Repository navigation
0.12.1 - #1465
Merged
Merged
0.12.1#1465
Conversation
Upgrading to v0.12.0 copies existing Milvus data once so it can use native hybrid search. The new maintenance page explains what happens during that first start, how long it takes on real hardware and why a restart midway starts it over. It also includes a script to follow the progress, and the Milvus warning in the environment reference now links to it.
The old descriptions did not say which of the two settings decides how many chunks reach the model. Top K is the number of chunks retrieved, and with hybrid search on those are the candidates the reranker scores. Top K Reranker is how many of them the model gets, so the context budget tip now points to it too.
The Pipelines pages, their images and the overview mentions are gone, and every link that pointed there now goes to the matching Functions page or is dropped. The pattern for rendering tool runs of self-contained agents moved into the Pipe Function page, since that is where readers need it.
The Langfuse guide used the separate Pipelines server, which the docs no longer cover. It now ships a tested Filter Function that admins add under Functions and that sends every chat reply to Langfuse with its input, output, model, token usage, user and chat. The page also explains why the Langfuse SDK stays on version 3 and when Open WebUI needs a restart after changing the keys.
Tool Search has the model call deferred tools that are not in the tool list sent to the provider. Backends that only allow calls to listed tools, such as the llama.cpp server, can never make those calls, so the docs now say so and name the two settings that work there.
Testing against real providers showed that the OpenAI API, on both Chat Completions and Responses, and xAI also never call tools missing from the tool list, the same as the llama.cpp server. The Tool Search backend note now names them.
The Tool Search backend note now also lists the providers where the model does call deferred tools: Anthropic, Google Gemini, Mistral, Alibaba (Qwen) and CoreWeave (Llama 3.3). Through OpenRouter the result follows the provider the request is routed to.
Adds screenshots of two-factor sign-in (the admin switch, the authenticator setup and the recovery codes), signing a user out of every device, nested groups and group default models, replying in shared chats, the knowledge base file browser, model and prompt version history and the Tool Search settings, next to the sections that describe them.
Redis Sentinel now treats the master name as case sensitive and passes the URL's credentials to the Sentinels, which the env reference and the Redis guide now explain. Voice avatars work in Standard calls too, so they got their own section on the voice page. The default follow-up prompt, automation timing, translation deletes, citation links for external knowledge, folder links and MCP connection logging are updated where readers look for them.
Admins can now connect external apps that answer mentions and button clicks in channels. A new page covers adding and using apps, installing and configuring them, private network setups and troubleshooting, and a second page explains how to build one with the oix SDK. The env reference, database schema, hardening guide, security policy and Channels page now point to them.
The reserved arguments page said the files entry in the request metadata is an empty list when nothing is attached. It is None in that case, which pipes and filters reading it directly need to handle.
Skills now give the model a preview and a heading index for very long instructions, read files in smaller pieces and can search a skill's files with the new grep tool, so the Skills and Tools pages describe the new limits and when these tools are offered. Smaller updates cover Generate Message Pair, calendar alerts after moving an event, editing other people's messages in shared chats and model versions that use notes as knowledge.
Chats in shared folders now show the pictures of the people who wrote in them, and hovering a folder shows a short preview with its chats, default model and knowledge. The folders page describes both, with screenshots of the pictures, a profile card and the preview.
Apps now confirm a mention within ten seconds and post their answer as a channel message, and a failed mention is reported in the channel. The Apps pages describe this, the new error messages and the Requested scopes editor in the app settings. The Skills page now explains which symbolic links a skill import skips or refuses.
Channels can now be muted down to mentions, typed @NAMEs and #channels turn into real mentions, and channel webhooks reach members of group channels and direct messages. Skills can be exported as a single .skill file, and skills, prompts, system prompts, chat text and terminal files can be compared against a revision side by side. Analytics, memory in automations, compaction during long tool runs and empty file lists for pipes and tools are updated too.
A connected Realtime call now sends the Voice Mode Prompt to the chat model like a Standard call, including for messages typed during the call. The Prompt Caching page notes that a call started after typed messages rebuilds the cached prefix once on its first spoken turn.
On v0.12.0, people created through SCIM could not sign in through SSO and got a 401. v0.12.1 gives these accounts a sign-in record and adds the missing ones on startup, so the SCIM page, the SSO troubleshooting and the update notes explain the cause, the fix and the backup step. The SCIM page now also says that deleting a user removes the account.
Realtime voice calls can now use Azure OpenAI with an API key or Microsoft Entra ID, and admins can pass extra audio input settings such as noise reduction. The voice page explains the Azure setup and the new settings, and the env reference lists the three new variables. It also covers how calls now handle background noise, the generated titles and tags of call chats and the avatar controls.
Deploying with
|
| Status | Name | Latest Commit | Updated (UTC) |
|---|---|---|---|
| ✅ Deployment successful! View logs |
open-webui-docs-search | 31e3154 | Oct 14 2026, 05:31 PM |
Six audits checked the 0.12.1 docs against the code and against what a first-time reader needs. This corrects the long skill preview, the compare and apply steps, channel notification and webhook conditions, the Azure Realtime credential setup and the env reference, and fixes stale wording on the pages that still described older behaviour. A broken database recovery command and two duplicate env entries are fixed too.
The hardening guide said community sharing controls chat share links, but it only controls sharing to and browsing the Open WebUI Community. It also recommended rolling updates for several replicas, which are unsupported across the schema changes in the 0.12 releases, so it now says to update every instance at once.
Since v0.12.1 the chat model is also offered the avatar gesture tool during Realtime calls. Only the voice model plays gestures there, so the chat model's calls come back unavailable, and the Tools and voice pages now say so.
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.