fix(container): update image ghcr.io/open-webui/open-webui (v0.13.0 ➔ v0.13.3) #2610

Open
renovate-bot wants to merge 1 commit from renovate/ghcr.io-open-webui-open-webui-0.x into main
Collaborator

This PR contains the following updates:

Package Update Change
ghcr.io/open-webui/open-webui patch v0.13.0 → v0.13.3

⚠️ Warning

Some dependencies could not be looked up. Check the Dependency Dashboard for more information.


Release Notes

open-webui/open-webui (ghcr.io/open-webui/open-webui)

v0.13.3

Compare Source

Added
  • 🦾 Agents in automations. The automation editor now lets you choose a terminal and working folder, then select a coding agent available there. Scheduled runs use that saved terminal and folder, and check that the terminal and agent are still available before starting. Agent automations send their results to a chat; channel destinations remain available for ordinary models. Commit
  • 📂 Change terminal and folder during a chat. You can now choose a different terminal or working folder after sending messages, including in temporary chats. Both controls remain disabled while a reply is being generated. Commit
  • 🇧🇷 Brazilian Portuguese translations. Added missing translations for terminal and folder controls, Git actions, spreadsheet editing, voice notes, tab shortcuts, notification sounds and tool settings. Updated the explanations for document retrieval, tool permissions and voice settings. #​33945
Fixed
  • 📂 Files tabs follow terminal changes. Switching a chat to another terminal now updates its open Files tabs to show that terminal's files, instead of leaving the previous terminal's folders on screen. With a terminal set to "Per chat", the Files tab also switches to the chat's own files after the first message. #​33954
  • 🧭 Agents in unsupported model pickers. Connecting a terminal no longer adds its coding agents to Admin Models, base-model selectors, Arena candidates, task-model settings, group defaults, sub-agent settings, channel mentions or the Playground. Notes also stop choosing a terminal agent for title generation. Agents cannot be saved or imported as custom models or their base models, and their chat menu no longer offers Edit. Commit, Commit
  • 🔌 Agent list after switching terminals. Changing terminals immediately removes the previous terminal's agents from the chat's model list while the new list loads. A model-list request from a chat you have left can no longer replace the current list. Commit
  • 📋 Copy buttons over plain HTTP. At a plain HTTP address on your local network, copying comment links, search links, shared-chat links, webhook URLs, chat IDs, Git branch names, commit hashes and spreadsheet cells now works. Copy Formatted Text preserves formatting, and copied text keeps its final line break. #​33939
  • 📝 Clipboard prompts and Auto-Copy Response over plain HTTP. Prompts using the clipboard placeholder now insert into chats and channels even when the browser blocks clipboard access, with a message explaining that the clipboard could not be read. Auto-Copy Response stays off and shows a permission message instead of silently failing to turn on; turning it off no longer requires clipboard access. #​33939
  • 🧠 DeepSeek thinking mode with skills, knowledge and memories. These chats no longer fail with a missing reasoning error when the connection's Provider is set to "llama.cpp". DeepSeek thinking mode still requires that setting; connections using "Default" are unchanged. #​33950
  • 🏷️ Labels on grouped tool calls. With Tool Call Labels enabled, a group of tool calls keeps the labels that are present even if one call has none. The unlabeled call shows its tool name instead of making the whole group lose its labels. #​33932
  • 💬 Tool labels in later replies. The model now sees the labels it wrote for earlier tool calls when continuing a reply or answering another message, removing a reason it could stop writing labels as the conversation continued. #​33953
  • 🔎 Knowledge searches containing punctuation. Searches in knowledge files now find text such as "Art. 40" when the model marks the period as literal punctuation, instead of returning no matches. #​33936
  • 🪟 Side Chat with Temporary Chat disabled. Leaving Enforce Temporary Chat switched on while disabling Temporary Chat no longer prevents users from sending and saving side chats or dragging an existing chat into the side panel. #​33930
  • 🏠 Home folder tab title. A Files tab at the connected machine's home directory now says "Home" instead of showing the directory name, including when reopening saved tabs. Other folders keep their names. Commit
  • 🦙 Missing Ollama in Docker images. Building the image that includes Ollama now fails if Ollama was not installed, instead of producing an image that cannot start Ollama after its download failed. #​33944
  • 📄 Errors after leaving a chat. Closing or leaving a chat while its tab layout is being saved no longer fills the server log with an error report for the dropped connection. #​33929

v0.13.2

Compare Source

Fixed
  • 🔓 Chat over plain HTTP. Opening Open WebUI over plain HTTP at a LAN or Docker address now loads new and existing chats instead of staying on a spinner with "crypto.randomUUID is not a function". #​33919
  • 📝 Comments and Markdown editing over plain HTTP. Creating, replying to and editing comments, and editing Markdown files containing rendered blocks such as Mermaid diagrams or callouts, no longer fail with the same UUID error. Commit
  • 🎛️ Side Chat permission switch. The Allow Side Chat switch is now visible even when Temporary Chat is off, so administrators can configure the two permissions independently. #​33920

v0.13.1

Compare Source

Added
  • 🦾 Coding agents as models. On an Open Terminal that offers agents, coding agents such as Claude Code, Codex and Pi now appear in the model picker. You can chat with them or call them by voice, choose their reasoning effort and service tier, and approve their tool requests in the chat. Each conversation uses one agent, and agent chats cannot be regenerated, continued or edited and resent. If an agent's terminal goes away, its chats and new chats fall back to the default model and earlier replies keep the agent's name. Autocompletion in agent chats uses the task model, or is skipped when there is none. #​33843, #​33860, #​33880, #​33857
  • 🪢 Introducing Tether. Tether is the new desktop companion app for macOS, Windows and Linux that ties your own computer to Open WebUI. Sign in once and your computer shows up under Your devices in the terminal menu, so models, coding agents and automations can work in your real files, repositories and tools, and you can reach them from any browser, including your phone. Tether connects out to Open WebUI, so there are no ports to open, its terminal output resumes after a dropped connection, and it delivers notifications to your desktop even when no Open WebUI page is open. It signs in with a renewable session that expires after 90 days without use by default, set with "SESSION_EXPIRES_IN".
  • 💸 Much cheaper tool chats. Chats that use tools now cost much less. Citation instructions, memory recall, the skills list and attached knowledge are no longer rewritten into the system prompt or your message on every tool round, so the start of the conversation stays the same between requests and providers can serve it from their prompt cache. Custom pipes see a different message layout as a result; see Context as hidden tool results under Changed. #​33852
  • 🚦 Auto tool permissions. Administrators can now turn on Auto, a third tool permission mode next to Full access and Ask for approval (off by default). A review model checks each tool call and runs it right away only when it judges that no confirmation is needed. The review sends the tool, its arguments and your last few messages to the configured scoring API, and any call it cannot review, because of an error, a timeout or too much context, asks for approval.
  • 📚 Full text for small files. Files whose extracted text is shorter than a new threshold (50,000 characters by default, set with "RAG_FULL_TEXT_CHAR_THRESHOLD" or in the admin Documents settings) are now included in full instead of through retrieval. This applies both when Preload files adds the content to the conversation and when the model reads an attached file with its file search tool, and the file viewer shows when it happens.
  • ⏱️ Faster large chats. Large chats now open about twice as fast, and marking a chat as read or unread or removing a tag from it is 4 to 6 times faster. #​33665, #​33666
  • 🏎️ Faster chat access checks. Sending messages, typing indicators, shared links, comments and file access no longer load the whole conversation just to check who may access it, which helps most on large chats. #​33667, #​33669
  • 🔧 More forgiving tool calls. Tool calls with arguments wrapped in code fences, encoded twice or sent as text instead of numbers and booleans now run, and tool calls a model writes out as plain text markup are recognised and executed. #​33868
  • 🛑 Tool loop stops. A tool call that fails twice in a row with the same arguments now ends the tool loop so the model writes a final answer, and a reply cut off by the token limit no longer runs its unfinished tool calls.
  • 🛠️ Default tools for models. The Model Defaults section in the admin Models settings now has a Tools picker, models without their own tool list start new chats with those tools enabled, and the new model form preselects them. #​33749, #​33787
  • 🚀 Faster channels. The channel list, the channel messages and the replied-to messages in a channel now load much faster, especially with many channels or many replies, and new database indexes speed up channel threads and reactions. #​33690, #​33691, #​33688
  • 🪶 Lighter reasoning streams. Long reasoning replies from providers such as OpenRouter now send far less data to the browser and to Redis while they stream. #​33698
  • 🗂️ Faster sidebar chat list. The chat list in the sidebar now appears as soon as its own data is ready instead of waiting for the channel list. #​33686
  • ⌨️ Faster typing after large pastes. Typing in the chat input after pasting 16,000 lines of text is now about 117 times faster, down from 8.6 seconds to 74 milliseconds per keystroke. #​33729, #​12087
  • 🧮 Faster Redis filter lookups. With Redis enabled, chat requests now look up the filters of a model faster. #​33668
  • 🪟 Split tab panes. On large screens, dragging a tab or a chat to the edge of the workspace, or pressing the new Split screen button, splits it into resizable panes, so several tabs, including the chat, stay visible side by side. Tabs can also be reordered and moved between panes with a long press and drag on touch screens. Commit
  • 🧳 Tabs remembered per chat. The open tabs, split layout and selected terminal of each chat are now saved privately for each user on the server and restored when you reopen the chat, even on another device. Signing out warns you about unsaved file edits. #​33899, #​33885
  • 🪣 Batched Tool Search calls. With Tool Search on, the model can run several independent tools it found in a single call_tool request, which saves model turns and so lowers the cost of tool-heavy chats.
  • 🧭 Tool Search engines. Tool Search in the admin Interface settings can now match tools with embeddings from the Documents embedding model (kept in memory, not stored) or by sending the query and tool definitions to a relevance-scoring API, besides the default keyword search.
  • 🏷️ Readable tool call labels. Administrators can now turn on Tool Call Labels in Admin Settings > Interface or with "ENABLE_TOOL_CALL_LABELS", off by default, to ask the model for short descriptions in the user's language of what each tool call is doing and what it did. Commit
  • ✂️ Chat read limit. Admins can limit how much of an earlier chat the Chat History tools return, with "VIEW_CHAT_MAX_CHARS" or the new Chat Read Limit field in Admin Settings > General, and the newest messages are kept while older ones are cut. #​33695, #​33608
  • 🌍 Client IP header placeholders. Custom headers of connections can now use "{{USER_IP}}" and "{{X_FORWARDED_FOR}}" to pass the client address on to the upstream provider, and both are empty for automations and background jobs. #​33833, #​33819
  • ✉️ Non-ASCII email addresses. The sign-up page and the add and edit user dialogs now accept email addresses with non-ASCII characters. #​33693, #​33681
  • 🕒 Reply finish time. Assistant replies now show the time they finished instead of the time the prompt was sent. #​33895
  • 🧐 Channel message search. Each channel header has a Search button that searches the messages of that channel.
  • 🔔 Notifications open their chat. Clicking a desktop notification now brings Open WebUI to the front and opens the related chat, channel or calendar.
  • 📮 MCP Apps message the chat. An MCP App can now send a text message to the chat, which you confirm first in a Confirm Prompt from Embed dialog. #​33907
  • 🖱️ Drag chats into side panel. Dragging a chat from the sidebar onto the chat area opens it as a side chat tab, and a side chat can now also be started from an empty new chat.
  • 🪄 Quick open files. A new shortcut, Cmd+P (Ctrl+P elsewhere), opens a search box that finds files in the selected terminal by name or path and opens the one you pick in a tab.
  • 🎹 Tab keyboard shortcuts. New shortcuts, which can be changed in Settings > Keyboard, open the Files, Desktop, Comments, new chat, new file, new terminal and new browser tabs, close and switch between tabs and toggle the split screen.
  • 🗺️ Working directory picker. The message input has a working directory menu with folder search, a Recently Used list kept per user on the terminal and an Open Folder dialog.
  • 📂 Folder default terminal. A folder can now have a default terminal and working directory in its Advanced settings, used for new chats in that folder.
  • 🧾 Editable spreadsheets. Excel files in the file view can now be edited in the preview, with a formula bar, pasting from other spreadsheets and resizable columns, and saved back to the file, while formula cells and protected sheets stay read-only.
  • 🖋️ Markdown and code editing. Markdown files in the file view now open in a rich text editor that saves back to Markdown, where pasted or dropped images are uploaded to the terminal, and the code editor suggests words that already appear in the file. Other text files and the Source view of previews now use the code editor directly, without a separate Edit step. Commit
  • 🌳 Clone Git repositories. The file browser and the folder pickers can now clone a Git repository from an HTTP(S) or SSH URL into the chosen folder of the selected terminal.
  • 🧬 Git changes in the editor. The code editor now marks the lines of a file that were added or changed since the last commit.
  • 🔖 File browser Git status. Files and folders in the file browser are marked with a Git status letter such as M, A, D or U, and folders show that they contain changes.
  • 🌿 Git panel feedback and resizing. The Git panel now shows what it is doing, such as staging, pushing or fetching, and its file list can be resized.
  • 🧹 Discard any Git change. The Git panel can now discard staged and new files too, and its confirmation warns that files not in the last commit are deleted and that this cannot be undone.
  • 🧲 Drag files into chat. Files from the terminal file browser can be dragged, one or several at once, into the message input to attach them, also with a long press on touch screens.
  • 🗃️ Multiple Files tabs. The + menu can now open several Files tabs for the same terminal, each with its own folder and the folder name as its tab title, and moving files refreshes the other Files tabs on that terminal. Commit
  • 🔲 Drag to select files. Several files in the file browser can now be selected by dragging a selection box over them.
  • 🪂 Spring-loaded Files tabs. Holding a dragged file over a Files tab for half a second switches to that tab, so files can be moved into folders of another tab.
  • ⚡ Faster terminal rendering. The terminal tab now renders with WebGL where the browser supports it and takes in bursts of output in larger batches, so heavy output is handled faster.
  • 🩺 Terminal system info. The terminal menu now has a System info view with the terminal's CPU, memory, disk, network, uptime and top processes.
  • 🖌️ Edit images from the terminal. The image edit tool can now take source images from the selected terminal by file path, as PNG, JPEG, WebP or GIF up to 25 megapixels. #​33824
  • 🎙️ Voice notes in files. A new shortcut, Cmd+Alt+M (Ctrl+Alt+M elsewhere), records a voice note, can transcribe it, and saves the recording with a Markdown note in the current folder of the selected terminal, then opens the note in a tab. Escape cancels a recording. #​33902, #​33903
  • 📢 Model availability events. Open WebUI now publishes model.available and model.unavailable events when a model appears in or disappears from the fetched model list.
  • 🗒️ Descriptions in model pickers. The tool, skill, filter, action and sub-agent pickers in the model editor now show each item's description under its name. #​33723
  • 🔢 Grouped token counts. Whole numbers in the token usage tooltip are now grouped by thousands for the interface language, for example 26,678,623. #​33896
  • 🎼 Notification sound styles. The notification sound setting now offers Chime, Soft, Bell and Tap sounds, with a play button to preview each one.
  • 🔄 General improvements. Various improvements were implemented across the application to enhance performance, stability, and security.
  • 🌐 Translation updates. Translations for Portuguese (Brazil), Azerbaijani, Basque, Bengali, Bosnian, Bulgarian, Cebuano, Croatian, Czech, Danish, Dutch, Estonian, Faroese, Filipino, Galician, Georgian, German, Greek, Hebrew, Hindi, Hungarian, Indonesian, Irish, Japanese, Kabyle, Korean, Lithuanian, Malay, Norwegian Bokmål, Persian, Polish, Punjabi, Romanian, Russian, Serbian, Slovak, Slovenian, Spanish (Spain), Swedish, Tamil, Thai, Tibetan, Turkish, Turkmen, Ukrainian, Urdu, Uyghur and Vietnamese were enhanced and expanded.
Fixed
  • 🛡️ Security Advisory: This release contains security fixes. We strongly recommend updating all deployments as soon as possible to have the latest security fixes in place. These fixes are not listed in this changelog; we will publish the security advisories for them immediately following CVE assignment. Advisories
  • 🗄️ Redis Cluster chats. Sending a chat message on a Redis Cluster deployment no longer fails while the personal terminal relay starts. #​33766
  • 🚪 Automatic SSO sign-in. With automatic SSO redirect on, an expired session now signs you back in without a state mismatch error. #​33791, #​33783
  • 🎨 OpenAI image default model. Image generation with the OpenAI engine now defaults to GPT-IMAGE 2 instead of the shut-down DALL·E 2, and the shut-down DALL·E models are no longer offered. #​33694, #​33578
  • 🖼️ Default GPT image requests. OpenAI image generation and editing with no model explicitly set no longer send an unsupported response format to the default GPT image model. #​33875
  • 🔐 Terminal Session authentication. Terminal connections set to Session authentication now accept an API key, so Verify Connection, the policy settings and the terminal tools work again. #​33699, #​30001, #​33579
  • 🦙 Ollama broken tool calls. Chats with Ollama models no longer get stuck after a tool call with cut-off, rejected or empty arguments. #​33697, #​33671
  • 🧠 Complete compaction summaries. Context compaction summaries are no longer cut off at 1000 tokens, which could leave reasoning models with an empty or truncated summary. #​33726
  • 💾 Prompt cache on tool results. Tool results sent through OpenAI-compatible gateways now keep their prompt cache marker, so Anthropic models behind OpenRouter or LiteLLM can cache the prefix in tool loops. #​33757, #​33755
  • 📖 File read size limit. Reading a file by line range now stops at the same size limit as other reads, so one read can no longer overflow the context window of the model. #​33696, #​33607
  • 🧰 Tool Search retries. A malformed tool call with Tool Search on no longer disables tools for the rest of the reply, so the model can correct its arguments and try again. #​33872
  • 🧩 Split tool calls. When a model sends several tool calls glued into one, the chat no longer shows an extra tool call that never gets a result. #​33779, #​33778
  • 🔎 External knowledge base queries. External knowledge bases now search with every generated query at the same time instead of using only the results of the first one. #​33704, #​33184
  • 📊 Large spreadsheet uploads. Uploading a large, sparsely filled spreadsheet no longer takes minutes and stalls the server. #​33770, #​33768
  • 🧱 Milvus knowledge base uploads. Adding a file to a knowledge base on Milvus no longer processes the file a second time or stores its chunks twice. #​33745, #​33744
  • 📄 MinerU 4.1 servers. The MinerU document loader in local mode now works with MinerU 4.1 servers, where every document used to fail. #​33719, #​33718
  • 🔑 MCP OAuth stays signed in. The login to an MCP tool server that uses OAuth is no longer lost when the provider has a temporary outage while the token is refreshed. #​33805, #​33803
  • 📍 Location lookup delays. With location access turned on, sending a message no longer waits for a slow or unanswered browser location request.
  • 🗓️ Automation schedule examples. Models creating automations through the built-in tool no longer hit "Invalid RRULE" errors caused by a wrong example schedule format. #​33732, #​33730
  • ♿ Keyboard and screen reader fixes. Keyboard focus now stays where it belongs: on buttons while they save, in place when you cancel a message edit, switch message versions or sort and page a workspace or notes list, on the first and last page buttons, and in the title field while a note title is generated. Closing a channel thread, a note panel or a chat menu returns focus to where it was, and the sign-in skip link moves focus to the form. The show password button is now wide enough to hit, and holding Shift no longer swaps row buttons for Delete while a keyboard user has focus.
  • 🕵️ Audit log secret masking. With request auditing on, the audit log now masks API keys, tokens and secrets saved in the admin settings, not only passwords. #​33746
  • 🔇 Queries kept out of logs. The default server log no longer contains your search queries or the provider address of a failing external reranker. #​33701
  • 💬 Chat message search. Workspace search now finds message text and shows matching excerpts in chats whose messages are stored separately from the chat record. Commit
  • 🔃 Search results stay current. Notes, archived and shared chats, the admin user list and the knowledge and file pickers in the message input no longer let an older search replace the results of a newer one. Commit
  • 👎 Feedback details saved. A thumbs-down reason and comment are no longer lost when saved before the feedback tags finish generating. #​33716
  • 🎒 Side chat default tools. New side chats now start with the same default tools, skills, filters and web search as a normal chat with that model. #​33788
  • ✅ Side chat tool approvals. Tool calls in side chats now show working approval controls and resume the reply after approval. #​33889
  • ❓ Side chat questions. Questions asked through the model's ask_user tool now appear in side chats and can be answered or dismissed, including after reopening the chat. #​33901
  • ⏩ Continued replies on Responses connections. Continuing a reply on a Responses API connection now adds the new text to the existing reply.
  • 🛂 Model required before sending. The file ask box, follow-up suggestions and queued messages now report "Model not selected" instead of trying to send when no model is selected. #​33838
  • 👻 Temporary chat button with tabs. The Temporary Chat button is back on new chats when the tabbed layout is on. #​33680
  • 🪧 Stable MCP app views. MCP apps in chat no longer reload and lose their state while the message around them updates.
  • 🎴 Artifacts keep the chat visible. Opening an HTML or SVG preview from a reply now uses the side pane instead of replacing the visible chat tab. #​33894
  • 📣 Banners in tabbed view. Admin banners and license notices now show on the new chat page when the terminal tabs split the screen. #​33789
  • 📡 Tabs over plain HTTP. The New File, New Terminal and Browser tabs in the side panel now open when Open WebUI is reached over plain HTTP on a LAN or Docker address. #​33847, #​33845
  • 🕶️ Per-chat terminals in temporary chats. Per-chat workspace terminals are no longer offered in temporary chats, where sending to them failed, and one already selected is deselected when Temporary mode is turned on. #​33866, #​33861
  • 🖥️ Terminal kept on model switch. Switching to a model without its own default terminal no longer turns off the terminal selected in the chat.
  • 📁 Working folder in new chats. The terminal working folder of a new chat is no longer reset to the root when the chat is created before the file browser has finished loading.
  • 📞 Return to call. The Return to call button now reopens the active voice call after switching to the chat tab. #​33898
  • 🔕 Channel notifications clear on read. Opening a channel now marks its notifications as read.
  • 🪝 Webhooks window stability. The channel Webhooks window no longer jumps back to the webhook list when a new message arrives, so edits in progress are kept. #​33776, #​33775
  • 🎧 Background audio on iOS. With Auto-Playback Response off, tapping in Open WebUI or opening a chat no longer stops audio playing in other apps on iOS. #​33780
  • 🖨️ Scanned PDF preview. PDF previews no longer show blank pages for scanned PDFs that store their pages as JPEG2000 images. #​33906, #​33905
  • 🔬 Preview zoom stability. Zooming in PDF, Word and PowerPoint previews no longer resets or drifts when you switch tabs, change slides or resize the pane. #​33836, #​33835
  • 📌 File preview from Controls. A file preview opened from the Controls menu no longer closes when you click inside it. #​33831, #​33828
  • 🔭 Tab bar and UI scale. Tab content no longer slides under the tab bar when the UI scale is above 1. #​33830, #​33829
  • 🔘 Desktop app top clicks. In the desktop app, banners and menus near the top of the window now take clicks instead of dragging the window.
  • 📱 Mobile menu keyboard. Closing the + menu on a phone no longer opens the keyboard. #​33893
  • 📝 File rename in terminal. Renaming a file or folder in the Open Terminal file browser no longer shows a false error after the rename succeeded. #​33786
  • 📋 Stable file selection. Selecting files no longer shifts the file rows when the bulk action bar appears, and clicking an empty area of the file list clears the selection. Commit
  • 🫥 Selection popups on scroll. Scrolling the file view dismisses its Ask box, and scrolling a rich text editor dismisses its floating formatting menus until the selection changes. Commit, Commit
  • 🙈 Empty control groups hidden. When every control of a group is hidden with Visible controls, the empty group heading or submenu no longer shows. #​33867
  • 🌄 Chat background behind tabs. The chat tab bar no longer covers the chat background with a solid strip. #​33849
  • 🌫️ Top bar blur. The top bars of Workspace, Admin Panel, Playground and Calendar no longer blur a page background image. #​33799, #​33798
  • 🗑️ Feedback model filter. Deleting the last feedback of a model in the Evaluations feedback list now clears the model filter. #​33794, #​33793
  • 🔤 Visible controls labels. Six toggles under Visible controls in the Admin Interface settings now show their names instead of raw translation keys. #​33734
  • 🤫 Quiet terminal disconnects. Closing the file panel or a terminal view right after opening it no longer logs an error traceback on the server. #​33863, #​33850
  • 📜 Quieter MCP connection logs. Logs no longer fill with a warning for every MCP server a user has not connected yet each time they load Open WebUI. #​33663
  • 🪪 OAuth token refresh warnings. Refreshing OAuth tokens with a client secret no longer logs deprecation warnings on every call. #​33700, #​33297
Changed
  • ⚠️ Database Migrations: This release includes database schema changes; we strongly recommend backing up your database and all associated data before upgrading in production environments. If you are running a multi-worker, multi-server, or load-balanced deployment, all instances must be updated simultaneously, rolling updates are not supported and will cause application failures due to schema incompatibility.
  • 💥 Preload files off by default. With native function calling, the default, attached files are no longer added to the conversation automatically, and models read them with their file tools instead. Models without tool support or without the Files built-in tool can no longer see attached files until you turn on Preload files in the model's Capabilities. Legacy function calling, and API requests that do not come from a chat in the web UI, keep the old behavior. #​33815
  • 🎯 Retrieval defaults. Unless you saved your own values, "RAG_TOP_K" and "RAG_TOP_K_RERANKER" now default to 10 instead of 3, and "CHUNK_SIZE" to 1024 instead of 1000. Each query now retrieves more text, so prompts with knowledge get longer. The new chunk size applies to files processed from now on; existing knowledge keeps its chunks until it is reindexed.
  • 🧷 Context as hidden tool results. With native function calling, memory recall, the skills list and the attached knowledge list now arrive as hidden tool results after your message instead of in the system prompt. This is what keeps tool chats cacheable, and it means custom pipes can see a tool message as the last message.
  • 🗨️ Side chats are saved. Side chats, which were always temporary in 0.13.0, are now saved as regular chats and appear in the chat list. They stay temporary only when you turn on Temporary Chat in the side chat or your admin enforces temporary chats.
  • 🚧 Tool call limit. Reaching the tool call limit now makes the model write a final answer without tools, instead of ending the reply with the "Tool-call limit reached" error.
  • 🔦 Tool Search prompt. With Tool Search on, the model no longer sees a list of the hidden tools and finds them only by searching.
  • ✏️ Always Use Full Text. The Full Context Mode setting in the admin Documents settings is now called Always Use Full Text and includes complete documents regardless of the full text threshold.
  • ↔️ Always full-width tabs. On large screens the tabbed workspace now always fills the chat area, with the chat as one of its tabs, and the full-width toggle added in 0.13.0 is gone.
  • 🔀 Integrations moved into + menu. The Integrations button is now part of the + menu in the message input, and the new Show Input Selections setting, off by default, controls whether selected tools, skills, filters and capabilities are shown beside it.
  • 🔓 Side chat permission. The Allow Side Chat permission no longer requires the Temporary Chat permission.
  • 🕸️ Chrome first in Browser tab. The Browser tab now starts in Chrome mode when the terminal offers it, and switches to Proxy mode with a message if Chrome fails to start.
  • ⛔ Terminal select setting removed. The Show Files on Terminal Select setting has been removed from the Interface settings. #​33855, #​33840

Configuration

📅 Schedule: (in timezone Europe/Rome)

  • Branch creation
    • At any time (no schedule defined)
  • Automerge
    • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about these updates again.


  • If you want to rebase/retry this PR, check this box

This PR has been generated by Mend Renovate CLI.

This PR contains the following updates: | Package | Update | Change | |---|---|---| | [ghcr.io/open-webui/open-webui](https://github.com/open-webui/open-webui) | patch | `v0.13.0` → `v0.13.3` | --- > ⚠️ **Warning** > > Some dependencies could not be looked up. Check the [Dependency Dashboard](issues/6) for more information. --- ### Release Notes <details> <summary>open-webui/open-webui (ghcr.io/open-webui/open-webui)</summary> ### [`v0.13.3`](https://github.com/open-webui/open-webui/releases/tag/v0.13.3) [Compare Source](https://github.com/open-webui/open-webui/compare/v0.13.2...v0.13.3) ##### Added - 🦾 **Agents in automations.** The automation editor now lets you choose a terminal and working folder, then select a coding agent available there. Scheduled runs use that saved terminal and folder, and check that the terminal and agent are still available before starting. Agent automations send their results to a chat; channel destinations remain available for ordinary models. [Commit](https://github.com/open-webui/open-webui/commit/21fc016141c73a0e11b428c61bde20527bbc228a) - 📂 **Change terminal and folder during a chat.** You can now choose a different terminal or working folder after sending messages, including in temporary chats. Both controls remain disabled while a reply is being generated. [Commit](https://github.com/open-webui/open-webui/commit/a518ff1630b74ed6edd62538c5da4bb26cc6dd20) - 🇧🇷 **Brazilian Portuguese translations.** Added missing translations for terminal and folder controls, Git actions, spreadsheet editing, voice notes, tab shortcuts, notification sounds and tool settings. Updated the explanations for document retrieval, tool permissions and voice settings. [#&#8203;33945](https://github.com/open-webui/open-webui/pull/33945) ##### Fixed - 📂 **Files tabs follow terminal changes.** Switching a chat to another terminal now updates its open Files tabs to show that terminal's files, instead of leaving the previous terminal's folders on screen. With a terminal set to "Per chat", the Files tab also switches to the chat's own files after the first message. [#&#8203;33954](https://github.com/open-webui/open-webui/pull/33954) - 🧭 **Agents in unsupported model pickers.** Connecting a terminal no longer adds its coding agents to Admin Models, base-model selectors, Arena candidates, task-model settings, group defaults, sub-agent settings, channel mentions or the Playground. Notes also stop choosing a terminal agent for title generation. Agents cannot be saved or imported as custom models or their base models, and their chat menu no longer offers Edit. [Commit](https://github.com/open-webui/open-webui/commit/21fc016141c73a0e11b428c61bde20527bbc228a), [Commit](https://github.com/open-webui/open-webui/commit/6d3123b5b4a90962bff44c69c7e0bae0c65557d5) - 🔌 **Agent list after switching terminals.** Changing terminals immediately removes the previous terminal's agents from the chat's model list while the new list loads. A model-list request from a chat you have left can no longer replace the current list. [Commit](https://github.com/open-webui/open-webui/commit/21fc016141c73a0e11b428c61bde20527bbc228a) - 📋 **Copy buttons over plain HTTP.** At a plain HTTP address on your local network, copying comment links, search links, shared-chat links, webhook URLs, chat IDs, Git branch names, commit hashes and spreadsheet cells now works. Copy Formatted Text preserves formatting, and copied text keeps its final line break. [#&#8203;33939](https://github.com/open-webui/open-webui/pull/33939) - 📝 **Clipboard prompts and Auto-Copy Response over plain HTTP.** Prompts using the clipboard placeholder now insert into chats and channels even when the browser blocks clipboard access, with a message explaining that the clipboard could not be read. Auto-Copy Response stays off and shows a permission message instead of silently failing to turn on; turning it off no longer requires clipboard access. [#&#8203;33939](https://github.com/open-webui/open-webui/pull/33939) - 🧠 **DeepSeek thinking mode with skills, knowledge and memories.** These chats no longer fail with a missing reasoning error when the connection's Provider is set to "llama.cpp". DeepSeek thinking mode still requires that setting; connections using "Default" are unchanged. [#&#8203;33950](https://github.com/open-webui/open-webui/pull/33950) - 🏷️ **Labels on grouped tool calls.** With Tool Call Labels enabled, a group of tool calls keeps the labels that are present even if one call has none. The unlabeled call shows its tool name instead of making the whole group lose its labels. [#&#8203;33932](https://github.com/open-webui/open-webui/pull/33932) - 💬 **Tool labels in later replies.** The model now sees the labels it wrote for earlier tool calls when continuing a reply or answering another message, removing a reason it could stop writing labels as the conversation continued. [#&#8203;33953](https://github.com/open-webui/open-webui/pull/33953) - 🔎 **Knowledge searches containing punctuation.** Searches in knowledge files now find text such as "Art. 40" when the model marks the period as literal punctuation, instead of returning no matches. [#&#8203;33936](https://github.com/open-webui/open-webui/pull/33936) - 🪟 **Side Chat with Temporary Chat disabled.** Leaving Enforce Temporary Chat switched on while disabling Temporary Chat no longer prevents users from sending and saving side chats or dragging an existing chat into the side panel. [#&#8203;33930](https://github.com/open-webui/open-webui/pull/33930) - 🏠 **Home folder tab title.** A Files tab at the connected machine's home directory now says "Home" instead of showing the directory name, including when reopening saved tabs. Other folders keep their names. [Commit](https://github.com/open-webui/open-webui/commit/241c8c0e47e19b0adeb13d577ca538d2252b0ad4) - 🦙 **Missing Ollama in Docker images.** Building the image that includes Ollama now fails if Ollama was not installed, instead of producing an image that cannot start Ollama after its download failed. [#&#8203;33944](https://github.com/open-webui/open-webui/pull/33944) - 📄 **Errors after leaving a chat.** Closing or leaving a chat while its tab layout is being saved no longer fills the server log with an error report for the dropped connection. [#&#8203;33929](https://github.com/open-webui/open-webui/pull/33929) ### [`v0.13.2`](https://github.com/open-webui/open-webui/releases/tag/v0.13.2) [Compare Source](https://github.com/open-webui/open-webui/compare/v0.13.1...v0.13.2) ##### Fixed - 🔓 **Chat over plain HTTP.** Opening Open WebUI over plain HTTP at a LAN or Docker address now loads new and existing chats instead of staying on a spinner with "crypto.randomUUID is not a function". [#&#8203;33919](https://github.com/open-webui/open-webui/issues/33919) - 📝 **Comments and Markdown editing over plain HTTP.** Creating, replying to and editing comments, and editing Markdown files containing rendered blocks such as Mermaid diagrams or callouts, no longer fail with the same UUID error. [Commit](https://github.com/open-webui/open-webui/commit/3c400952cbb30e8eb72652d83b2af2f75a3b9341) - 🎛️ **Side Chat permission switch.** The Allow Side Chat switch is now visible even when Temporary Chat is off, so administrators can configure the two permissions independently. [#&#8203;33920](https://github.com/open-webui/open-webui/pull/33920) ### [`v0.13.1`](https://github.com/open-webui/open-webui/releases/tag/v0.13.1) [Compare Source](https://github.com/open-webui/open-webui/compare/v0.13.0...v0.13.1) ##### Added - 🦾 **Coding agents as models.** On an Open Terminal that offers agents, coding agents such as Claude Code, Codex and Pi now appear in the model picker. You can chat with them or call them by voice, choose their reasoning effort and service tier, and approve their tool requests in the chat. Each conversation uses one agent, and agent chats cannot be regenerated, continued or edited and resent. If an agent's terminal goes away, its chats and new chats fall back to the default model and earlier replies keep the agent's name. Autocompletion in agent chats uses the task model, or is skipped when there is none. [#&#8203;33843](https://github.com/open-webui/open-webui/pull/33843), [#&#8203;33860](https://github.com/open-webui/open-webui/pull/33860), [#&#8203;33880](https://github.com/open-webui/open-webui/pull/33880), [#&#8203;33857](https://github.com/open-webui/open-webui/pull/33857) - 🪢 **Introducing Tether.** [Tether](https://github.com/open-webui/tether) is the new desktop companion app for macOS, Windows and Linux that ties your own computer to Open WebUI. Sign in once and your computer shows up under Your devices in the terminal menu, so models, coding agents and automations can work in your real files, repositories and tools, and you can reach them from any browser, including your phone. Tether connects out to Open WebUI, so there are no ports to open, its terminal output resumes after a dropped connection, and it delivers notifications to your desktop even when no Open WebUI page is open. It signs in with a renewable session that expires after 90 days without use by default, set with "SESSION_EXPIRES_IN". - 💸 **Much cheaper tool chats.** Chats that use tools now cost much less. Citation instructions, memory recall, the skills list and attached knowledge are no longer rewritten into the system prompt or your message on every tool round, so the start of the conversation stays the same between requests and providers can serve it from their prompt cache. Custom pipes see a different message layout as a result; see Context as hidden tool results under Changed. [#&#8203;33852](https://github.com/open-webui/open-webui/issues/33852) - 🚦 **Auto tool permissions.** Administrators can now turn on Auto, a third tool permission mode next to Full access and Ask for approval (off by default). A review model checks each tool call and runs it right away only when it judges that no confirmation is needed. The review sends the tool, its arguments and your last few messages to the configured scoring API, and any call it cannot review, because of an error, a timeout or too much context, asks for approval. - 📚 **Full text for small files.** Files whose extracted text is shorter than a new threshold (50,000 characters by default, set with "RAG_FULL_TEXT_CHAR_THRESHOLD" or in the admin Documents settings) are now included in full instead of through retrieval. This applies both when Preload files adds the content to the conversation and when the model reads an attached file with its file search tool, and the file viewer shows when it happens. - ⏱️ **Faster large chats.** Large chats now open about twice as fast, and marking a chat as read or unread or removing a tag from it is 4 to 6 times faster. [#&#8203;33665](https://github.com/open-webui/open-webui/pull/33665), [#&#8203;33666](https://github.com/open-webui/open-webui/pull/33666) - 🏎️ **Faster chat access checks.** Sending messages, typing indicators, shared links, comments and file access no longer load the whole conversation just to check who may access it, which helps most on large chats. [#&#8203;33667](https://github.com/open-webui/open-webui/pull/33667), [#&#8203;33669](https://github.com/open-webui/open-webui/pull/33669) - 🔧 **More forgiving tool calls.** Tool calls with arguments wrapped in code fences, encoded twice or sent as text instead of numbers and booleans now run, and tool calls a model writes out as plain text markup are recognised and executed. [#&#8203;33868](https://github.com/open-webui/open-webui/issues/33868) - 🛑 **Tool loop stops.** A tool call that fails twice in a row with the same arguments now ends the tool loop so the model writes a final answer, and a reply cut off by the token limit no longer runs its unfinished tool calls. - 🛠️ **Default tools for models.** The Model Defaults section in the admin Models settings now has a Tools picker, models without their own tool list start new chats with those tools enabled, and the new model form preselects them. [#&#8203;33749](https://github.com/open-webui/open-webui/pull/33749), [#&#8203;33787](https://github.com/open-webui/open-webui/pull/33787) - 🚀 **Faster channels.** The channel list, the channel messages and the replied-to messages in a channel now load much faster, especially with many channels or many replies, and new database indexes speed up channel threads and reactions. [#&#8203;33690](https://github.com/open-webui/open-webui/pull/33690), [#&#8203;33691](https://github.com/open-webui/open-webui/pull/33691), [#&#8203;33688](https://github.com/open-webui/open-webui/pull/33688) - 🪶 **Lighter reasoning streams.** Long reasoning replies from providers such as OpenRouter now send far less data to the browser and to Redis while they stream. [#&#8203;33698](https://github.com/open-webui/open-webui/pull/33698) - 🗂️ **Faster sidebar chat list.** The chat list in the sidebar now appears as soon as its own data is ready instead of waiting for the channel list. [#&#8203;33686](https://github.com/open-webui/open-webui/pull/33686) - ⌨️ **Faster typing after large pastes.** Typing in the chat input after pasting 16,000 lines of text is now about 117 times faster, down from 8.6 seconds to 74 milliseconds per keystroke. [#&#8203;33729](https://github.com/open-webui/open-webui/pull/33729), [#&#8203;12087](https://github.com/open-webui/open-webui/issues/12087) - 🧮 **Faster Redis filter lookups.** With Redis enabled, chat requests now look up the filters of a model faster. [#&#8203;33668](https://github.com/open-webui/open-webui/pull/33668) - 🪟 **Split tab panes.** On large screens, dragging a tab or a chat to the edge of the workspace, or pressing the new Split screen button, splits it into resizable panes, so several tabs, including the chat, stay visible side by side. Tabs can also be reordered and moved between panes with a long press and drag on touch screens. [Commit](https://github.com/open-webui/open-webui/commit/b5eaa59d0a5f85a85940bd2ec1afb4aa4ff21a80) - 🧳 **Tabs remembered per chat.** The open tabs, split layout and selected terminal of each chat are now saved privately for each user on the server and restored when you reopen the chat, even on another device. Signing out warns you about unsaved file edits. [#&#8203;33899](https://github.com/open-webui/open-webui/pull/33899), [#&#8203;33885](https://github.com/open-webui/open-webui/pull/33885) - 🪣 **Batched Tool Search calls.** With Tool Search on, the model can run several independent tools it found in a single call_tool request, which saves model turns and so lowers the cost of tool-heavy chats. - 🧭 **Tool Search engines.** Tool Search in the admin Interface settings can now match tools with embeddings from the Documents embedding model (kept in memory, not stored) or by sending the query and tool definitions to a relevance-scoring API, besides the default keyword search. - 🏷️ **Readable tool call labels.** Administrators can now turn on Tool Call Labels in Admin Settings > Interface or with "ENABLE_TOOL_CALL_LABELS", off by default, to ask the model for short descriptions in the user's language of what each tool call is doing and what it did. [Commit](https://github.com/open-webui/open-webui/commit/2416b1fc0b83571652e0efff83f9591c89b62a64) - ✂️ **Chat read limit.** Admins can limit how much of an earlier chat the Chat History tools return, with "VIEW_CHAT_MAX_CHARS" or the new Chat Read Limit field in Admin Settings > General, and the newest messages are kept while older ones are cut. [#&#8203;33695](https://github.com/open-webui/open-webui/pull/33695), [#&#8203;33608](https://github.com/open-webui/open-webui/issues/33608) - 🌍 **Client IP header placeholders.** Custom headers of connections can now use "{{USER_IP}}" and "{{X_FORWARDED_FOR}}" to pass the client address on to the upstream provider, and both are empty for automations and background jobs. [#&#8203;33833](https://github.com/open-webui/open-webui/pull/33833), [#&#8203;33819](https://github.com/open-webui/open-webui/issues/33819) - ✉️ **Non-ASCII email addresses.** The sign-up page and the add and edit user dialogs now accept email addresses with non-ASCII characters. [#&#8203;33693](https://github.com/open-webui/open-webui/pull/33693), [#&#8203;33681](https://github.com/open-webui/open-webui/issues/33681) - 🕒 **Reply finish time.** Assistant replies now show the time they finished instead of the time the prompt was sent. [#&#8203;33895](https://github.com/open-webui/open-webui/pull/33895) - 🧐 **Channel message search.** Each channel header has a Search button that searches the messages of that channel. - 🔔 **Notifications open their chat.** Clicking a desktop notification now brings Open WebUI to the front and opens the related chat, channel or calendar. - 📮 **MCP Apps message the chat.** An MCP App can now send a text message to the chat, which you confirm first in a Confirm Prompt from Embed dialog. [#&#8203;33907](https://github.com/open-webui/open-webui/pull/33907) - 🖱️ **Drag chats into side panel.** Dragging a chat from the sidebar onto the chat area opens it as a side chat tab, and a side chat can now also be started from an empty new chat. - 🪄 **Quick open files.** A new shortcut, Cmd+P (Ctrl+P elsewhere), opens a search box that finds files in the selected terminal by name or path and opens the one you pick in a tab. - 🎹 **Tab keyboard shortcuts.** New shortcuts, which can be changed in Settings > Keyboard, open the Files, Desktop, Comments, new chat, new file, new terminal and new browser tabs, close and switch between tabs and toggle the split screen. - 🗺️ **Working directory picker.** The message input has a working directory menu with folder search, a Recently Used list kept per user on the terminal and an Open Folder dialog. - 📂 **Folder default terminal.** A folder can now have a default terminal and working directory in its Advanced settings, used for new chats in that folder. - 🧾 **Editable spreadsheets.** Excel files in the file view can now be edited in the preview, with a formula bar, pasting from other spreadsheets and resizable columns, and saved back to the file, while formula cells and protected sheets stay read-only. - 🖋️ **Markdown and code editing.** Markdown files in the file view now open in a rich text editor that saves back to Markdown, where pasted or dropped images are uploaded to the terminal, and the code editor suggests words that already appear in the file. Other text files and the Source view of previews now use the code editor directly, without a separate Edit step. [Commit](https://github.com/open-webui/open-webui/commit/b5eaa59d0a5f85a85940bd2ec1afb4aa4ff21a80) - 🌳 **Clone Git repositories.** The file browser and the folder pickers can now clone a Git repository from an HTTP(S) or SSH URL into the chosen folder of the selected terminal. - 🧬 **Git changes in the editor.** The code editor now marks the lines of a file that were added or changed since the last commit. - 🔖 **File browser Git status.** Files and folders in the file browser are marked with a Git status letter such as M, A, D or U, and folders show that they contain changes. - 🌿 **Git panel feedback and resizing.** The Git panel now shows what it is doing, such as staging, pushing or fetching, and its file list can be resized. - 🧹 **Discard any Git change.** The Git panel can now discard staged and new files too, and its confirmation warns that files not in the last commit are deleted and that this cannot be undone. - 🧲 **Drag files into chat.** Files from the terminal file browser can be dragged, one or several at once, into the message input to attach them, also with a long press on touch screens. - 🗃️ **Multiple Files tabs.** The + menu can now open several Files tabs for the same terminal, each with its own folder and the folder name as its tab title, and moving files refreshes the other Files tabs on that terminal. [Commit](https://github.com/open-webui/open-webui/commit/0b7e7b3f707023c626ae51cc66a4db1c19ed7056) - 🔲 **Drag to select files.** Several files in the file browser can now be selected by dragging a selection box over them. - 🪂 **Spring-loaded Files tabs.** Holding a dragged file over a Files tab for half a second switches to that tab, so files can be moved into folders of another tab. - ⚡ **Faster terminal rendering.** The terminal tab now renders with WebGL where the browser supports it and takes in bursts of output in larger batches, so heavy output is handled faster. - 🩺 **Terminal system info.** The terminal menu now has a System info view with the terminal's CPU, memory, disk, network, uptime and top processes. - 🖌️ **Edit images from the terminal.** The image edit tool can now take source images from the selected terminal by file path, as PNG, JPEG, WebP or GIF up to 25 megapixels. [#&#8203;33824](https://github.com/open-webui/open-webui/issues/33824) - 🎙️ **Voice notes in files.** A new shortcut, Cmd+Alt+M (Ctrl+Alt+M elsewhere), records a voice note, can transcribe it, and saves the recording with a Markdown note in the current folder of the selected terminal, then opens the note in a tab. Escape cancels a recording. [#&#8203;33902](https://github.com/open-webui/open-webui/pull/33902), [#&#8203;33903](https://github.com/open-webui/open-webui/pull/33903) - 📢 **Model availability events.** Open WebUI now publishes model.available and model.unavailable events when a model appears in or disappears from the fetched model list. - 🗒️ **Descriptions in model pickers.** The tool, skill, filter, action and sub-agent pickers in the model editor now show each item's description under its name. [#&#8203;33723](https://github.com/open-webui/open-webui/pull/33723) - 🔢 **Grouped token counts.** Whole numbers in the token usage tooltip are now grouped by thousands for the interface language, for example 26,678,623. [#&#8203;33896](https://github.com/open-webui/open-webui/pull/33896) - 🎼 **Notification sound styles.** The notification sound setting now offers Chime, Soft, Bell and Tap sounds, with a play button to preview each one. - 🔄 **General improvements.** Various improvements were implemented across the application to enhance performance, stability, and security. - 🌐 **Translation updates.** Translations for Portuguese (Brazil), Azerbaijani, Basque, Bengali, Bosnian, Bulgarian, Cebuano, Croatian, Czech, Danish, Dutch, Estonian, Faroese, Filipino, Galician, Georgian, German, Greek, Hebrew, Hindi, Hungarian, Indonesian, Irish, Japanese, Kabyle, Korean, Lithuanian, Malay, Norwegian Bokmål, Persian, Polish, Punjabi, Romanian, Russian, Serbian, Slovak, Slovenian, Spanish (Spain), Swedish, Tamil, Thai, Tibetan, Turkish, Turkmen, Ukrainian, Urdu, Uyghur and Vietnamese were enhanced and expanded. ##### Fixed - 🛡️ **Security Advisory**: This release contains **security fixes**. We **strongly recommend updating** all deployments **as soon as possible** to have the latest security fixes in place. These fixes are **not listed in this changelog**; we will publish the security advisories for them immediately following CVE assignment. [Advisories](https://github.com/open-webui/open-webui/security) - 🗄️ **Redis Cluster chats.** Sending a chat message on a Redis Cluster deployment no longer fails while the personal terminal relay starts. [#&#8203;33766](https://github.com/open-webui/open-webui/pull/33766) - 🚪 **Automatic SSO sign-in.** With automatic SSO redirect on, an expired session now signs you back in without a state mismatch error. [#&#8203;33791](https://github.com/open-webui/open-webui/pull/33791), [#&#8203;33783](https://github.com/open-webui/open-webui/issues/33783) - 🎨 **OpenAI image default model.** Image generation with the OpenAI engine now defaults to GPT-IMAGE 2 instead of the shut-down DALL·E 2, and the shut-down DALL·E models are no longer offered. [#&#8203;33694](https://github.com/open-webui/open-webui/pull/33694), [#&#8203;33578](https://github.com/open-webui/open-webui/issues/33578) - 🖼️ **Default GPT image requests.** OpenAI image generation and editing with no model explicitly set no longer send an unsupported response format to the default GPT image model. [#&#8203;33875](https://github.com/open-webui/open-webui/pull/33875) - 🔐 **Terminal Session authentication.** Terminal connections set to Session authentication now accept an API key, so Verify Connection, the policy settings and the terminal tools work again. [#&#8203;33699](https://github.com/open-webui/open-webui/pull/33699), [#&#8203;30001](https://github.com/open-webui/open-webui/issues/30001), [#&#8203;33579](https://github.com/open-webui/open-webui/issues/33579) - 🦙 **Ollama broken tool calls.** Chats with Ollama models no longer get stuck after a tool call with cut-off, rejected or empty arguments. [#&#8203;33697](https://github.com/open-webui/open-webui/pull/33697), [#&#8203;33671](https://github.com/open-webui/open-webui/issues/33671) - 🧠 **Complete compaction summaries.** Context compaction summaries are no longer cut off at 1000 tokens, which could leave reasoning models with an empty or truncated summary. [#&#8203;33726](https://github.com/open-webui/open-webui/pull/33726) - 💾 **Prompt cache on tool results.** Tool results sent through OpenAI-compatible gateways now keep their prompt cache marker, so Anthropic models behind OpenRouter or LiteLLM can cache the prefix in tool loops. [#&#8203;33757](https://github.com/open-webui/open-webui/pull/33757), [#&#8203;33755](https://github.com/open-webui/open-webui/issues/33755) - 📖 **File read size limit.** Reading a file by line range now stops at the same size limit as other reads, so one read can no longer overflow the context window of the model. [#&#8203;33696](https://github.com/open-webui/open-webui/pull/33696), [#&#8203;33607](https://github.com/open-webui/open-webui/issues/33607) - 🧰 **Tool Search retries.** A malformed tool call with Tool Search on no longer disables tools for the rest of the reply, so the model can correct its arguments and try again. [#&#8203;33872](https://github.com/open-webui/open-webui/pull/33872) - 🧩 **Split tool calls.** When a model sends several tool calls glued into one, the chat no longer shows an extra tool call that never gets a result. [#&#8203;33779](https://github.com/open-webui/open-webui/pull/33779), [#&#8203;33778](https://github.com/open-webui/open-webui/issues/33778) - 🔎 **External knowledge base queries.** External knowledge bases now search with every generated query at the same time instead of using only the results of the first one. [#&#8203;33704](https://github.com/open-webui/open-webui/pull/33704), [#&#8203;33184](https://github.com/open-webui/open-webui/issues/33184) - 📊 **Large spreadsheet uploads.** Uploading a large, sparsely filled spreadsheet no longer takes minutes and stalls the server. [#&#8203;33770](https://github.com/open-webui/open-webui/pull/33770), [#&#8203;33768](https://github.com/open-webui/open-webui/issues/33768) - 🧱 **Milvus knowledge base uploads.** Adding a file to a knowledge base on Milvus no longer processes the file a second time or stores its chunks twice. [#&#8203;33745](https://github.com/open-webui/open-webui/pull/33745), [#&#8203;33744](https://github.com/open-webui/open-webui/issues/33744) - 📄 **MinerU 4.1 servers.** The MinerU document loader in local mode now works with MinerU 4.1 servers, where every document used to fail. [#&#8203;33719](https://github.com/open-webui/open-webui/pull/33719), [#&#8203;33718](https://github.com/open-webui/open-webui/issues/33718) - 🔑 **MCP OAuth stays signed in.** The login to an MCP tool server that uses OAuth is no longer lost when the provider has a temporary outage while the token is refreshed. [#&#8203;33805](https://github.com/open-webui/open-webui/pull/33805), [#&#8203;33803](https://github.com/open-webui/open-webui/issues/33803) - 📍 **Location lookup delays.** With location access turned on, sending a message no longer waits for a slow or unanswered browser location request. - 🗓️ **Automation schedule examples.** Models creating automations through the built-in tool no longer hit "Invalid RRULE" errors caused by a wrong example schedule format. [#&#8203;33732](https://github.com/open-webui/open-webui/pull/33732), [#&#8203;33730](https://github.com/open-webui/open-webui/issues/33730) - ♿ **Keyboard and screen reader fixes.** Keyboard focus now stays where it belongs: on buttons while they save, in place when you cancel a message edit, switch message versions or sort and page a workspace or notes list, on the first and last page buttons, and in the title field while a note title is generated. Closing a channel thread, a note panel or a chat menu returns focus to where it was, and the sign-in skip link moves focus to the form. The show password button is now wide enough to hit, and holding Shift no longer swaps row buttons for Delete while a keyboard user has focus. - 🕵️ **Audit log secret masking.** With request auditing on, the audit log now masks API keys, tokens and secrets saved in the admin settings, not only passwords. [#&#8203;33746](https://github.com/open-webui/open-webui/pull/33746) - 🔇 **Queries kept out of logs.** The default server log no longer contains your search queries or the provider address of a failing external reranker. [#&#8203;33701](https://github.com/open-webui/open-webui/pull/33701) - 💬 **Chat message search.** Workspace search now finds message text and shows matching excerpts in chats whose messages are stored separately from the chat record. [Commit](https://github.com/open-webui/open-webui/commit/07029af63db59c9ce868b4562aa3f1c848e8b9e7) - 🔃 **Search results stay current.** Notes, archived and shared chats, the admin user list and the knowledge and file pickers in the message input no longer let an older search replace the results of a newer one. [Commit](https://github.com/open-webui/open-webui/commit/53b1b1928b1ee08175c0218884d7fe40c6aa909c) - 👎 **Feedback details saved.** A thumbs-down reason and comment are no longer lost when saved before the feedback tags finish generating. [#&#8203;33716](https://github.com/open-webui/open-webui/pull/33716) - 🎒 **Side chat default tools.** New side chats now start with the same default tools, skills, filters and web search as a normal chat with that model. [#&#8203;33788](https://github.com/open-webui/open-webui/pull/33788) - ✅ **Side chat tool approvals.** Tool calls in side chats now show working approval controls and resume the reply after approval. [#&#8203;33889](https://github.com/open-webui/open-webui/pull/33889) - ❓ **Side chat questions.** Questions asked through the model's ask_user tool now appear in side chats and can be answered or dismissed, including after reopening the chat. [#&#8203;33901](https://github.com/open-webui/open-webui/pull/33901) - ⏩ **Continued replies on Responses connections.** Continuing a reply on a Responses API connection now adds the new text to the existing reply. - 🛂 **Model required before sending.** The file ask box, follow-up suggestions and queued messages now report "Model not selected" instead of trying to send when no model is selected. [#&#8203;33838](https://github.com/open-webui/open-webui/pull/33838) - 👻 **Temporary chat button with tabs.** The Temporary Chat button is back on new chats when the tabbed layout is on. [#&#8203;33680](https://github.com/open-webui/open-webui/pull/33680) - 🪧 **Stable MCP app views.** MCP apps in chat no longer reload and lose their state while the message around them updates. - 🎴 **Artifacts keep the chat visible.** Opening an HTML or SVG preview from a reply now uses the side pane instead of replacing the visible chat tab. [#&#8203;33894](https://github.com/open-webui/open-webui/pull/33894) - 📣 **Banners in tabbed view.** Admin banners and license notices now show on the new chat page when the terminal tabs split the screen. [#&#8203;33789](https://github.com/open-webui/open-webui/pull/33789) - 📡 **Tabs over plain HTTP.** The New File, New Terminal and Browser tabs in the side panel now open when Open WebUI is reached over plain HTTP on a LAN or Docker address. [#&#8203;33847](https://github.com/open-webui/open-webui/pull/33847), [#&#8203;33845](https://github.com/open-webui/open-webui/issues/33845) - 🕶️ **Per-chat terminals in temporary chats.** Per-chat workspace terminals are no longer offered in temporary chats, where sending to them failed, and one already selected is deselected when Temporary mode is turned on. [#&#8203;33866](https://github.com/open-webui/open-webui/pull/33866), [#&#8203;33861](https://github.com/open-webui/open-webui/issues/33861) - 🖥️ **Terminal kept on model switch.** Switching to a model without its own default terminal no longer turns off the terminal selected in the chat. - 📁 **Working folder in new chats.** The terminal working folder of a new chat is no longer reset to the root when the chat is created before the file browser has finished loading. - 📞 **Return to call.** The Return to call button now reopens the active voice call after switching to the chat tab. [#&#8203;33898](https://github.com/open-webui/open-webui/pull/33898) - 🔕 **Channel notifications clear on read.** Opening a channel now marks its notifications as read. - 🪝 **Webhooks window stability.** The channel Webhooks window no longer jumps back to the webhook list when a new message arrives, so edits in progress are kept. [#&#8203;33776](https://github.com/open-webui/open-webui/pull/33776), [#&#8203;33775](https://github.com/open-webui/open-webui/issues/33775) - 🎧 **Background audio on iOS.** With Auto-Playback Response off, tapping in Open WebUI or opening a chat no longer stops audio playing in other apps on iOS. [#&#8203;33780](https://github.com/open-webui/open-webui/issues/33780) - 🖨️ **Scanned PDF preview.** PDF previews no longer show blank pages for scanned PDFs that store their pages as JPEG2000 images. [#&#8203;33906](https://github.com/open-webui/open-webui/pull/33906), [#&#8203;33905](https://github.com/open-webui/open-webui/issues/33905) - 🔬 **Preview zoom stability.** Zooming in PDF, Word and PowerPoint previews no longer resets or drifts when you switch tabs, change slides or resize the pane. [#&#8203;33836](https://github.com/open-webui/open-webui/pull/33836), [#&#8203;33835](https://github.com/open-webui/open-webui/issues/33835) - 📌 **File preview from Controls.** A file preview opened from the Controls menu no longer closes when you click inside it. [#&#8203;33831](https://github.com/open-webui/open-webui/pull/33831), [#&#8203;33828](https://github.com/open-webui/open-webui/issues/33828) - 🔭 **Tab bar and UI scale.** Tab content no longer slides under the tab bar when the UI scale is above 1. [#&#8203;33830](https://github.com/open-webui/open-webui/pull/33830), [#&#8203;33829](https://github.com/open-webui/open-webui/issues/33829) - 🔘 **Desktop app top clicks.** In the desktop app, banners and menus near the top of the window now take clicks instead of dragging the window. - 📱 **Mobile menu keyboard.** Closing the + menu on a phone no longer opens the keyboard. [#&#8203;33893](https://github.com/open-webui/open-webui/pull/33893) - 📝 **File rename in terminal.** Renaming a file or folder in the Open Terminal file browser no longer shows a false error after the rename succeeded. [#&#8203;33786](https://github.com/open-webui/open-webui/pull/33786) - 📋 **Stable file selection.** Selecting files no longer shifts the file rows when the bulk action bar appears, and clicking an empty area of the file list clears the selection. [Commit](https://github.com/open-webui/open-webui/commit/10f1226f0904a78a59daffeacbb7fd6404450364) - 🫥 **Selection popups on scroll.** Scrolling the file view dismisses its Ask box, and scrolling a rich text editor dismisses its floating formatting menus until the selection changes. [Commit](https://github.com/open-webui/open-webui/commit/555125b3697781fa593a27cd496a4d84c772552a), [Commit](https://github.com/open-webui/open-webui/commit/2416b1fc0b83571652e0efff83f9591c89b62a64) - 🙈 **Empty control groups hidden.** When every control of a group is hidden with Visible controls, the empty group heading or submenu no longer shows. [#&#8203;33867](https://github.com/open-webui/open-webui/pull/33867) - 🌄 **Chat background behind tabs.** The chat tab bar no longer covers the chat background with a solid strip. [#&#8203;33849](https://github.com/open-webui/open-webui/pull/33849) - 🌫️ **Top bar blur.** The top bars of Workspace, Admin Panel, Playground and Calendar no longer blur a page background image. [#&#8203;33799](https://github.com/open-webui/open-webui/pull/33799), [#&#8203;33798](https://github.com/open-webui/open-webui/issues/33798) - 🗑️ **Feedback model filter.** Deleting the last feedback of a model in the Evaluations feedback list now clears the model filter. [#&#8203;33794](https://github.com/open-webui/open-webui/pull/33794), [#&#8203;33793](https://github.com/open-webui/open-webui/issues/33793) - 🔤 **Visible controls labels.** Six toggles under Visible controls in the Admin Interface settings now show their names instead of raw translation keys. [#&#8203;33734](https://github.com/open-webui/open-webui/pull/33734) - 🤫 **Quiet terminal disconnects.** Closing the file panel or a terminal view right after opening it no longer logs an error traceback on the server. [#&#8203;33863](https://github.com/open-webui/open-webui/pull/33863), [#&#8203;33850](https://github.com/open-webui/open-webui/issues/33850) - 📜 **Quieter MCP connection logs.** Logs no longer fill with a warning for every MCP server a user has not connected yet each time they load Open WebUI. [#&#8203;33663](https://github.com/open-webui/open-webui/pull/33663) - 🪪 **OAuth token refresh warnings.** Refreshing OAuth tokens with a client secret no longer logs deprecation warnings on every call. [#&#8203;33700](https://github.com/open-webui/open-webui/pull/33700), [#&#8203;33297](https://github.com/open-webui/open-webui/issues/33297) ##### Changed - ⚠️ **Database Migrations**: This release includes database schema changes; we strongly recommend backing up your database and all associated data before upgrading in production environments. If you are running a multi-worker, multi-server, or load-balanced deployment, all instances must be updated simultaneously, rolling updates are not supported and will cause application failures due to schema incompatibility. - 💥 **Preload files off by default.** With native function calling, the default, attached files are no longer added to the conversation automatically, and models read them with their file tools instead. Models without tool support or without the Files built-in tool can no longer see attached files until you turn on Preload files in the model's Capabilities. Legacy function calling, and API requests that do not come from a chat in the web UI, keep the old behavior. [#&#8203;33815](https://github.com/open-webui/open-webui/pull/33815) - 🎯 **Retrieval defaults.** Unless you saved your own values, "RAG_TOP_K" and "RAG_TOP_K_RERANKER" now default to 10 instead of 3, and "CHUNK_SIZE" to 1024 instead of 1000. Each query now retrieves more text, so prompts with knowledge get longer. The new chunk size applies to files processed from now on; existing knowledge keeps its chunks until it is reindexed. - 🧷 **Context as hidden tool results.** With native function calling, memory recall, the skills list and the attached knowledge list now arrive as hidden tool results after your message instead of in the system prompt. This is what keeps tool chats cacheable, and it means custom pipes can see a tool message as the last message. - 🗨️ **Side chats are saved.** Side chats, which were always temporary in 0.13.0, are now saved as regular chats and appear in the chat list. They stay temporary only when you turn on Temporary Chat in the side chat or your admin enforces temporary chats. - 🚧 **Tool call limit.** Reaching the tool call limit now makes the model write a final answer without tools, instead of ending the reply with the "Tool-call limit reached" error. - 🔦 **Tool Search prompt.** With Tool Search on, the model no longer sees a list of the hidden tools and finds them only by searching. - ✏️ **Always Use Full Text.** The Full Context Mode setting in the admin Documents settings is now called Always Use Full Text and includes complete documents regardless of the full text threshold. - ↔️ **Always full-width tabs.** On large screens the tabbed workspace now always fills the chat area, with the chat as one of its tabs, and the full-width toggle added in 0.13.0 is gone. - 🔀 **Integrations moved into + menu.** The Integrations button is now part of the + menu in the message input, and the new Show Input Selections setting, off by default, controls whether selected tools, skills, filters and capabilities are shown beside it. - 🔓 **Side chat permission.** The Allow Side Chat permission no longer requires the Temporary Chat permission. - 🕸️ **Chrome first in Browser tab.** The Browser tab now starts in Chrome mode when the terminal offers it, and switches to Proxy mode with a message if Chrome fails to start. - ⛔ **Terminal select setting removed.** The Show Files on Terminal Select setting has been removed from the Interface settings. [#&#8203;33855](https://github.com/open-webui/open-webui/pull/33855), [#&#8203;33840](https://github.com/open-webui/open-webui/issues/33840) </details> --- ### Configuration 📅 **Schedule**: (in timezone Europe/Rome) - Branch creation - At any time (no schedule defined) - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about these updates again. --- - [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check this box --- This PR has been generated by [Mend Renovate CLI](https://github.com/renovatebot/renovate). <!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0NC4xOTkuMiIsInVwZGF0ZWRJblZlciI6IjQ0LjIyMi41IiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6WyJyZW5vdmF0ZS9jb250YWluZXIiLCJ0eXBlL3BhdGNoIl19-->
fix(container): update image ghcr.io/open-webui/open-webui (v0.13.0 ➔ v0.13.1)
All checks were successful
Konflate 2 resources changed
Labeler / Labeler (pull_request) Successful in 2m42s
Lint / yamllint (pull_request) Successful in 2m49s
Lint / actionlint (pull_request) Successful in 2m55s
AI PR Review / AI PR Review (pull_request) Successful in 5m38s
dc9e89f459
Collaborator

Routine: only container-image and chart-version changes; 2 resources across 2 apps

image from to upstream
ghcr.io/open-webui/open-webui v0.13.0@sha256:2272b0… v0.13.3@sha256:46b591… ❔

konflate 0.6.6 · rendered eaae3da · full diff →

<!-- konflate:pr-2610 --> **Routine**: only container-image and chart-version changes; 2 resources across 2 apps | image | from | to | upstream | |---|---|---|---| | `ghcr.io/open-webui/open-webui` | `v0.13.0@sha256:2272b0…` | `v0.13.3@sha256:46b591…` | ❔ | <sub>konflate 0.6.6 · rendered `eaae3da` · [full diff →](https://konflate.aresu.eu/#/pr/2610)</sub>

AI Automated Review

Analysis engine: (model unavailable) · pr-reviewer-action v3.8.0

AI review could not run

The configured model endpoint(s) did not return a usable review for this run (reason: Fallback model failed).

This is an automated notice, not a substantive review. Re-run the workflow once the endpoint is reachable; see the action logs for the underlying error.

<!-- ai-pr-reviewer --> <!-- ai-pr-reviewer:{"version":1,"head_sha":"eaae3daffd30a8d17dab62ab24d64c8ff60975f8","base_sha":"3da074be12d693d8caaa815c229b5a1bdfbe02f4","review_result":"issues","review_route":"primary","ci_state":"pending","tool_budget":24,"tool_budget_source":"tier-default","tool_calls":0,"tool_rounds":0,"max_rounds":24,"context_budget":65536,"context_peak":9435,"action_version":"v3.8.0","degraded_gate_bypass":true,"tool_stop_reason":"request-error"} --> <!-- ai-pr-review-sha:eaae3daffd30a8d17dab62ab24d64c8ff60975f8 --> <!-- ai-pr-review-fingerprint:3aa5fab89ccd4e085994519fdfcb83adbaa0dafe593e3b3914f053bc9a7a324c|cfg:5e0aeeaae5d6309ddbdd547a004fede6a6150e4f63a0b4e4debe12c7e45f8edc --> # AI Automated Review _Analysis engine: (model unavailable) · pr-reviewer-action v3.8.0_ ## AI review could not run The configured model endpoint(s) did not return a usable review for this run (reason: Fallback model failed). This is an automated notice, not a substantive review. Re-run the workflow once the endpoint is reachable; see the action logs for the underlying error.
renovate-bot changed title from fix(container): update image ghcr.io/open-webui/open-webui (v0.13.0 ➔ v0.13.1) to fix(container): update image ghcr.io/open-webui/open-webui (v0.13.0 ➔ v0.13.2) 2026-11-08 14:07:56 +01:00
renovate-bot force-pushed renovate/ghcr.io-open-webui-open-webui-0.x from dc9e89f459
All checks were successful
Konflate 2 resources changed
Labeler / Labeler (pull_request) Successful in 2m42s
Lint / yamllint (pull_request) Successful in 2m49s
Lint / actionlint (pull_request) Successful in 2m55s
AI PR Review / AI PR Review (pull_request) Successful in 5m38s
to 6664bcb7a4
Some checks failed
Labeler / Labeler (pull_request) Successful in 2m14s
Lint / actionlint (pull_request) Successful in 6m0s
Lint / yamllint (pull_request) Successful in 6m29s
AI PR Review / AI PR Review (pull_request) Failing after 9m20s
Konflate 2 resources changed
2026-11-08 14:08:14 +01:00
Compare
renovate-bot force-pushed renovate/ghcr.io-open-webui-open-webui-0.x from 6664bcb7a4
Some checks failed
Labeler / Labeler (pull_request) Successful in 2m14s
Lint / actionlint (pull_request) Successful in 6m0s
Lint / yamllint (pull_request) Successful in 6m29s
AI PR Review / AI PR Review (pull_request) Failing after 9m20s
Konflate 2 resources changed
to eaae3daffd
All checks were successful
Labeler / Labeler (pull_request) Successful in 9s
Lint / actionlint (pull_request) Successful in 32s
Lint / yamllint (pull_request) Successful in 32s
AI PR Review / AI PR Review (pull_request) Successful in 3m13s
Konflate 2 resources changed
2026-11-09 09:02:31 +01:00
Compare
renovate-bot changed title from fix(container): update image ghcr.io/open-webui/open-webui (v0.13.0 ➔ v0.13.2) to fix(container): update image ghcr.io/open-webui/open-webui (v0.13.0 ➔ v0.13.3) 2026-11-09 09:02:40 +01:00
All checks were successful
Labeler / Labeler (pull_request) Successful in 9s
Lint / actionlint (pull_request) Successful in 32s
Lint / yamllint (pull_request) Successful in 32s
AI PR Review / AI PR Review (pull_request) Successful in 3m13s
Konflate 2 resources changed
This pull request can be merged automatically.
This branch is out-of-date with the base branch
You are not authorized to merge this pull request.
View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u origin "renovate/ghcr.io-open-webui-open-webui-0.x:renovate/ghcr.io-open-webui-open-webui-0.x"
git switch "renovate/ghcr.io-open-webui-open-webui-0.x"

Merge

Merge the changes and update on Forgejo.

Warning: The "Autodetect manual merge" setting is not enabled for this repository, you will have to mark this pull request as manually merged afterwards.

git switch "main"
git merge --no-ff "renovate/ghcr.io-open-webui-open-webui-0.x"
git switch "renovate/ghcr.io-open-webui-open-webui-0.x"
git rebase "main"
git switch "main"
git merge --ff-only "renovate/ghcr.io-open-webui-open-webui-0.x"
git switch "renovate/ghcr.io-open-webui-open-webui-0.x"
git rebase "main"
git switch "main"
git merge --no-ff "renovate/ghcr.io-open-webui-open-webui-0.x"
git switch "main"
git merge --squash "renovate/ghcr.io-open-webui-open-webui-0.x"
git switch "main"
git merge --ff-only "renovate/ghcr.io-open-webui-open-webui-0.x"
git switch "main"
git merge "renovate/ghcr.io-open-webui-open-webui-0.x"
git push origin "main"
Sign in to join this conversation.
No description provided.