{"id":15579,"date":"2026-11-17T11:03:48","date_gmt":"2026-11-17T02:03:48","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/?p=15579"},"modified":"2026-11-17T11:13:20","modified_gmt":"2026-11-17T02:13:20","slug":"open-webui-v0131-release","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/open-webui-v0131-release\/","title":{"rendered":"Open WebUI v0.13.1 Released with Security Fixes and Tether"},"content":{"rendered":"<p><!-- lmw:facts --><\/p>\n<h2>At a Glance<\/h2>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Value<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Repository<\/td>\n<td><a href=\"https:\/\/github.com\/open-webui\/open-webui\">open-webui\/open-webui<\/a><\/td>\n<\/tr>\n<tr>\n<td>Version<\/td>\n<td><a href=\"https:\/\/github.com\/open-webui\/open-webui\/releases\/tag\/v0.13.1\">v0.13.1<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-11-08<\/td>\n<\/tr>\n<tr>\n<td>License<\/td>\n<td>NOASSERTION<\/td>\n<\/tr>\n<tr>\n<td>Source type<\/td>\n<td>Primary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Values determined by this site&#8217;s code when the information was collected. Dates are JST.<\/em><\/p>\n<p><!-- \/lmw:facts --><\/p>\n<h2>Overview<\/h2>\n<p>The latest version &#8220;v0.13.1&#8221; of &#8220;open-webui&#8221;, an open-source software providing an AI interface, has been released. open-webui is a Web UI tool that allows users to connect Ollama, OpenAI-compatible APIs, and others to use them directly from a browser.<\/p>\n<p>Since this version includes security fixes whose details are scheduled to be published at a later date, prompt updating is strongly recommended for all production environments. In terms of features, major additions and performance improvements have been introduced, including the ability to call coding agents like Claude Code directly from the model selection screen, the introduction of the desktop integration app &#8220;Tether&#8221;, and cost reductions for tool chats leveraging prompt caching.<\/p>\n<h2>Breaking Changes and Deprecations<\/h2>\n<p>Changes have been made to the message structure when using tools, as well as default values associated with the discontinuation of integrated services.<\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Previous Behavior<\/th>\n<th>Behavior from v0.13.1 Onward<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>System prompt updates during tool execution<\/td>\n<td>Citation instructions, memory recall, skill lists, and attached knowledge were rewritten into the system prompt and messages in each round of tool calls<\/td>\n<td>Changed to a message layout that does not rewrite them, fixing the structure at the beginning of the conversation to leverage prompt caching<\/td>\n<\/tr>\n<tr>\n<td>OpenAI image generation engine default model<\/td>\n<td>DALL\u00b7E 2 (discontinued)<\/td>\n<td>GPT-IMAGE 2 (discontinued DALL\u00b7E models are excluded from choices)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p>This change affects developers who create and operate custom pipes that customize the Open WebUI message pipeline, as well as users who utilize image generation via the OpenAI engine. If you use custom pipes, operation checks are required to match the differences in the message layout.<\/p>\n<h2>Key Changes<\/h2>\n<h3>Model Selection Support for Coding Agents<\/h3>\n<p>In the Open Terminal environment providing agent features, coding agents such as Claude Code, Codex, and Pi are now displayed directly on the model selection screen. They can be called in chats or voice calls just like regular models, and support specifying reasoning effort and service tiers, as well as tool execution approval within chat. Chats with agents are fixed to one agent per conversation and do not support regeneration or mid-chat editing. If the terminal is disconnected, it falls back to the default model.<\/p>\n<h3>Introduction of the &#8220;Tether&#8221; Desktop Companion App<\/h3>\n<p>&#8220;Tether&#8221;, a desktop companion app supporting macOS, Windows, and Linux, is now available. Once you log in to Tether on your local PC, your device will appear in the &#8220;Tether&#8221; menu under terminal settings. Because it establishes an outbound connection to the Open WebUI side without requiring port forwarding, it allows you to directly operate local files, repositories, and development tools from a browser or smartphone. The default session timeout when inactive is set to 90 days.<\/p>\n<h3>Caching Efficiency and Cost Reduction for Tool Execution Chats<\/h3>\n<p>In chats utilizing tools, the process of re-inserting citation instructions, memory, and knowledge information into the system prompt on every turn has been deprecated. This keeps the conversation prefix unchanged across requests, allowing prompt caching on the API provider side to function effectively and significantly reducing API usage costs. Additionally, a fix has been applied to maintain prompt caching markers even for tool results via OpenAI-compatible gateways.<\/p>\n<h3>Automatic Tool Permission Mode and Execution Loop Safety Improvements<\/h3>\n<p>In addition to the conventional &#8220;Full access&#8221; and &#8220;Ask for approval&#8221; modes, an &#8220;Auto&#8221; mode has been added to the tool execution permissions configurable by administrators. A review model evaluates the tool call and immediate context, executing it immediately only when it determines that confirmation is unnecessary. Furthermore, processing to automatically stop execution loops for tool calls that fail twice consecutively with the same arguments, and fallback processing to automatically fix and execute malformed formats such as arguments enclosed in Markdown code fences, have been introduced.<\/p>\n<h3>Accelerated Chat Processing and Access Right Checks<\/h3>\n<p>Processing to open large conversations has been accelerated by approximately 2x, while switching read\/unread statuses and deleting tags have been made 4 to 6x faster. Previously, the entire conversation data (including embedded images) was read from the database every time message sending, typing indicators, or file access permission checks were performed; this has been revised to retrieve only owner and folder information. This eliminates unnecessary read loads in conversations scaling to tens of megabytes including images.<\/p>\n<h3>Fix for Tool Call Stack During Ollama Integration<\/h3>\n<p>In chats with Ollama models, a bug where subsequent message sending completely stopped when tool arguments were cut off halfway or rejected has been fixed. Improvements have been made to allow Ollama to continue processing as empty arguments even if invalid or empty arguments occur, preventing conversations from stalling.<\/p>\n<h2>Supported Models and Hardware<\/h2>\n<p>This version updates several support statuses regarding external services, integration tools, and backend engines. Changes to supported models and integration targets in local and cloud environments are as follows:<\/p>\n<h3>Integration of Coding Agent Models<\/h3>\n<p>As agent features provided via Open Terminal, coding agents such as Claude Code, Codex, and Pi are now available on the model selection screen. Conversations can be started via voice or text just like regular language models, supporting reasoning effort and service tier settings, as well as tool execution approval within chat. One agent is assigned per conversation, and it automatically falls back to the default model if the terminal connection is lost.<\/p>\n<h3>Updates to Image Generation Models and Providers<\/h3>\n<p>In OpenAI&#8217;s image generation engine, the standard model has been switched from the discontinued DALL\u00b7E 2 to &#8220;GPT-IMAGE 2&#8221;. As a result, legacy DALL\u00b7E series models whose provision has ended have been removed from the UI choices. Moreover, an issue where unnecessary base64 output parameters were sent and caused errors in requests without explicitly specified model names has been resolved, allowing safe utilization of GPT image models under default settings. Furthermore, image files (PNG, JPEG, WebP, GIF up to 25 megapixels) within the terminal can now be specified by path to be used as input sources for image editing tools.<\/p>\n<h3>Improvements to Inference Backends and Local Integration<\/h3>\n<p>In integration with Ollama, which is widely used for local LLM operations, an issue where conversations became unprogressable due to tool calls with cut-off or empty arguments has been fixed, enabling stable continuous conversations. In addition, in OpenAI-compatible gateway connections via OpenRouter or LiteLLM, prompt caching markers for models such as Anthropic are now retained even within tool execution loops, allowing users to receive the maximum benefits of context caching. Data transfer volume to the browser and Redis when processing long reasoning streams has also been reduced.<\/p>\n<h3>Document Parsing and External Knowledge Integration<\/h3>\n<p>In the local mode document loader, connection to the &#8220;MinerU 4.1&#8221; server, which previously caused processing errors, is now supported. Also, a bug where file processing and chunk saving were performed twice during knowledge base registration to the external vector database Milvus has been resolved. External knowledge base searches have been improved to a method where all generated queries are searched simultaneously in parallel, aiming to improve search accuracy.<\/p>\n<h3>Hardware and Client Environments<\/h3>\n<p>The newly provided desktop companion app &#8220;Tether&#8221; supports macOS, Windows, and Linux platforms. Additionally, WebGL rendering has been introduced to the browser-side terminal tab rendering engine, improving rendering performance when receiving large amounts of output.<\/p>\n<h2>How to Get It<\/h2>\n<p>Since this release includes security fixes, prompt updating is recommended in all production environments. Follow the steps corresponding to your environment to perform the update.<\/p>\n<h3>Using Docker<\/h3>\n<p>Stop and remove the currently running container, pull the latest image, and restart it.<\/p>\n<pre><code class=\"language-bash\">docker stop open-webui\ndocker rm open-webui\ndocker pull ghcr.io\/open-webui\/open-webui:main\ndocker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:\/app\/backend\/data --name open-webui --restart always ghcr.io\/open-webui\/open-webui:main\n<\/code><\/pre>\n<p>If using GPU acceleration, introduce the Nvidia Container Toolkit, add <code>--gpus all<\/code> to the startup options, and specify an image with the <code>:cuda<\/code> or <code>:ollama<\/code> tag.<\/p>\n<h3>Using a Python Environment (uvx \/ pip)<\/h3>\n<p>In macOS or Linux environments where uv is installed, you can run the latest version directly with the following command:<\/p>\n<pre><code class=\"language-bash\">DATA_DIR=~\/.open-webui uvx --python 3.12 open-webui@latest serve\n<\/code><\/pre>\n<p>If you are using pip to install in a Python 3.12 environment, update the package and restart the server:<\/p>\n<pre><code class=\"language-bash\">pip install --upgrade open-webui\nopen-webui serve\n<\/code><\/pre>\n<p>After startup, access the specified port in your browser (<code>http:\/\/localhost:3000<\/code> in the Docker command example, or <code>http:\/\/localhost:8080<\/code> for direct Python startup) to check operation.<\/p>\n<h3>Using the Desktop App<\/h3>\n<p>If you are using the desktop app version for macOS, Windows, or Linux, obtain the latest installer from the official repository and update.<\/p>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/open-webui-v0-13-0\/\">Open WebUI v0.13.0 Released: Tabbed Workspaces &amp; Security Fixes<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/open-webui-v0-12-3-released\/\">Open WebUI v0.12.3 Released with E2EE Chats and Redis Streaming<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/open-webui-v012-1-released\/\">Open WebUI v0.12.1 Released: External Apps and More<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/open-webui-v0125-released\/\">Open WebUI v0.12.5 Released with Security Patches and Multi-Page Notes<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:related --><\/p>\n<p><!-- lmw:next-steps --><\/p>\n<h2>What to Read Next<\/h2>\n<ul>\n<li><strong>Follow this tool<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-open-webui-en\/\">Open WebUI overview and release history (181 releases tracked)<\/a><\/li>\n<li><strong>Other uis and desktop apps<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-jan-en\/\">Jan<\/a> \/ <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-text-generation-webui-en\/\">TextGen<\/a><\/li>\n<li><strong>Engines mentioned in this article<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-ollama-en\/\">Ollama<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/github.com\/open-webui\/open-webui\/releases\/tag\/v0.13.1\">https:\/\/github.com\/open-webui\/open-webui\/releases\/tag\/v0.13.1<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/open-webui\/open-webui\/pull\/33667\">https:\/\/github.com\/open-webui\/open-webui\/pull\/33667<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/open-webui\/open-webui\/pull\/33669\">https:\/\/github.com\/open-webui\/open-webui\/pull\/33669<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/open-webui\/open-webui\/pull\/33694\">https:\/\/github.com\/open-webui\/open-webui\/pull\/33694<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/open-webui\/open-webui\/pull\/33697\">https:\/\/github.com\/open-webui\/open-webui\/pull\/33697<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/open-webui\/open-webui\/pull\/33766\">https:\/\/github.com\/open-webui\/open-webui\/pull\/33766<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/open-webui\/open-webui\/pull\/33875\">https:\/\/github.com\/open-webui\/open-webui\/pull\/33875<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/open-webui\/tether\">https:\/\/github.com\/open-webui\/tether<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/open-webui\/desktop\">https:\/\/github.com\/open-webui\/desktop<\/a><\/li>\n<li><a href=\"https:\/\/docs.openwebui.com\/getting-started\/\">https:\/\/docs.openwebui.com\/getting-started\/<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Open WebUI v0.13.1 has been released, featuring security fixes, the Tether desktop app, coding agent support, and performance improvements.<\/p>\n","protected":false},"author":1,"featured_media":15578,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[105],"tags":[4764,1475,1093,1917,4766,1547],"class_list":["post-15579","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-engines-and-tools","tag-claude-code-en","tag-docker-en","tag-ollama-en","tag-open-webui-en","tag-tether-en","tag-verified"],"lang":"en","translations":{"en":15579,"ja":15577},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/15579","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=15579"}],"version-history":[{"count":2,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/15579\/revisions"}],"predecessor-version":[{"id":18072,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/15579\/revisions\/18072"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/15578"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=15579"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=15579"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=15579"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}