fix(hermes): isolate voice runtime environment

This commit is contained in:
jenkins 2026-08-10 00:49:52 -03:00
parent 1b2053511b
commit 495638903a
2 changed files with 10 additions and 0 deletions

View File

@ -23,6 +23,8 @@ spec:
ai.bstein.dev/gpu: titan-21 dedicated speech time-slice
spec:
runtimeClassName: nvidia
automountServiceAccountToken: false
enableServiceLinks: false
nodeSelector:
kubernetes.io/hostname: titan-21
containers:
@ -34,6 +36,7 @@ spec:
env:
- {name: HOME, value: /tmp}
- {name: XDG_CACHE_HOME, value: /tmp/cache}
- {name: HERMES_STT_PORT, value: "9000"}
- {name: HERMES_STT_MODEL, value: large-v3-turbo}
- {name: HERMES_STT_CACHE, value: /opt/models/whisper}
- {name: NVIDIA_VISIBLE_DEVICES, value: all}
@ -112,6 +115,8 @@ spec:
ai.bstein.dev/model: piper-en-us-lessac-high
ai.bstein.dev/gpu: CPU-only on routing node
spec:
automountServiceAccountToken: false
enableServiceLinks: false
nodeSelector:
kubernetes.io/hostname: titan-20
containers:
@ -123,6 +128,7 @@ spec:
env:
- {name: HOME, value: /tmp}
- {name: XDG_CACHE_HOME, value: /tmp/cache}
- {name: HERMES_TTS_PORT, value: "9001"}
- {name: HERMES_TTS_VOICE, value: en_US-lessac-high}
- {name: HERMES_TTS_CACHE, value: /opt/models/piper}
startupProbe:

View File

@ -225,6 +225,10 @@ def test_voice_workloads_have_deliberate_xavier_placement():
assert "@sha256:" in tts["containers"][0]["image"]
assert stt["nodeSelector"] == {"kubernetes.io/hostname": "titan-21"}
assert tts["nodeSelector"] == {"kubernetes.io/hostname": "titan-20"}
assert stt["automountServiceAccountToken"] is False
assert tts["automountServiceAccountToken"] is False
assert stt["enableServiceLinks"] is False
assert tts["enableServiceLinks"] is False
assert stt["runtimeClassName"] == "nvidia"
stt_resources = stt["containers"][0]["resources"]
assert stt_resources["requests"]["nvidia.com/gpu.shared"] == 1