Fix Gemma 4 vLLM startup and improve partial-provisioning UX

- provisioning.rb: auto-install pytest before pre_start_cmd to fix missing
dependency in nightly vLLM images that crashes EngineCore on cupy import
- cli.rb: relax active_config_loaders/vm1_alive? to match by vm_id+public_ip
instead of requiring status==ACTIVE, so watch/status/test work on VMs
that were interrupted mid-provisioning
- vm_lifecycle.rb: persist API status/vm_state back to state file during
status calls; add 'Provisioning: incomplete' notice when provisioned_at
is missing
- provisioning_orchestrator.rb: add wait_for_ssh polling before
ensure_trusted_host to fix race between SSH daemon readiness and keyscan

1 files changed, 3 insertions, 0 deletions

diff --git a/lib/hyperstack/provisioning.rb b/lib/hyperstack/provisioning.rb
index 19a3d33..dfe5b8d 100644
--- a/lib/hyperstack/provisioning.rb
+++ b/lib/hyperstack/provisioning.rb
@@ -162,6 +162,9 @@ module HyperstackVM
       # When set, --entrypoint bash is used so the command can patch dependencies at runtime
       # (e.g. upgrading transformers for Gemma 4, which requires transformers>=5.x).
       pre_cmd = (cfg.key?('pre_start_cmd') ? cfg['pre_start_cmd'] : nil) || @config.vllm_pre_start_cmd
+      # vLLM nightly images may be missing pytest which cupy imports during engine init.
+      # Prepend a quiet install so any pre_start_cmd also satisfies this dependency.
+      pre_cmd = "pip install -q pytest 2>/dev/null; #{pre_cmd}" if pre_cmd
       port = @config.ollama_port
 
       docker_args = [