Open WebUI with API429: server environment, model allowlist and background task controls

Prepare the server connection and shared key boundary

Decide who administers your Open WebUI deployment and which users will receive access to the configured connection before entering credentials. A server-level API key pays for requests made by every authorized user in the workspace, so restrict model and workspace permissions inside Open WebUI first. A personal user key and a shared deployment credential solve different operational problems: never paste a shared secret into staff onboarding notes, exported chat transcripts or public repositories. Record the nonsecret fields of any existing connection before changing environment variables so you can restore the previous state cleanly.

Enable ENABLE_OPENAI_API on the server and set OPENAI_API_BASE_URLS to https://gateway.api429.com/v1. Supply OPENAI_API_KEYS through the protected process environment of your deployment. The reference block below uses the literal placeholder YOUR_API429_KEY, which must never be sent as a live credential. It is an administrator field sheet rather than an installer script, and it does not assume that docker --env-file expands shell variables inside an environment file. Also note that the official Open WebUI environment reference documents a v0.12.1 baseline on the review date rather than proving the behavior of every release.

Expose one explicit model ID in the connection index

In OPENAI_API_CONFIGS, the key "0" addresses the first entry in your semicolon-separated OPENAI_API_BASE_URLS list. Setting enable to true activates that specific connection index, while model_ids restricts the visible list to gemini-3.8-flash. Omit prefix_id during your initial verification so the client sends the exact model identifier without a custom prefix. If your server already defines multiple OpenAI-compatible connections, do not overwrite them blindly with index zero; coordinate the semicolon order of URLs, keys and index objects with your administrator.

Multiple base URLs and multiple keys in Open WebUI environment variables are separated by semicolons, and their positional order maps directly to the numeric keys in OPENAI_API_CONFIGS. For a clean initial test, configure a single base URL and a single key. Declaring a manual model allowlist keeps the selector uncluttered, but it does not grant account entitlement. Always verify that gemini-3.8-flash appears in an authenticated GET /v1/models check for your key in addition to the public catalog, and keep the secret out of diagnostic logs.

Check persistent ConfigVar settings and background tasks

The official Open WebUI reference marks these connection and task settings as persistent ConfigVar values. After the first server startup, values stored in the application database can take precedence over edited environment variables. If changing OPENAI_API_BASE_URLS or OPENAI_API_CONFIGS in your container configuration does not update the active endpoint, inspect the saved connection inside the Admin Panel instead of repeating paid chat prompts. Saving changes in the interface updates persistent state, whereas disabling persistent configuration globally affects the entire installation; you do not need to wipe your database or user history just to update one endpoint.

Conversation title generation, tag generation and follow-up prompt suggestions are separate documented background tasks that can trigger extra model calls. The configuration sheet disables them explicitly with ENABLE_TITLE_GENERATION=False, ENABLE_TAGS_GENERATION=False and ENABLE_FOLLOW_UP_GENERATION=False. TASK_MODEL_PARAMS sets max_tokens to 256 for any remaining background task requests, but it does not cap the main chat completion response. Check the active interface switches as well, because disabling those three variables does not prove that web search, pipelines or third-party extensions cannot issue additional requests.

Send one bounded text prompt and verify usage

Open a fresh standard chat without file uploads, web search, tool calling or custom pipelines, and select gemini-3.8-flash. In the model parameters exposed by your installed version, set max_tokens to 256, disable streaming for the initial check and send one short nonconfidential sentence. The verification JSON below illustrates the target API429 Chat Completions request body rather than an importable Open WebUI settings file. Confirm that your client version actually applies those request parameters; a single click in a web interface should never be assumed to equal exactly one billable API request without checking usage logs.

Compare the returned answer with the usage entry in your API429 account dashboard. Every model generation is a paid request, including any auxiliary calls triggered by the client interface. Capping output tokens with max_tokens limits a single completion response, not your total deployment budget: prompt context length, conversation history growth and multi-user traffic all contribute to total spend. If you encounter a 404 routing error, verify that the final request targets /v1/chat/completions; if you receive a 401 or model error, verify the key and exact ID before sending another manual test.

Keep passthrough disabled and separate advanced features

Leave ENABLE_OPENAI_API_PASSTHROUGH set to False. Standard server-backed chat does not require it, and the upstream reference describes passthrough as a separate broad proxy mode that exposes administrative credentials to direct API forwarding. Manage model visibility and user access independently through Open WebUI permissions. A working text reply confirms only the basic Chat Completions contract; it does not validate function calling, image inputs, structured JSON schemas, agent loops or streaming behavior across every Open WebUI release. Evaluate each advanced feature separately against the API429 contract before enabling it for users.

Server environment reference; API key is a literal placeholder

ENABLE_OPENAI_API=True
OPENAI_API_BASE_URLS=https://gateway.api429.com/v1
OPENAI_API_KEYS=YOUR_API429_KEY
OPENAI_API_CONFIGS={"0":{"enable":true,"model_ids":["gemini-3.8-flash"]}}
ENABLE_TITLE_GENERATION=False
ENABLE_TAGS_GENERATION=False
ENABLE_FOLLOW_UP_GENERATION=False
TASK_MODEL_PARAMS={"max_tokens":256}
ENABLE_OPENAI_API_PASSTHROUGH=False

Target Chat Completions request contract; not an Open WebUI import file

{
  "model": "gemini-3.8-flash",
  "messages": [
    {
      "role": "user",
      "content": "Answer in one sentence: what is an API?"
    }
  ],
  "max_tokens": 256,
  "stream": false
}

Payment and a $10 test

• Russian bank card. Top up your balance in the account.
• Invoice for companies. Bank transfer with closing documents.
• USDT and TON. Top up in crypto.

$10 test credit. Subscribe to our Telegram channel through the bot and get $10 on your balance for the first requests. Get $10 in the bot