Open WebUI¶

Feature-rich, self-hosted web interface for interacting with LLMs. Provides a ChatGPT-like experience with support for multiple models, RAG, tools, and collaborative features.

Overview¶

Open WebUI serves as the primary user interface for the starter kit: - Chat Interface: Conversational UI with streaming, markdown, and code highlighting - Multi-Model: Switch between models via LiteLLM integration - RAG: Built-in document upload and retrieval-augmented generation - Tools & Functions: Extensible pipe functions for custom integrations - User Management: Multi-user support with roles and permissions - Conversation History: Persistent chat history and conversation export - Model Presets: Save and share model configurations

Installation¶

Prerequisites¶

Configure in .env:

LITELLM_API_KEY=sk-1234567890abcdef
OPENWEBUI_ADMIN_EMAIL=admin@example.com
OPENWEBUI_ADMIN_PASSWORD=your-password

Install¶

./cli gui-app openwebui install

The installer: 1. Adds the Open WebUI Helm chart repository 2. Renders Helm values with LiteLLM connection and admin credentials 3. Deploys Open WebUI via Helm chart (version 12.8.1) 4. Creates the openwebui namespace

Verification¶

# Check pods
kubectl get pods -n openwebui

# Check service
kubectl get svc -n openwebui

# Check ingress
kubectl get ingress -n openwebui

# Port-forward for local access
kubectl port-forward svc/openwebui 3000:8080 -n openwebui --address 0.0.0.0 &

# Access Open WebUI
open http://localhost:3000

Log in with OPENWEBUI_ADMIN_EMAIL and OPENWEBUI_ADMIN_PASSWORD.

Configuration¶

Environment Variables¶

LiteLLM Backend¶

Open WebUI connects to LiteLLM as its backend:

OPENAI_API_BASE_URL=http://litellm.litellm:4000/v1
OPENAI_API_KEY=sk-1234

All models configured in LiteLLM (vLLM, SGLang, TGI, Ollama, Bedrock) appear in the Open WebUI model selector.

OpenClaw Integration¶

Open WebUI integrates with OpenClaw via pipe functions to enable AI agent capabilities directly from the chat interface.

Features¶

Document Upload (RAG)¶

1. Click the attachment icon in the chat input
2. Upload a document (PDF, TXT, DOCX, etc.)
3. Ask questions about the uploaded document
4. Open WebUI uses embedded models to retrieve relevant content

Custom Pipe Functions¶

Extend Open WebUI with custom Python functions:

# Example: OpenClaw integration pipe
class Pipe:
    def __init__(self):
        self.valves = Valves(
            OPENCLAW_URL="http://openclaw.openclaw:8080",
            OPENCLAW_TOKEN="openclaw-gateway-token"
        )

    def pipe(self, body):
        # Route request to OpenClaw agent
        pass

Troubleshooting¶

Cannot connect to models¶

# Check LiteLLM is running
kubectl get pods -n litellm

# Check Open WebUI environment
kubectl exec -it -n openwebui <openwebui-pod> -- env | grep OPENAI

# Check LiteLLM connectivity from Open WebUI
kubectl exec -it -n openwebui <openwebui-pod> -- \
  curl http://litellm.litellm:4000/v1/models

Login fails¶

# Check admin credentials in values
kubectl get secret -n openwebui -o yaml

# Check Open WebUI logs
kubectl logs -n openwebui -l app=openwebui

Slow responses¶

# Check LLM backend health
kubectl get pods -n vllm
kubectl get pods -n sglang

# Check Open WebUI pod resources
kubectl top pod -n openwebui

Learn More¶

• Open WebUI Documentation
• Open WebUI GitHub
• Open WebUI Helm Charts
• Pipe Functions Guide