scriptoriumadmin/open-webui
1
0
Fork 0
mirror of https://github.com/open-webui/open-webui.git synced 2026-11-07 08:17:07 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

fix: keep the selected model in sync when regenerating a response

Browse source 
Switching the model in the dropdown and regenerating a response left the chat's stored model list pointing at the model that was used when the user message was first sent. Leaving the chat and coming back restored that stale entry, so the dropdown kept snapping back to the original model (and could even show a model that is no longer used by any response in the chat, or one that no longer exists).

The chat's model list is persisted from the user message's `models` field, which is written when the message is created and was never refreshed afterwards. Regeneration is the one send path that reuses an existing user message, so it now refreshes that field from the current selection.

Side-by-side turns are left untouched: those regenerate the clicked column and keep their full model list, which is also what the column grouping in MultiResponseMessages relies on.

Fixes #25052
This commit is contained in:
Classic298 2026-07-28 23:10:33 +02:00
parent 29541cbb52
commit 15d09e21ca
1 changed files with 7 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
src/lib/components/chat/Chat.svelte
View file

@ -3554,6 +3554,12 @@
scrollToBottom();
}
// Single-model turns follow the current selection; side-by-side turns regenerate the clicked column.
const isSideBySideTurn = (userMessage.models ?? selectedModels).length > 1;
if (!isSideBySideTurn) {
userMessage.models = selectedModels;
}
await sendMessage(history, userMessage.id, {
...(suggestionPrompt
? {
@ -3561,9 +3567,8 @@
regenerationPrompt: suggestionPrompt
}
: {}),
...((userMessage?.models ?? [...selectedModels]).length > 1
...(isSideBySideTurn
? {
// If multiple models are selected, use the model from the message
modelId: message.model,
modelIdx: message.modelIdx
}