Open WebUI v0.12.3 Released with E2EE Chats and Redis Streaming

Open WebUI v0.12.3 Released with E2EE Chats and Redis Streaming

At a Glance

Item Value
Repository open-webui/open-webui
Version v0.12.3
Published 2026-10-16
License NOASSERTION
Source type Primary source (the publisher itself)

Values determined by this site’s code when the information was collected. Dates are JST.

Overview

The latest version “v0.12.3" of open-webui, an open-source software that provides an AI interface, has been released. open-webui is a Python-based WebUI environment for intuitively operating local and remote LLMs such as Ollama and OpenAI API.

The highlights of this update are the addition of end-to-end encrypted (E2EE) chat features for direct messages and private channels, and a significant reduction in streaming overhead using Redis. This is a beneficial update for administrators looking to enhance security requirements in multi-user collaborative environments or engineers operating large-scale instances.

Key Changes

Addition of End-to-End Encrypted Chats via MLS Protocol

It is now possible to start temporary end-to-end encrypted chats in direct messages and private group channels. Messages are encrypted in the browser using MLS (Messaging Layer Security), and text up to 16 KiB is relayed without being stored on the server side. Participants in the chat can compare verification codes to verify each other, and typing indicators for the other party are also supported. Logs are immediately erased upon window reload or exit.

This affects teams that need to handle highly confidential communications. This feature is disabled by default. To use it, administrators must enable the setting in Admin Settings > General > Encrypted channels or set the environment variable ENABLE_ENCRYPTED_CHANNELS, and grant user permissions (USER_PERMISSIONS_FEATURES_ENCRYPTED_CHANNELS). Additionally, HTTPS and disabling WebSocket server logging are required in production environments.

Significant Reduction in Streaming Processing and Subscription Load when Connected to Redis

Data writing processes during streaming generation have been significantly revised in configurations using Redis as the backend. Instead of rewriting the entire response to a shared hash on every update as before, the behavior has been changed to prepare an independent key for each response and append differential text. Furthermore, when room channels are enabled, each instance now subscribes only to the rooms of users connected to itself, eliminating the reception of unrelated live updates.

This directly impacts administrators running scale-out environments using multiple instances or Redis clusters. Unnecessary traffic between instances and Redis load are greatly reduced. Compatibility is also ensured, allowing rolling updates while maintaining streaming responses started on older instances.

Scoped-Down Request Permissions for Google Drive Integration

Permission requests when using the Google Drive file picker have been improved. The traditional request for read-only access to the entire drive (drive.readonly) has been changed to request the drive.file scope, which allows access only to files explicitly selected by the user.

This affects administrators who have configured Google Drive integration in self-hosted environments. Unlike the previous scope, Google’s restricted scope security review is no longer required, so the warning “App hasn’t been verified by Google" will no longer appear during integration.

Model Queries from File Browser and Enhanced Markdown Preview

When text is selected in a file opened via the terminal file browser, an “Ask or edit…" popup appears, allowing users to send instructions or modification requests to the model with the file attached. Additionally, the preview display for Markdown files has been unified with the chat response rendering engine, adding support for math formulas, footnotes, Mermaid diagrams, and code highlighting.

This is useful for users who perform code editing and document referencing using models directly within the browser. It has also become possible to update files directly by clicking checkboxes in task lists.

Bug Fix: Overwriting Public Settings by Users with Collaborative Edit Permissions

A bug has been fixed where collaborators with write permissions for shared knowledge bases, prompts, and tools would unintentionally remove the public sharing setting set by the owner, making it private when editing and saving names or descriptions. Even if users update content whose sharing scope they cannot change with their own permissions, existing sharing settings are safely maintained.

This benefits teams managing knowledge and tools collaboratively among multiple people. No special reconfiguration is required after updating.

Bug Fix: Accidental Message Sending Upon IME Conversion Confirmation in Japanese and Other Languages

An issue has been fixed in channel input fields where pressing the Enter key to confirm text conversion candidates using IMEs (Input Method Editors) for Japanese, Chinese, Korean, etc., would accidentally send and clear the message simultaneously.

This affects all users utilizing the channel feature in Japanese input environments. Applying the update allows for smooth input without the stress of accidental message sending.

Bug Fix: PDF Export Failures for Long Texts

A bug has been fixed where exporting lengthy chats or notes as PDFs caused errors such as “Canvas exceeds max size" in Firefox, or generated entirely black PDFs in Chrome. This improves usability for users who routinely archive long text logs or create reports.

How to Get It

Two methods are provided for installing or updating open-webui: using Python’s package management tool pip, or using Docker containers. Choose the appropriate procedure according to your environment and LLM infrastructure (such as Ollama) configuration.

Installation and Update via Python pip

When installing directly into a Python environment, use the pip command. To avoid compatibility issues, the use of Python 3.11 is recommended as the execution environment.

To install and update to the latest package, run the following command:

pip install open-webui

After installation is complete, launch the Open WebUI server with the following command:

open-webui serve

Once launched, you can access the management screen and chat functions by visiting http://localhost:8080 in your browser.

Installation and Update via Docker

When building or updating your environment using Docker containers, you must include the persistent volume option -v open-webui:/app/backend/data in the command to prevent the loss of database data. The Web port for access is set to http://localhost:3000 by default.

1. Integrating with Ollama Running on a Local PC

If you are running Ollama on the same machine, configure it to assign the host network and run the following command:

docker 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

2. Integrating with Ollama on an External Server

When connecting to Ollama running on a separate server, specify the destination URL in the environment variable OLLAMA_BASE_URL.

docker run -d -p 3000:8080 -e OLLAMA_BASE_URL=https://example.com -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main

3. Using OpenAI API Only

If you do not use local models and rely solely on the OpenAI API, start by setting the environment variable OPENAI_API_KEY.

docker run -d -p 3000:8080 -e OPENAI_API_KEY=your_secret_key -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main

4. Using Nvidia GPU (CUDA)

To use GPU acceleration, install the Nvidia CUDA container toolkit on your host system (Linux/WSL) and run with the :cuda tag and --gpus all option.

docker run -d -p 3000:8080 --gpus all --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:cuda

5. Using the Ollama-Bundled Image (All-in-One)

A pre-built image that starts Open WebUI and Ollama together in a single container is also available. Specify -v ollama:/root/.ollama as the data storage destination for Ollama as well.

  • When using GPU:
    bash docker run -d -p 3000:8080 --gpus=all -v ollama:/root/.ollama -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:ollama

  • When using CPU only:
    bash docker run -d -p 3000:8080 -v ollama:/root/.ollama -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:ollama

When updating containers in a Docker environment, follow the official update guides and other resources to fetch the latest image and restart while preserving persistent data (the open-webui volume).

Releases Since Our Last Article

Compiled by Local Model Watch from the project’s GitHub releases: the versions between this release and the last one we covered, which did not get separate articles. Full history: release tracker.

Version Released Release notes
v0.12.2 2026-10-15 GitHub

Related Articles

What to Read Next

Sources