Python, ONNX, Torch, TensorFlow, Rust (via WASM)

  • Networking: REST, gRPC, ZeroMQ (coming)

  • Storage: IPFS, Arweave, HTTP fallback

  • Token Layer: EVM-compatible (initial), with rollup abstraction planned

Reading the Documentation

This documentation is organized into five primary sections:

  1. Introduction Learn about the vision, problem space, and evolution of InfraMind as a compute layer.

  2. Architecture Deep dive into the technical design — from container standards to job orchestration.

  3. Running a Node Complete walkthrough for becoming a node operator, earning rewards, and contributing to the mesh.

  4. Deploying Models How to prepare, containerize, and publish AI workloads to the network.

  5. Economics & Governance Token design, staking mechanics, reputation systems, and DAO transition path.

Community & Support

InfraMind is not infrastructure as a service. It’s intelligence as a right — served by anyone, anywhere.

NextWhat is InfraMind

Last updated