Skip to content

environments.md

Latest commit

 

History

History
36 lines (28 loc) · 1.95 KB

environments.md

File metadata and controls

36 lines (28 loc) · 1.95 KB

Environments

cpu-local

  • Build with make vllm-sr-dev
  • Start with vllm-sr serve --image-pull-policy never
  • Use this for the default local Docker workflow
  • Default smoke config: config.agent-smoke.cpu.yaml
  • If you need a non-default config, run make agent-serve-local ENV=cpu AGENT_SERVE_CONFIG=<config>
  • For isolated parallel local stacks, add AGENT_STACK_NAME=<name> and AGENT_PORT_OFFSET=<n>, for example: make agent-serve-local ENV=cpu AGENT_STACK_NAME=lane-a AGENT_PORT_OFFSET=0 and make agent-serve-local ENV=cpu AGENT_STACK_NAME=lane-b AGENT_PORT_OFFSET=200
  • Use the same AGENT_STACK_NAME and AGENT_PORT_OFFSET values with make agent-smoke-local and make agent-stop-local

amd-local

  • Build with make vllm-sr-dev VLLM_SR_PLATFORM=amd
  • Start with vllm-sr serve --image-pull-policy never --platform amd
  • Use this for ROCm/AMD validation and platform-default image checks
  • Default smoke config: config.agent-smoke.amd.yaml
  • If you need a non-default config, run make agent-serve-local ENV=amd AGENT_SERVE_CONFIG=<config>
  • The same AGENT_STACK_NAME=<name> and AGENT_PORT_OFFSET=<n> overrides work for isolated AMD-local stacks
  • For real AMD model deployment and backend container setup, read deploy/amd/README.md
  • Use deploy/amd/config.yaml as the reference YAML-first AMD routing profile
  • See amd-local.md

ci-k8s

  • Run local profile checks with make e2e-test E2E_PROFILE=<profile>
  • CI expands to the standard kind/Kubernetes matrix in integration-test-k8s.yml

Selection Rule

  • Default to cpu-local
  • Use amd-local when platform behavior, ROCm image selection, or AMD defaults are affected
  • Use ci-k8s for merge-gate coverage and all profile-sensitive routing/deploy behavior