Skip to main content
This guide shows how to turn any self-hosted Open WebUI instance into a usage-based, billable chat experience powered by OpenStack.
You’ll point Open WebUI’s OpenAI-compatible connector to OpenStack, ensure every chat request carries the right user identity for metering, and end up with automatic per-user pricing, rate limits, and reporting—without changing your users’ workflow.Open WebUI with OpenStack

​
Prerequisites

  • OpenStack API key (Dashboard → Paywall → API keys)
  • Open WebUI admin access (so you can edit Connections and Filters)

​
1. Add OpenStack as the OpenAI-compatible Connection

1

Open the Connections page

In Open WebUI click your avatar → Admin Panel → Connections → + New Connection.Open WebUI Connections
2

Configure the endpoint

Fill the form with the OpenStack details:

  • Provider: OpenAI Compatible - Base URL: https://api.openstack.ai/v1
    • API Key: sk-openstack-... (copied from the dashboard)
    • Default model: pick one of the OpenStack backed models, e.g. openai/gpt-4o-mini or leave empty to load all models dynamically.
Save the connection and (optionally) set it as the default for your workspace so every model inherits it automatically.
Open WebUI Configure Endpoint

​
2. Capture user identity for accurate billing

OpenStack meters usage per end user. Every request must include a stable user identifier via the body-level user field or the X-Openstack-User header (see Pass user ID for the full rationale).
Open WebUI filters can rewrite the JSON payload that is sent to every model. We’ll add one that copies the signed-in user’s WebUI ID into the OpenStack-required user field so every completion and tool call is attributed (and billed) correctly.Open WebUI Filters
1

Add a new filter

Go to Admin Panel → Functions → Filters → + New Filter, name it OpenStack user injector, choose Python, and paste the snippet below.
import re
from pydantic import BaseModel, Field

class Filter:
  class Valves(BaseModel):
      prefix: str = Field(
          default="webui",
          description="Prepended to every user handle to keep it unique per deployment.",
      )

  def __init__(self):
      self.valves = self.Valves()

  def inlet(self, body: dict, __user__: dict) -> dict:
      # numeric WebUI user id with prefix and safe chars
      login = f"id_{__user__.get('id')}"
      safe = re.sub(r"[^A-Za-z0-9_-]", "_", login) or "unknown"
      body["user"] = f"{self.valves.prefix}_{safe}"
      return body
Open WebUI passes the signed-in user’s metadata via user when running filters (see Filter Function docs). Returning the modified body makes the change effective for the upstream API request.
2

Attach the filter to your models

  • 1. Go to Workspace → Models and open each chat model that should charge through OpenStack.
    1. Scroll to Tools / Filters / Actions, enable Filters, and add OpenStack user injector to the active list.
  • 3. Save the modelfile. Repeat for every model that should hit OpenStack.
Open WebUI Attach Filter

​
3. Test the integration

1

Start a new chat

Open WebUI main interface and start a new chat with a model that uses the OpenStack connection.
2

Send a prompt

Send a short prompt, e.g. Hello!. The model should respond normally, but if your OpenStack balance is zero you’ll get a topup message instead.
3

Top up and check billing log

Once you top up your OpenStack balance, send the message again. In the OpenStack dashboard → Paywall → Reports view you should see the billing log with the user you injected (e.g. id_1234567).OpenStack Billing Log

Pass user ID

Full overview of all supported OpenStack user identity mechanisms.

Workspace → Models

Open WebUI’s reference for managing models, filters, and tools.