Skip to main content

Documentation Index

Fetch the complete documentation index at: https://vektorcompute-77d08130.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

The Vektor network is secured and powered by independent Tier III and Tier IV data centers around the world. By provisioning hardware to the network, Node Operators capture a continuous stream of $VKTR yield generated by global AI inference demand.

​
Hardware Requirements

Vektor does not support consumer-grade GPUs (e.g., RTX 4090s or Mac Studio clusters). To maintain our strict < 50ms latency and high-throughput SLAs for enterprise clients, the network requires data-center-grade Tensor Core architecture. Currently, the Vektor Routing Engine only accepts nodes running the following verified hardware:
  • NVIDIA H100 (80GB / 96GB) - Tier 1 Multiplier
  • NVIDIA A100 (80GB) - Tier 2 Multiplier
  • NVIDIA H200 - (Pending Mainnet Integration)
Because AI models and inference outputs can be incredibly data-heavy, Node Operators must meet strict networking thresholds to avoid bottlenecking the mesh:
  • Minimum Uplink: 10 Gbps dedicated symmetric fiber.
  • Redundancy: Dual-homed network configurations with BGP routing.
  • Storage: NVMe SSD arrays for rapid model weight loading into VRAM.

​
Service Level Agreements (SLAs) & Slashing

The Vektor Protocol requires absolute reliability. Node Operators must stake a minimum of 50,000 $VKTR as collateral to join the active routing pool. To ensure network health, the protocol enforces strict cryptographic SLAs:
  1. The 99.9% Uptime Rule: If a node drops offline and fails to respond to health-check pings for more than 4 consecutive epochs (approx. 1 hour), its collateral is subject to a 0.5% slashing penalty.
  2. Latency Gating: If a node consistently returns inference outputs slower than the regional mesh average, the Routing Engine will temporarily demote the node, reducing its request flow and subsequent yield until the bottleneck is resolved.

​
The Onboarding Process

Due to overwhelming demand from independent data centers, onboarding is currently gated. If you represent an enterprise facility with available H100/A100 clusters, you must submit a formal provisioning request through the Nodes Dashboard. Our infrastructure team will reach out to verify your hardware architecture and IP allocations before granting your cryptographic access keys.