Debug report uploaded:
Report https://dpaste.com/CBPLWUM95
agent.log https://dpaste.com/FEMD7N3BS
gateway.log https://dpaste.com/8DK96KG7D
#MiMo v2.5 issue
inner roost
frosty oar
hermes config set model.max_tokens 131072 for that model
The first failure is MiMo rejecting the output cap Hermes sent:
max_tokens is too large: 262144. This model supports at most 131072 completion tokens, whereas you provided 262144.
Check your config for model.max_tokens. The compact debug report does not print that field, so it is not visible in the paste.
Run:
hermes config show
If model.max_tokens is 262144, lower it to the provider’s stated cap:
hermes config set model.max_tokens 131072
Then restart Hermes and retry.
The later Context length exceeded: 5,673 tokens. Cannot compress further. message is a confusing follow-on recovery message. The prompt is not actually too large; the first problem is the requested output cap.