Open WebUI is an open source LLM user interface that supports Ollama and OpenAI-compatible APIs and is suitable for unified management of multi-model access, knowledge bases, and dialogue workflows within the team.

OpenWebUI

Core parameters and statistics of Open WebUI

Open WebUI is positioned as a "user-friendly, self-hosted AI platform that can run offline". Its core value lies in integrating multi-model access to the RAG knowledge base, team permission management and plug-in extensions into the same interface, avoiding the team from duplicating the construction of front-end and session management on each model platform.

Projects	Public Information
Product positioning	Self-hosted AI platform / multi-model unified interface
Code repository	open-webui/open-webui
Main languages	Python (backend), Svelte (frontend)
License Information	Open WebUI License (with brand retention clause + historical MIT components)
Community size	146k+ stars, 21.1k+ forks, 827+ contributors
Latest version	v0.10.2 (2026-07-04)
Deployment form	Docker / pip / Kubernetes / Desktop App
Model access	Ollama, OpenAI compatible API, LMStudio, vLLM, GroqCloud, etc.
Vector database	ChromaDB, PGVector, Qdrant, Milvus, Elasticsearch, Pinecone, etc. 9

Capability Boundary: Open WebUI is a unified interface for the "model usage layer" and "team collaboration layer" and does not directly participate in model training, fine-tuning or data annotation processes. Its knowledge base capability relies on external model embedding and does not provide model training or distillation functions. For lightweight personal scenarios that only require a single model and a single user, the friction of using Ollama WebUI or ChatGPT directly is lower. The permission management and plug-in system of Open WebUI become a deployment burden.

User and market recognition of Open WebUI

Open WebUI's recognition mainly comes from the adoption scale of the open source community and continuous high-frequency version iterations. There is no official disclosure of the number of paying enterprise customers or commercial revenue data.

GitHub ecological indicators: 146k+ stars and 21.1k+ forks make it occupy the top position in the LLM web interface track. The participation of 827 contributors shows that the project has healthy community maintenance capabilities. The release rhythm of 166 Releases and one version every 2 weeks reflects that the project is in a period of rapid feature expansion.

Adoption model: The typical adoption path starts with "One-click Docker deployment + Ollama local model" for individuals/small teams, and then gradually adds LDAP/OAuth authentication RBAC permissions, knowledge base and plug-in extensions after verifying the use experience, and finally becomes a department-level or enterprise-level AI portal. This "bottom-up" adoption model is similar to similar products such as Dify and MaxKB, but the advantage of Open WebUI lies in its deep integration with Ollama and richer plug-in system.

Organizational Adoption Notes: Due to the lack of public enterprise customer lists and commercial SLA descriptions, it is recommended to conduct a 2-4 week POC in the target environment before purchasing decisions, focusing on verifying: the stability of the LDAP/OIDC integration, RBAC matching with the expected organizational structure, and the database and Redis performance when large-scale user concurrency occurs.

Cost Advantages of Open WebUI

C-side/Individual: Zero license fee + own infrastructure

Open WebUI itself is completely open source and can be deployed by individual users for free through Docker or pip without any subscription fees. The only explicit cost is the maintenance overhead of the server (or local hardware). For users who already have Ollama or OpenAI API Key, building Open WebUI is equivalent to obtaining a complete AI interface with session management, knowledge base and network search at zero marginal cost.

Developer/API: Reduce duplication of front-end construction

When a team already holds multiple model APIs (e.g. OpenAI, Claude, native Ollama), Open WebUI can serve as a unified gateway to reuse the front end and session management layer. Compared with independently developing chat UI for each model, Open WebUI eliminates the repetitive work of front-end development, session storage, multi-model switching, etc., indirectly saving about 40-80 man-days of development resources. However, the access cost still needs to be evaluated: if the team has in-depth customization requirements for the interface (complete white label, customized workflow), the cost of self-development may be lower than secondary development of Open WebUI.

Enterprise / Privatization: Data is autonomous and controllable, and there are thresholds for operation and maintenance

Privatized deployment is the core value of Open WebUI in enterprise scenarios: data does not leave the intranet, meeting the data compliance requirements of finance, medical, government and other industries. However, its hidden costs of operation and maintenance cannot be ignored:

Cost Dimensions	Estimate Range	Description
Infrastructure	5,000-20,000 yuan/month	1-3 GPU servers or cloud hosts, depending on the amount of concurrency
Operation and maintenance manpower	0.5-1 person/month	Docker/host maintenance, database backup, version upgrade
Vector database service	0-3,000 yuan/month	Use PGVector to build your own for free; Pinecone/Qdrant cloud service requires payment
Technical support	Community support is free	Enterprise SLA requires contacting Enterprise Plan
Upgrade risk	High	Versions containing database migration (such as v0.9.0->v0.10.0) need to be tested and verified before going online

Hidden benefits: The unified model entrance can reduce the team's adaptation costs caused by model switching; RBAC and audit capabilities meet compliance review; the combination of knowledge base and dialogue can accumulate the team's tacit knowledge. These benefits are difficult to quantify directly, but typically become apparent 3-6 months after adoption.

Main functions of Open WebUI

The functional system of Open WebUI revolves around the four capability lines of "unified access -> knowledge enhancement -> team collaboration -> ecological expansion". It is not a simple chat interface encapsulation.

• Multi-model unified access and gateway proxy: Simultaneously access Ollama local models and any OpenAI compatible API (including LMStudio, vLLM, GroqCloud, Mistral, OpenRouter, etc.). Administrators can configure default models, user permissions, and model visibility in the backend. Synergy: The model access layer is linked with RBAC - different user groups can see different model lists, without the need for separate configuration on each model platform.

• Local RAG Knowledge Base: Built-in search enhances generation capabilities, supporting 9 vector databases (ChromaDB, PGVector, Qdrant, Milvus, Elasticsearch, OpenSearch, Pinecone, S3Vector, Oracle 23ai) and multiple content extraction engines (Tika, Docling, Mistral OCR, PaddleOCR-vl, MinerU). Supports hybrid retrieval (BM25 + vector) and reordering. Hidden linkage: The knowledge base can cooperate with the directory synchronization tool oikb to continuously and incrementally synchronize from 45+ data sources such as GitHub, Confluence, SharePoint, Notion, etc., to achieve "source update -> automatic refresh of knowledge base -> automatic reference of the latest content in conversations".

• Plug-in and Extension System: Supports 5 types of plug-in primitives - Filters (request/response filtering), Actions (interface action expansion), Pipes (pipeline conversion), Tools (tool functions), Skills (skill presets). The new Event function can hook into system events (login, file upload, role change, etc.) and trigger custom logic. Synergy: The plug-in system is linked with MCP (Model Context Protocol), OpenAPI tool server, and Event event system to realize enterprise-level scenarios such as approval flow embedding dialogue, alarm routing, and data outgoing auditing.

• Team collaboration workspace: Channels (channels) support real-time collaboration among multiple users and model response flow; Calendar provides schedule management and AI scheduling; Notes provides rich text notes and context injection into conversations; Automations support scheduled tasks (daily reports, data summary, regular reminders); Memories persists user preferences and is inherited across conversations. Expert View: These functions are not independent modules - Calendar can be linked with Automations to generate periodic summary tasks; Notes can Attach to Chat as an additional context; model replies in Channels can trigger Webhook events - when used in combination, the efficiency is doubled, but the configuration complexity also increases linearly.

• Desktop desktop application: The official native desktop App (Mac/Windows/Linux) released in v0.9.0, which can be run without Docker or configuration. Built-in system-level shortcut call bar (Shift+Cmd+I), global Push-to-Talk, and built-in llama.cpp offline reasoning engine. Value Point: Upgrading self-hosted AI from "browser bookmarks" to "operating system-level tools" lowers the threshold for use by non-technical team members.

• Voice/Video Calling and Multimodality: Integrated WebRTC voice/video calling, multiple STT (Whisper, Deepgram, Azure) and TTS (ElevenLabs, OpenAI, Azure) engines. Supports image generation (DALL-E, ComfyUI, AUTOMATIC1111) and image editing. Applicable Boundary: Call quality depends on the network environment and the deployment level of self-built STT/TTS services. Call stress testing is required before large-scale use.

Model and version evolution of Open WebUI

The version iteration rhythm of Open WebUI is about 2-3 Mondays, and in 2026 it will enter a stage of significantly improved functional density. The following is a summary of the main version history by milestones.

Recent mainline releases

Version number	Release date	Key changes
v0.10.2	2026-07-04	Streaming reasoning display, folder batch upload to knowledge base Memories system context switch STT request format optional Provider failure event API contextual variable configuration
v0.10.1	2026-06-27	Shared folder read-only session repair
v0.10.0	2026-06-27	Folder sharing and RBAC, long conversation automatic context compression Computer Agent integration, hybrid search speed-up External Knowledge Base, redo Memories system Event function and Webhook event system LDAP/OAuth management background configuration, chat import permission MCP connection timeout configuration
v0.9.6	2026-06-02	Official knowledge base synchronization tool oikb, knowledge base folder and file renaming, directory intelligent synchronization, knowledge base file system tool Emoji selector, per-conversation Skills switch, user/group access rights preview MinerU file type configuration, multi-instance model list deduplication writing
v0.9.5	2026-05-11	Security hardening project: SSRF redirection protection Iframe CSP, channel streaming inference and tool support Markdown rendering granularity control, multiple permissions repairs
v0.9.4	2026-05-09	Browser rollback fix: chat scroll position regression correction
v0.9.3	2026-05-09	Voice mode mute control, prompt word list speed, conversation history loading optimization, calendar creation stream Playground control panel MCP OAuth supports Brave LLM Context search, rich text editing optimization, assistant reply editing and continuation LaTeX copy shortcut keys
v0.9.2	2026-04-24	PaddleOCR-vl document extraction Firecrawl v2, custom API Key Header, OAuth session disconnection, asynchronous DB driver migration to psycopg v3, Docker ARM64 fix CSS virtualization instead of JS truncation
v0.9.1	2026-04-21	Dependency package repair (aiosqlite, asyncpg missing)
v0.9.0	2026-04-21	Milestone version: Desktop App release Automations scheduled task Calendar workspace task management tool Responses API support, multi-model dialogue Ollama model management PWA enhanced Mistral TTS, SCIM 2.0 preset OpenTelemetry observability Redis horizontal expansion support

Evolution features and upgrade strategies

Function density acceleration: v0.9.0 to v0.10.2 These 7 versions (about 2.5 months) have introduced major functions such as desktop App, Automations, Calendar, Event system Computer Agent, folder sharing, etc., indicating that the project is in a period of rapid growth. But this also means:

• Versions containing database migration (such as v0.9.0->v0.10.0) must be backed up first and then upgraded. Multi-node deployments must be upgraded at the same time. Rolling upgrades are not supported.
• API and configuration items (contextual variables) may undergo incompatible changes across major versions. It is recommended to test the context verification before upgrading the production instance.
• For production scenarios, it is recommended to fix major versions, only do timely follow-up on minor versions (such as v0.10.1->v0.10.2), and arrange a 1-2 week test window before major version upgrades (such as v0.9.x->v0.10.x).

Technical advantages of Open WebUI

The technical architecture of Open WebUI does not pursue "large model capabilities", but focuses on "standardization layer for multi-model access" and "infrastructure layer for team collaboration". Its value is particularly prominent in the context of fragmented model ecology.

Plug-in primitive system: hierarchical design of 5 extension points: Open WebUI defines 5 plug-in primitives (Filters, Actions, Pipes, Tools, Skills), plus the new Event function, covering the entire link from request preprocessing to system event response. Compared with LangChain's chain abstraction or Dify's workflow engine, Open WebUI's plug-in system is more "middleware mode" - each conversation is an HTTP request, and plug-ins can inject logic at any node on the request link. The advantage of this design is intuitive debugging and good failure isolation; the disadvantage is that complex multi-step workflow (multi-step conditional branching, human-computer approval) is not as intuitive as Workflow engines such as Dify, and is more suitable for "plug-in combination" rather than "process orchestration" scenarios.

Multi-layer pluggable design of RAG engine: The RAG link is split into seven stages: "Document extraction->Blocking->Embedding->Vector storage->Retrieval->Reordering->Context injection", each stage can be plugged and replaced. Supports 9 vector databases, 6+ document extraction engines, and multiple search strategies (BM25, vector search, hybrid search). The engineering value of this design is that enterprises can continue to use existing vector databases (such as Elasticsearch or Pinecone) without migrating; the document extraction engine can be configured differentially according to document types (such as Mistral OCR for scanned documents and Tika for ordinary PDFs).

Productization extension of the ecosystem: Open WebUI is not just a single application, but a complete product ecosystem built around it:

• Open Terminal: Provides a sandbox code execution environment for the model and supports self-hosted script running.
• Terminals Enterprise: User-level isolation execution context based on Docker/K8s.
• cptr: A mobile-first independent Computer Agent that can be used as the model upstream of Open WebUI.
• oikb: Knowledge base synchronization gateway, connecting to 45+ data sources.
• Desktop App: Native desktop, built-in llama.cpp inference engine.

This "core + satellite" architecture is unique among open source LLM interface projects. Compared with similar products:

Dimensions	Open WebUI	Dify	MaxKB
Core positioning	Multi-model unified interface + team collaboration	LLMOps workflow platform	Enterprise knowledge base Q&A
Plugins/Extensions	Filters/Actions/Pipes/Tools/Skills/Event	Workflow Nodes + Plugins	Limited
Multi-model access	Ollama + any OpenAI compatible API	Model provider abstraction layer	OpenAI compatible
Desktop	Official native App	None	None
Horizontal expansion	Redis + multiple workers	Yes	Limited
Operational complexity	Medium (Docker or pip, including database migration)	Medium-high (Docker + Workflow configuration)	Low
Most suitable for scenarios	Unified AI portal for teams	Complex AI workflow orchestration	Internal enterprise knowledge Q&A

Mechanism -> Effect -> Applicable Scenarios:

• Mechanism: Through the three-layer extension architecture of plug-in primitives + MCP protocol + Event event system, the "model dialogue" is changed from a closed chat box to a programmable platform.
• Effect: The same set of deployment can be used to create a simple chatbot, or it can also use the Event function to realize the automated onboarding process of automatically sending a Welcome message for user registration + creating a default channel + assigning initial permissions.
• Applicable scenarios: Medium and large teams that need flexible expansion, or organizations that need to integrate multiple business systems on a self-hosted platform.

How to use Open WebUI

Open WebUI provides multiple usage paths from zero-threshold desktop to production-level Kubernetes deployment.

Installation method comparison

Method	Command/Step	Suitable Scenario	Remarks
Docker (default)	docker run -d -p 3000:8080 -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main	Quick trial, individual/small team	Visit http://localhost:3000
Docker + Ollama built-in	docker run -d -p 3000:8080 -v ollama:/root/.ollama -v open-webui:/app/backend/data --name open-webui ghcr.io/open-webui/open-webui:ollama	Scenarios that require one-click inclusion of local models	Contains built-in Ollama
pip installation	pip install open-webui && open-webui serve	Already a Python-savvy developer	Requires Python 3.11
Docker Compose	Refer to the official docker-compose.yaml	Scenarios that require customized networks, volumes, and multiple containers	PostgreSQL can be superimposed
Kubernetes	Helm / Kustomize	Production-level multi-node deployment	Requires Redis for session management
Desktop App	Download the installation package from the official website (Mac/Win/Linux)	Non-technical users, no configuration requirements	Built-in llama.cpp, available offline

Access model

1. Existing Ollama: Automatically identify the host Ollama after starting the container (--add-host=host.docker.internal:host-gateway).
2. Cloud API only: Set the contextual variable OPENAI_API_KEY=your_key, or add an OpenAI compatible endpoint in the admin after logging in for the first time.
3. Hybrid Mode: Connect to local Ollama + cloud API at the same time, Open WebUI automatically summarizes into a model selection list.

Quickly enable knowledge base

• Create a knowledge base from the management panel -> Select a vector database -> Upload documents (PDF/DOCX/TXT/images, etc.) -> The system automatically completes Embedding.
• Use the # command to reference the knowledge base in a conversation, or have the model automatically retrieve it when needed.
• Use the oikb tool to connect to Confluence, Notion, GitHub, etc. to achieve continuous synchronization.

Configure team permissions

• Enable LDAP or OAuth/OIDC authentication -> Create user groups -> Assign model visibility, knowledge base access and function permissions to different groups.
• The event function can be used to realize the automated permission allocation process after user registration.

Product Pricing for Open WebUI

Open WebUI adopts a dual-track pricing model of "open source community version free + Enterprise Plan paid".

Community Edition (Free)

• License: Open WebUI License (including brand retention terms, the "Open WebUI" brand logo must be retained).
• Capability: All core functions are fully available - multi-model access to RAG knowledge base, plug-in system Automations, Calendar, Channels, etc.
• Restrictions: No official SLA, no LTS version, no dedicated technical support.

Enterprise Plan (Business Pricing)

The official enterprise plan is provided, including custom themes and brand customization SLA guarantees LTS long-term support version. The specific price has not been disclosed, and you need to contact sales through official channels (docs.openwebui.com/enterprise).

Summary of hidden costs

Projects	Community Edition	Enterprise Plan
Software License	Free (subject to brand retention requirements)	Pending commercial confirmation
Infrastructure	Prepare your own server/cloud host	Same as left, or optional official hosting (undisclosed)
Operation and maintenance	Self-responsible (upgrade, backup, monitoring)	Includes SLA guarantee
Commercial Authorization	Brand Retention Terms	Customizable White Label
Technical Support	GitHub Issues / Discord Community	Dedicated Support

Application scenarios of Open WebUI

Unified AI entrance for the team (quantified cost reduction and efficiency increase)

Scenario Description: Unify multiple models held by the company (Ollama local model + OpenAI API + Claude API + self-developed model) into one interface, so team members do not need to switch between different platforms.

Cost reduction and efficiency improvement deduction:

• Operation and Maintenance Team: Managing one Open WebUI instance vs maintaining N model independent front ends + N sets of certification systems -> From approximately one person-day/week of multi-platform maintenance to 0.5 person-days/week of single-instance operation and maintenance.
• Business users: No need to remember multiple model URLs and API Keys -> Model switching time is reduced from about 2 minutes/time to 10 seconds/time.
• IT managers: Configure model access permissions for all employees at once through RBAC, rather than managing them independently on each platform.

Human-machine collaboration boundary:

• Sections that can be 100% automated: model API availability detection, knowledge base incremental synchronization, scheduled Automations execution.
• There are areas that must be confirmed manually: replacement and rotation of model API keys, permission changes of high-privilege roles, creation and sharing of knowledge bases involving sensitive data, and database migration across major versions.

Enterprise intranet knowledge Q&A platform

Scenario description: Build an AI knowledge base based on internal documents (technical documents, rules and regulations, product manuals, project summaries), and employees can get answers with quotes by asking questions in natural language.

Implementation Tip: The quality of the knowledge base depends on the standardization of the upstream documents. It is recommended to start the pilot on a small scale (such as the technical team's Wiki) and establish a comprehensive process of "document update->oikb synchronization->knowledge base automatic refresh->user Q&A verification", and then expand to the entire company. For scanned documents and handwritten PDFs, it is recommended to use Mistral OCR or MinerU for in-depth document extraction. The accuracy of pure Tika extraction may not meet production requirements.

Unfit Boundary: For scenarios that require 100% answer accuracy (such as legal opinions, medical diagnosis suggestions), the AI ​​knowledge base can only be used as an auxiliary reference and may not independently output decision-making conclusions.

R&D and operation and maintenance automation assistance

Scenario description: Through the Automations function of Open WebUI, set up scheduled tasks to automatically summarize Jira work orders, generate daily reports, check system alarms, send notifications, etc.

Cost reduction and efficiency improvement deduction:

• Junior Operations Engineer: Write and maintain automated scripts -> From about 4 hours/week of manual inspections to 0.5 hours/week of script maintenance.
• Technical Director: Manually summarize the progress of each team every day -> Automatically generate daily reports through Automations -> Review from about 30 minutes/day to 5 minutes/day.

Human-machine collaboration boundary:

• Can be 100% automated: data summary, alarm analysis, daily report generation, scheduled reports.
• Manual confirmation is required: change operations involving production (such as pushing configurations to the production environment), customer communication content sent externally, and compliance audit reports.

More subdivided scenes

• Education Scenario: Teachers build private AI teaching assistants, and students ask questions through the course knowledge base to avoid the risk of data leakage caused by using public models.
• Medical and Scientific Research: Meet HIPAA/data compliance requirements through privatized deployment, AI-assisted literature review and experimental data analysis.
• Multi-language team collaboration: Using Open WebUI's multi-language i18n support, multi-language teams such as Chinese, English, Japanese and Korean can share the same platform.

Applicable groups of Open WebUI

• Technical team leader/IT decision-maker: It is necessary to provide the organization with a unified model access entrance and permission management capabilities. Adaptation value: Open WebUI's RBAC, LDAP/OAuth integration and audit logs can help organizations meet compliance review requirements while reducing management costs caused by model access fragmentation. Not suitable for boundaries: If an organization only uses a single model (such as only ChatGPT) and has no need for team collaboration or permission control, Open WebUI brings more operational complexity than the problems it solves.

• Platform and Operations Engineer: Responsible for the construction and daily maintenance of the self-hosted AI platform. Adaptation value: The Docker/Kubernetes deployment method is compatible with the existing operation and maintenance stack; OpenTelemetry observability can be integrated with the monitoring system; the multi-Worker + Redis architecture supports horizontal expansion. Implementation Tips: It is recommended to plan a version upgrade strategy - before major version upgrades (such as v0.9.x->v0.10.x), the database migration script must be tested and verified, and the database must be backed up before the upgrade window.

• Business Department AI Pilot Team: Hope to experience multi-model capabilities with low threshold and deploy private knowledge base within the department. Adaptation value: Desktop App eliminates the Docker threshold; one-click import of knowledge base + dialogue reference reduces the understanding cost of AI applications. Not suitable for the boundary: For the need for deeply customized workflows (such as multi-step conditional branching, human-machine approval process), Open WebUI's plug-in system is not as intuitive as Dify's Workflow engine. It is recommended to evaluate before selecting.

• Indie Developer/Individual Technology Enthusiast: An individual interested in Ollama local model and large model application deployment. Adaptation value: Completely free Docker one-click deployment, active community and fast response to issues. Prerequisites: At least one machine (or cloud server) capable of running Docker is required, as well as basic Docker operation capabilities.

Summary and Outlook

The core competitiveness of Open WebUI is to use open source and plug-in architecture to package "multi-model access + knowledge base enhancement + team collaboration" into a unified interface, and build around it a supporting ecosystem such as desktop, code execution context, and knowledge base synchronization tools. In an era of fragmented model ecology, the value of this "standardized access layer" will amplify as the number of model types and APIs owned by an organization increases.

Current core advantages: The community size (146k stars) and version rhythm (2 weekly versions) prove the vitality of the project; the plug-in system (5 primitives + Event function + MCP) provides flexible expansion capabilities; the multi-database and multi-engine support of the RAG engine adapts to the stock infrastructure of different organizations; the Desktop App lowers the deployment threshold for non-technical users.

Main current limitations: There is a lack of public enterprise customer cases and SLA information, and enterprises need to verify their own POC before purchasing; rapid iteration means that upgrade costs and risks of API incompatibility exist objectively; the plug-in system favors a middleware model, and complex workflow orchestration is not as intuitive as Dify; the brand retention clause requires that all derivative deployments must retain the "Open WebUI" logo, and a complete white label requires an Enterprise Plan.

Follow-up observation points: Whether the public pricing and SLA details of the Enterprise Plan will be launched; whether a mature community market can emerge in the plug-in ecosystem (similar to the VSCode extension market); whether the maturity of supporting projects such as oikb and Computer Agent can support enterprise-level scenarios.

Procurement and Adoption Risk Assessment: For individuals and small teams, the zero license fee Community Edition has no real risk and is worth using as the default UI for local models. For enterprises, it is recommended to follow the three-step process of "single-department pilot (4 weeks) -> cross-department expansion (4-8 weeks) -> organization-wide promotion": in the first stage, use Docker to quickly deploy, access 1-2 core models and a small knowledge base, and verify functional integrity and user acceptance; in the second stage, enable LDAP/OAuth authentication and RBAC permissions, load more knowledge bases and test multi-user concurrent performance; in the third stage, after confirming the upgrade strategy and operation and maintenance process, consider whether to upgrade the Enterprise Plan. Before purchasing an Enterprise Plan, you need to confirm with the official: the brand-customized support range, the update cycle and security fix commitments of the LTS version, and the response time and availability guarantees included in the SLA.

Related tools: DeepSeek, ChatGPT

Version Info

• Open WebUI v0.9.6 ：The latest public version of GitHub Releases continues the iteration of multi-model access and interface stability. （2026-06-01）
• Open WebUI v0.9.5 ：Mainline version update node. （2026-05-10）
• Open WebUI v0.9.4 ：The previous node in successive version iterations. （2026-05-09）