Gaia workers (eval)

A Gaia worker is the evaluation runner — a yasp-eval container that runs on the target GPU, self-detects its hardware, and registers with the Eval router. It backs the yasp-toolkit eval commands and the evaluation step of kernel generation, benchmarking every candidate routed to it. Deploy one before your first generation; it stays up and serves them all.

Fastest path: let the agent do it

From your reference machine, run yasp and say "deploy an eval worker to <host>" — the deploy-eval-worker skill handles the whole flow. The Docker guide is the manual equivalent, run on the target machine: it needs Docker, a working GPU runtime, and curl — not yasp-agent.

Deploy the worker

Pick the path that fits your infrastructure — each guide is self-contained (it mints the registry key, resolves the latest image, and starts the worker):

	Use when	Guide
Docker	one host — a single NVIDIA dGPU / Jetson, or AMD box	Deploy with Docker
Helm	a Kubernetes GPU cluster or fleet	Deploy with Helm

The worker's GPU is auto-detected on first registration, so you never set it. Confirm it came up with yasp-toolkit eval workers on your reference machine.

Tags & routing

The worker authenticates with your API key (YASP_API_KEY) or a worker key scoped just to registration, and can carry tags set at registration — --tag key=value (Docker) or worker.tags.* (Helm):

• host=<name> — target a specific box later.
• routing — which pool the worker joins:
  • default (no routing tag) — the shared stable pool, where evaluations route unless told otherwise.
  • routing=<name> — a separate pool for testing; target it with --routing on the yasp-toolkit eval commands, or --eval-routing on kernelgen.