vrr/spawn
2
0
Fork 0
mirror of https://github.com/OpenRouterTeam/spawn.git synced 2026-05-05 23:50:48 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 8
spawn/vastai/README.md
A 5cb837bbe3
feat: Add Vast.ai GPU marketplace as cloud provider with 3 agent scripts (#131) 
Adds Vast.ai as a new cloud provider (GPU marketplace with per-hour pricing).
Implements vastai/lib/common.sh with provider primitives using the vastai CLI,
and adds 3 initial agent scripts (claude, aider, codex). Remaining 11 agents
added as "missing" matrix entries.

Agent: cloud-scout

Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
2026-02-09 20:53:22 -08:00

1.5 KiB
Raw Blame History

Vast.ai

Vast.ai GPU marketplace via CLI. Vast.ai

Prerequisites

  1. A Vast.ai account with API key from Account Settings
  2. Python 3 with pip (for installing the vastai CLI)

Agents

Claude Code

bash <(curl -fsSL https://openrouter.ai/lab/spawn/vastai/claude.sh)

Aider

bash <(curl -fsSL https://openrouter.ai/lab/spawn/vastai/aider.sh)

Codex CLI

bash <(curl -fsSL https://openrouter.ai/lab/spawn/vastai/codex.sh)

Non-Interactive Mode

VASTAI_SERVER_NAME=dev-gpu \
VASTAI_API_KEY=your-api-key \
OPENROUTER_API_KEY=sk-or-v1-xxxxx \
  bash <(curl -fsSL https://openrouter.ai/lab/spawn/vastai/claude.sh)

Environment Variables

Variable Description Default
VASTAI_API_KEY Vast.ai API key (prompted)
VASTAI_SERVER_NAME Instance label (prompted)
VASTAI_GPU_TYPE GPU type to search for RTX_4090
VASTAI_DISK_GB Disk size in GB 40
VASTAI_IMAGE Docker image nvidia/cuda:12.1.0-devel-ubuntu22.04
OPENROUTER_API_KEY OpenRouter API key (prompted via OAuth)

Notes

  • Vast.ai is a GPU marketplace -- instances come with NVIDIA GPUs and CUDA pre-installed
  • The vastai CLI is installed automatically if not present (pip install vastai)
  • Instances are Docker containers; base tools are installed automatically on first run
  • SSH access is via dynamic port mapping (non-standard ports)
  • Pricing is per-hour, varies by GPU type and availability