Skip to content

Commit 1b83e36

Browse files
committed
Merge remote-tracking branch 'origin/v3' into v3-ai
2 parents 1c97210 + 1cbf935 commit 1b83e36

2 files changed

Lines changed: 34 additions & 0 deletions

File tree

deploy/env-uzh-prd/values.yaml

Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -205,6 +205,23 @@ chat:
205205
fallbackId: gpt-4.1-mini
206206

207207
modelRegistry:
208+
# "Auto" routes through the LiteLLM complexity-router, which self-selects the
209+
# tier per request (SIMPLE=gpt-4.1, MEDIUM=gpt-5.4-low, COMPLEX=gpt-5.4-medium,
210+
# REASONING=gpt-5.5-low). It is a normal selectable model (fallback: false), so
211+
# it does NOT change automatic selection for existing chatbots — they still
212+
# default to gpt-5.5. A chatbot can be pinned to Auto-only via its DB columns
213+
# (modelSelection=false + allowedModelIds=["auto"]).
214+
# cost mirrors the gpt-5.4 tier (the typical COMPLEX route); REASONING (gpt-5.5)
215+
# is rare. Tune if the routed mix shifts.
216+
- id: auto
217+
deploymentId: klickeruzh/azure/complexity-router
218+
name: Auto
219+
description: Automatic model selection (routes to the best model per request)
220+
supportsImageAttachments: true
221+
fallback: false
222+
cost:
223+
input: 1.25
224+
output: 10.0
208225
- id: gpt-4.1
209226
deploymentId: klickeruzh/azure/gpt-4.1
210227
name: GPT-4.1

deploy/env-uzh-stg/values.yaml

Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -173,6 +173,23 @@ chat:
173173
fallbackId: gpt-4.1-mini
174174

175175
modelRegistry:
176+
# "Auto" routes through the LiteLLM complexity-router, which self-selects the
177+
# tier per request (SIMPLE=gpt-4.1, MEDIUM=gpt-5.4-low, COMPLEX=gpt-5.4-medium,
178+
# REASONING=gpt-5.5-low). It is a normal selectable model (fallback: false), so
179+
# it does NOT change automatic selection for existing chatbots — they still
180+
# default to gpt-5.5. A chatbot can be pinned to Auto-only via its DB columns
181+
# (modelSelection=false + allowedModelIds=["auto"]); see the vorkurs chatbot.
182+
# cost mirrors the gpt-5.4 tier (the typical COMPLEX route); REASONING (gpt-5.5)
183+
# is rare. Tune if the routed mix shifts.
184+
- id: auto
185+
deploymentId: klickeruzh/azure/complexity-router
186+
name: Auto
187+
description: Automatic model selection (routes to the best model per request)
188+
supportsImageAttachments: true
189+
fallback: false
190+
cost:
191+
input: 1.25
192+
output: 10.0
176193
- id: gpt-4.1
177194
deploymentId: klickeruzh/azure/gpt-4.1
178195
name: GPT-4.1

0 commit comments

Comments
 (0)