Sirius Claw
9673b76fc7
fix: trigger KEDA scale-up on 1 request instead of 3
...
- Wake script handles cold start, so scale immediately on first request
- Reduces latency for /wake command and manual requests
2026-04-22 17:22:22 +00:00
Sirius Claw
240620c3a4
feat: add KEDA scaling for A100 vLLM
...
- Scale to 0 when idle (saves money)
- 15 min cooldown before scale-down
- Triggers on 3+ pending HTTP requests
- Targets openclaw-brain-vllm deployment
2026-04-22 17:07:05 +00:00
Sirius Claw
cecd14123e
chore: optimize vLLM parameters for A100 by removing artificial limits
2026-04-20 02:09:15 +00:00
sirius0xdev
46f576dcda
dial in vllm
2026-04-20 02:04:04 +00:00
sirius0xdev
7c72d229ac
switch back claw image
2026-04-20 01:52:50 +00:00
sirius0xdev
0dbf223df0
fix vllm config
2026-04-20 01:49:36 +00:00
sirius0xdev
33495a40d0
arch_claw image
2026-04-19 23:43:50 +00:00
sirius0xdev
ee3cf7c7ee
dial in vllm settings
2026-04-19 23:41:15 +00:00
sirius0xdev
28e701e27f
add tools to vllm
2026-04-19 23:33:43 +00:00
sirius0xdev
2d8e4b9476
fix config map
2026-04-19 22:32:22 +00:00
sirius0xdev
14551d44e3
reconfigure openclaw
2026-04-19 22:18:59 +00:00
sirius0xdev
283e7e92e1
configmap
2026-04-19 07:21:58 +00:00
sirius0xdev
f0a3d3b179
fix oc crash
2026-04-19 07:01:07 +00:00
sirius0xdev
1732e32a78
fix configmap
2026-04-19 06:51:09 +00:00
sirius0xdev
64ebcaba60
fix typo
2026-04-19 06:45:33 +00:00
sirius0xdev
e40a804d85
fix gateway
2026-04-19 06:41:34 +00:00
sirius0xdev
5139ece7ec
fix configmap
2026-04-19 05:58:32 +00:00
sirius0xdev
43c0a04b18
deactivate analyst
2026-04-19 05:53:21 +00:00
sirius0xdev
5ecd7a66be
fix config map
2026-04-19 05:51:42 +00:00
sirius0xdev
98da1db5fd
Merge branch 'master' of github.com:sirius0xdev/gcloud-lab
...
Call me
2026-04-19 05:47:02 +00:00
sirius0xdev
4f3093d0f8
fix openclaw.json
2026-04-19 05:46:49 +00:00
sirius0xdev
e17a96f43d
Update configmap.yaml
2026-04-18 20:29:16 -04:00
sirius0xdev
7784b786e6
Merge pull request #24 from sirius0xdev/feat/qwen35-27b-config
...
feat: route OpenClaw and Proxy Bot to Qwen3.5-27B-heretic
2026-04-18 19:06:42 -04:00
SiriusClaw
9b15b09eed
feat: sync openclaw and proxy bot configs to use new qwen3.5-27b-heretic model
2026-04-18 22:48:48 +00:00
sirius0xdev
d3d18ab28a
model switch
2026-04-18 22:29:44 +00:00
sirius0xdev
8aad1555a5
try new model
2026-04-18 22:25:26 +00:00
sirius0xdev
8ec068c841
vllm errors
2026-04-18 22:12:29 +00:00
sirius0xdev
3a924b4a10
fix vllm args
2026-04-18 22:08:56 +00:00
sirius0xdev
2fbd6bdf81
change resource limits
2026-04-18 22:04:41 +00:00
sirius0xdev
741a07bbc4
vllm errors models size might be too big
2026-04-18 21:58:39 +00:00
sirius0xdev
197212aacb
fix vllm errors
2026-04-18 21:53:00 +00:00
sirius0xdev
755bdb39ba
Merge pull request #22 from sirius0xdev/fix/vllm-gemma-crash
...
fix: enforce eager execution to stop gemma CUDA graph crash
2026-04-18 17:33:29 -04:00
SiriusClaw
1ebab4f6e5
fix: enforce eager execution to stop gemma 2 CUDA graph crash
2026-04-18 21:32:03 +00:00
sirius0xdev
4a6b30101b
Merge pull request #21 from sirius0xdev/feat/supergemma-26b-abliterated
...
feat: switch brain to supergemma4-26b-abliterated-multimodal
2026-04-18 17:19:03 -04:00
SiriusClaw
603465033b
feat: switch infrastructure to supergemma4-26b-abliterated-multimodal
2026-04-18 21:17:29 +00:00
sirius0xdev
b4f9099d19
Merge pull request #19 from sirius0xdev/fix/vllm-xformers-bug
...
fix: remove XFORMERS override
2026-04-18 16:54:03 -04:00
SiriusClaw
1cc2d8e112
fix: remove XFORMERS attention backend override to fix Venice gibberish
2026-04-18 20:52:26 +00:00
sirius0xdev
eac5d21894
Update kustomization.yaml
2026-04-18 16:37:48 -04:00
sirius0xdev
73a4ee6408
Merge pull request #18 from sirius0xdev/fix/vllm-trust-remote-code-tokenizer
...
fix: revert tokenizer mode to auto
2026-04-18 16:35:51 -04:00
SiriusClaw
6df1e12ab0
fix: revert tokenizer mode to auto
2026-04-18 20:34:15 +00:00
sirius0xdev
518127c711
Merge pull request #17 from sirius0xdev/fix/vllm-mistral-regex
...
fix: enforce mistral tokenizer mode
2026-04-18 16:11:01 -04:00
sirius0xdev
de475b3073
Merge pull request #16 from sirius0xdev/feat/vllm-pvc-cache
...
feat: attach persistent volume claim to vLLM to cache weights
2026-04-18 16:10:48 -04:00
SiriusClaw
0f5d0b17e5
fix: enforce mistral tokenizer mode to bypass regex corruption bug
2026-04-18 20:09:35 +00:00
SiriusClaw
9b7fc44ab2
feat: attach persistent volume claim to vLLM to cache HuggingFace weights
2026-04-18 20:07:30 +00:00
sirius0xdev
f2472307cf
Merge pull request #15 from sirius0xdev/fix/vllm-dtype-auto
...
fix: set vLLM dtype to auto to stop weight corruption
2026-04-18 15:57:11 -04:00
SiriusClaw
576ba503b0
fix: set vLLM dtype to auto to resolve weight casting corruption
2026-04-18 19:55:55 +00:00
sirius0xdev
bdf2e1d1a4
Merge pull request #14 from sirius0xdev/fix/vllm-trust-remote-code-2
...
fix: adjust memory parameters to stop gibberish output
2026-04-18 15:40:22 -04:00
SiriusClaw
7826ff0398
fix: adjust max-model-len and memory utilization for Venice
2026-04-18 19:35:20 +00:00
sirius0xdev
aa40f87885
Merge pull request #13 from sirius0xdev/fix/vllm-trust-remote-code
...
fix: add trust-remote-code to vLLM args
2026-04-18 15:08:07 -04:00
SiriusClaw
bd6416e6a6
fix: add trust-remote-code to vLLM args for mistral architecture
2026-04-18 19:01:53 +00:00
