Hey <@U0BME18CF7S> How are you doing?
# automations-debug
m
Hey @fresh-machine-37269 How are you doing?
👀 1
f
🤖 On it! View progress here: oss-agent-canvas.ngrok.dev/conversations/…
✅ Done! Doing great, thanks for asking! 😊 I'm here and ready to help — what can I do for you today?
c
@fresh-machine-37269 Check your VM and tell me, what llm profile is currently applied in the Github automation? Maybe look at a completed run from yesterday , its conversation base state json
👀 1
f
✅ Done! The currently applied LLM profile is
glm
, resolving to model
openhands/glm-5.2
. I confirmed this two ways: 1. From a completed run's base state JSON (yesterday, Aug 23) — e.g. conversation
1bc4ebde-…
(the "Fix browser SingletonLock collisions" PR review run) shows
agent.llm.model = "openhands/glm-5.2"
. Several other Aug 23 runs all show the same model. 2. From the profile config —
agent-profiles/default.json
(id
ff4c8b19-…
, which is the active agent profile) has
llm_profile_ref: "glm"
, and
profiles/glm.json
(plus
kimi-k3.json
) both define
"model": "openhands/glm-5.2"
. Note: the automations table itself has
model: None
for the "Weekly High-Impact Regression Hunt" entry, so it inherits the workspace-level active profile rather than overriding it — which is why it runs on
glm-5.2
.