LobeChat / LobeHub with API429: OpenAI server env, model list and canary caveats

Check your deployed release tag and canary documentation provenance

Record the exact image tag or release version of your self-hosted LobeChat / LobeHub deployment before changing server variables. This guide targets server-level environment configuration on an instance you administer; logging into a third-party hosted workspace does not let you modify its container environment. If another administrator manages the server, hand over only the nonsecret field sheet and agree on user access boundaries first. Never send a live API key in plain-text chat messages or screenshots.

The official documentation sources reviewed for this guide reside on the mutable canary branch of the upstream lobehub/lobehub repository. Canary documentation tracks active development and can diverge from a pinned stable release tag. Before restarting your service, compare these variable names against the release notes or source tree for your installed tag. Treating canary docs as a universal guarantee for every historical LobeChat build can mask configuration drift.

Configure the OpenAI-compatible server environment

In the environment of the running application process, set ENABLED_OPENAI=1 and set OPENAI_PROXY_URL to https://gateway.api429.com/v1 without appending /chat/completions. Supply your API429 credential via OPENAI_API_KEY through your deployment secret store. Using the OpenAI provider variables selects the OpenAI-compatible Chat Completions transport inside LobeHub; it does not require an OpenAI account key or alter the underlying Google Gemini billing family for gemini-3.8-flash.

In the parameter sheet below, YOUR_API429_KEY is an inert placeholder. Inject the real credential through your container orchestrator or process manager rather than passing it on a public command line or committing it to version control. Confirm that the variables reach the actual LobeHub server process rather than an unrelated interactive shell. The example is a plain key-value reference sheet and makes no assumptions about variable expansion inside docker --env-file.

Hide default models and expose only the literal model ID

Set OPENAI_MODEL_LIST to -all,+gemini-3.8-flash. In LobeHub model list syntax, -all removes the built-in default OpenAI model entries first, and +gemini-3.8-flash adds the exact model ID separated by a standard comma. Keep the identifier free of custom display aliases or angle-bracket capability suffixes during your first test so the UI selector, public catalog entry and wire request field match character for character. Declaring an ID in OPENAI_MODEL_LIST only changes UI visibility; it does not grant key access.

Do not append capability tags such as <fc>, <vision> or <imageOutput> to the model entry. While the upstream canary documentation describes bracketed tags for declaring function calling or multimodal UI controls, adding a tag in LobeHub does not create backend support or verify how your client version serializes those requests. Keep the initial check strictly limited to plain text without plugins, file uploads, web search or agent tools.

Verify catalog access, request parameters and initial response

Restart or redeploy the service using your standard operational workflow, open a clean chat session and select gemini-3.8-flash. If the default OpenAI model list still appears, check whether browser-side settings or cached user preferences are overriding the server environment before sending any prompt. Verify that gemini-3.8-flash is present in the API429 public catalog and separately confirm access for your specific credential via an authenticated GET /v1/models check.

For your first manual message, set max_tokens to 256 and disable streaming if your installed interface exposes those request controls, then send one short nonconfidential question. The JSON block below shows the target API429 Chat Completions wire payload rather than an importable LobeHub configuration file. If your LobeHub version does not expose an explicit non-streaming toggle or max_tokens field in the UI, inspect the actual outbound payload for your release tag rather than assuming the UI slider matches the wire body.

Account for shared server credentials and total spend

Every model generation is a paid API call. Because OPENAI_API_KEY in the server environment acts as a shared credential for all users allowed into your LobeHub instance, protect the deployment with authentication and access controls before opening it to a team. A per-request max_tokens setting bounds only a single completion reply, not your overall budget: multi-turn conversation history, topic naming requests, client retries and concurrent users all add to total cost. If a test fails, inspect the base URL, key and exact model ID before issuing another request.

Server environment reference; API key is a literal placeholder

ENABLED_OPENAI=1
OPENAI_API_KEY=YOUR_API429_KEY
OPENAI_PROXY_URL=https://gateway.api429.com/v1
OPENAI_MODEL_LIST=-all,+gemini-3.8-flash

Target API429 Chat Completions contract; not a LobeHub 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