From 157cece6b9067fb9d9d2124773230807c9daba39 Mon Sep 17 00:00:00 2001 From: zangguodong <237974119@qq.com> Date: Mon, 2 Mar 2026 11:21:24 +0800 Subject: [PATCH 1/2] feat: add openweb ui doc --- .../How_to_Deploy_and_use_OpenWebUI.md | 142 ++++++++++++++++++ 1 file changed, 142 insertions(+) create mode 100644 docs/en/solutions/How_to_Deploy_and_use_OpenWebUI.md diff --git a/docs/en/solutions/How_to_Deploy_and_use_OpenWebUI.md b/docs/en/solutions/How_to_Deploy_and_use_OpenWebUI.md new file mode 100644 index 00000000..137e78db --- /dev/null +++ b/docs/en/solutions/How_to_Deploy_and_use_OpenWebUI.md @@ -0,0 +1,142 @@ +--- +products: +- Alauda AI +kind: +- Solution +ProductsVersion: +- 4.x +--- + +# OpenWebUI + +## Overview +OpenWebUI is an open-source AI Web interface that supports docking with multiple OpenAI protocol-compatible inference backends (such as vLLM, MLServer, XInference, etc.) through a unified entry point. It is used for scenarios such as text generation, multimodal input, and voice input. It provides an extensible external tool mechanism to facilitate the integration of retrieval, function calling, and third-party services. It is suitable for deployment in containers locally or in the cloud, supporting persistent data and Ingress-based HTTPS access. + +## Basic Features +- **Conversation & Text Generation**: Support system prompts, adjustable parameters (temperature, length, etc.), and session management. +- **Multimodal & Voice**: Images/documents as context, voice input/transcription (dependent on backend capabilities). +- **External Tool Extension**: Can call retrieval, databases, HTTP APIs, etc., to build tool-enhanced workflows. +- **Data & Security**: Sessions and configurations can be persisted; can integrate with authentication, rate limiting, logging/monitoring. + +## Backend Integration +- **Protocol Compatibility**: Support OpenAI API style backends (such as vLLM, MLServer, XInference, TGI, etc.). +- **Connection Parameters**: Base URL (e.g., `http(s)://{backend}/v1`), API Key, model name, and default inference parameters. +- **Multiple Backends**: configured in the UI, allowing switching between different inference service backends. + +## Deployment Scheme +Create the following resources in order. In this case, choose an independent `open-webui-ns` namespace. You can choose an available namespace as needed. + +### Namespace +```bash +kubectl create ns open-webui-ns +``` + +### Create the specific deployment +```yaml +apiVersion: apps/v1 +kind: Deployment +metadata: + labels: + app: open-webui + name: open-webui + namespace: open-webui-ns +spec: + replicas: 1 + selector: + matchLabels: + app: open-webui + template: + metadata: + labels: + app: open-webui + spec: + volumes: + - name: webui-data + emptyDir: {} + containers: + - image: ghcr.io/open-webui/open-webui + name: open-webui + ports: + - containerPort: 8080 + env: + - name: ENABLE_DIRECT_CONNECTIONS + value: "true" + - name: OPENAI_API_BASE_URL + value: http://example-predictor/v1 # REPLACE with actual inference service URL + - name: PORT + value: "8080" + volumeMounts: + - name: webui-data + mountPath: /app/backend/data + resources: + requests: + cpu: 1000m + memory: 128Mi + limits: + cpu: 2000m + memory: 1Gi +``` + +## Important environment values + +Relative environment values should be configured. + +### ENABLE_DIRECT_CONNECTIONS +* Set to true to enable external connections. +* Purpose: Allows adding additional external inference service backends within OpenWebUI. + +### OPENAI_API_BASE_URL +* Specifies the default inference service endpoint. +* If OpenWebUI and the inference service are deployed in the same cluster, use the service’s internal cluster address. +* For the address details, refer to: **AML Business View / Inference Service / Inference Service Details / Access Method**. +* Value format: `{{Cluster Internal URL}}/v1`. + + +### Verification +```bash +kubectl get deployment open-webui -n open-webui-ns -w +``` +Wait until the deployment status is `1/1 Ready`. + +## Access OpenWebUI + +### 1. View OpenWebUI via NodePort Service +Create the following resource: + +```yaml +apiVersion: v1 +kind: Service +metadata: + labels: + app: open-webui + name: svc-open-webui + namespace: open-webui-ns +spec: + type: NodePort + ports: + - port: 8080 + protocol: TCP + targetPort: 8080 + selector: + app: open-webui +``` +Check the relevant port and node IP to access the page. + +### 2. Initial Settings +When accessing OpenWebUI for the first time, you need to register. Choose a strong password for the administrator account. + +### 3. Add Inference Service +Go to **Settings -> Connections -> Add Connection**. +Here you will be required to add the inference service address. +You can obtain the cluster external access methods via **AML Business View / Inference Service / Inference Service Details / Access Method**. +Fill it in afterwards. Please use the cluster **external** access method. +In the **Add Connection** popup, fill in: +`{{Cluster External URL}}/v1` + +Click the icon on the right to verify connectivity. After success, click save. Return to the chat page to select the existing inference service for use. + +### 4. Use Inference Service +Enter the chat page, select the uploaded inference service, and explore more features, such as: +- Voice input +- Multimodal input +- External tools From cef963d16f5387b61ec90a914fcb043f006e594f Mon Sep 17 00:00:00 2001 From: zangguodong <237974119@qq.com> Date: Fri, 3 Apr 2026 23:25:23 +0800 Subject: [PATCH 2/2] fix: improve --- docs/en/solutions/How_to_Deploy_and_use_OpenWebUI.md | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) diff --git a/docs/en/solutions/How_to_Deploy_and_use_OpenWebUI.md b/docs/en/solutions/How_to_Deploy_and_use_OpenWebUI.md index 137e78db..1ad480a5 100644 --- a/docs/en/solutions/How_to_Deploy_and_use_OpenWebUI.md +++ b/docs/en/solutions/How_to_Deploy_and_use_OpenWebUI.md @@ -19,7 +19,7 @@ OpenWebUI is an open-source AI Web interface that supports docking with multiple - **Data & Security**: Sessions and configurations can be persisted; can integrate with authentication, rate limiting, logging/monitoring. ## Backend Integration -- **Protocol Compatibility**: Support OpenAI API style backends (such as vLLM, MLServer, XInference, TGI, etc.). +- **Protocol Compatibility**: Support OpenAI API-compatible backends (such as vLLM, MLServer, XInference, TGI, etc.). - **Connection Parameters**: Base URL (e.g., `http(s)://{backend}/v1`), API Key, model name, and default inference parameters. - **Multiple Backends**: configured in the UI, allowing switching between different inference service backends. @@ -85,6 +85,14 @@ Relative environment values should be configured. * Set to true to enable external connections. * Purpose: Allows adding additional external inference service backends within OpenWebUI. +### Security Considerations +**ENABLE_DIRECT_CONNECTIONS**: When set to `true`, users can configure OpenWebUI to connect to external inference services. Consider the following: +- Only enable this in trusted environments or when users are authenticated and authorized +- External connections may expose sensitive data or credentials +- Monitor outbound connections to prevent data exfiltration +- Consider using network policies to restrict egress traffic +- For production environments, consider setting this to `false` and pre-configuring allowed backends + ### OPENAI_API_BASE_URL * Specifies the default inference service endpoint. * If OpenWebUI and the inference service are deployed in the same cluster, use the service’s internal cluster address. @@ -129,7 +137,7 @@ When accessing OpenWebUI for the first time, you need to register. Choose a stro Go to **Settings -> Connections -> Add Connection**. Here you will be required to add the inference service address. You can obtain the cluster external access methods via **AML Business View / Inference Service / Inference Service Details / Access Method**. -Fill it in afterwards. Please use the cluster **external** access method. +Fill it in afterward. Please use the cluster **external** access method. In the **Add Connection** popup, fill in: `{{Cluster External URL}}/v1`