Skip to main content
Open WebUI is a self-hosted, ChatGPT-style workspace with persistent conversation storage and native support for OpenAI-compatible APIs. This guide shows how to connect Open WebUI directly to SambaNova’s OpenAI-compatible endpoints (with an optional LiteLLM proxy) so you can deliver fast inference from a private, customizable UI.
This guide was tested with Open WebUI v0.6.x. If you’re using a different version, some UI paths or features may vary.

​
Prerequisites

  • SambaCloud account and API key.
  • Python 3.11 or 3.12 (Open WebUI does not yet support Python 3.13+). Open WebUI recommends the uv runtime manager, but pip works too.
  • Optional: Docker for containerized deployment.
  • Optional: LiteLLM proxy credentials if you plan to route SambaNova traffic through LiteLLM instead of connecting directly.

​
Setup

1

Install Open WebUI

2

Run Open WebUI

The server launches at http://localhost:3000. For persistent storage, set DATA_DIR:
3

Create the admin account

  1. Open your browser to http://localhost:3000.
  2. Click Get Started.
  3. Follow the onboarding flow to set the admin email, password, and optional team name.
This account governs model visibility, team access, and connection settings.
Open WebUI admin register panel
4

Connect to SambaCloud

  1. Click your profile icon and go to Settings → Admin Settings → Connections.
  2. Select and edit the OpenAI-compatible option.
  3. Enter your SambaCloud base URL (https://api.sambanova.ai/v1) and API key.
  4. Click Save.
Open WebUI automatically fetches the available SambaNova models.
Open WebUI register provider panel
5

(Optional) Connect through LiteLLM proxy

If you already run LiteLLM as a unified gateway, add a LiteLLM connection pointing to your proxy host (for example, http://0.0.0.0:4000/v1) and use the proxy’s API key. SambaNova-backed models registered in LiteLLM appear alongside the direct connection.
6

Test the connection

  1. Start a new chat and select a SambaNova model (for example, Llama-4-Maverick-17B-128E-Instruct).
  2. Send a test prompt.
  3. Refresh the page to verify conversation persistence.
  4. Test tagging and export options.
  5. View the Chat Overview diagram to confirm token counts and tool usage.
Open WebUI chat panel

​
Chat capabilities

Open WebUI provides several features that enhance your SambaNova experience:
  • Model switching: Switch between SambaNova models mid-conversation without leaving the chat.
  • Multimodal prompts: Send images to SambaNova’s Maverick models or attach documents for quick RAG experiments.
  • Conversation controls: Tag, edit, duplicate, or export (JSON/Markdown) any thread for auditing or sharing.
  • Audio: Enable audio input/output in Settings → Admin Settings → Audio → OpenAI. Enter your SambaCloud base URL, API key, and whisper model (Whisper-Large-v3).
  • Code interpreter: Use the built-in code interpreter to prototype scripts with SambaNova models.
  • Document knowledge base: Upload PDFs or text files so Open WebUI’s built-in RAG features can ground SambaNova responses.
Whisper-Large-v3 is available on SambaStack deployments only — not on SambaCloud. See Model deprecations.

​
Administration

To access the admin panel, open Admin Panel from your account menu.

​
Users and groups

Create groups, assign members, and restrict which models, tools, or documents they can access.

​
Connections and permissions

Manage available models per team. Promote SambaNova models in model settings to “public” so every user sees the same curated list, or add models to specific user groups.

​
Feature toggles

Control access per group to plugins, image uploads, audio, and experimental features from Admin Settings.

​
Tools and community extensions

Browse the Open WebUI community catalog for tools, plugins, and automations that enhance SambaNova-powered workflows. To import remote tools (OpenAPI or MCP), add their hosted URL in Settings → Admin Settings → External Tools. You can make tools public or share them within specific teams.

​
Troubleshooting

​
Additional resources