LobeChat

Connect LobeChat (now LobeHub) from its settings or with self-hosting environment variables, using the OpenAI, Anthropic or Google request format

LobeChat is now called LobeHub: an open-source, self-hostable AI chat app. These steps were checked against the interface text and the official docs of LobeHub v2.3.3 (released 2026-11-12).

Get an API key#

Create one at qd.zsopc.com/dashboard/keys (sk-gpushare- plus 64 hex characters; you can re-reveal it on its detail page at any time).

Set it up in the app#

  1. App Settings → AI Providers → OpenAI
  2. Fill in:
    • API Key: your sk-gpushare-...
    • API Proxy URL: https://qd.zsopc.com/v1 (with /v1)
  3. Leave Use Client Request Mode off (see "Things to know" below)
  4. Click Check to test the connection, then Fetch models to pull the model list, or Add Model to enter model IDs by hand (such as claude-sonnet-4-6, gpt-6.1-sol or deepseek-v4.1-flash)

To list this platform as a provider of its own, use Add Custom Provider: any Provider ID (it can't be changed later) and Provider Name, the Request Format and Proxy URL from the table below, and the same key as API Key.

Request formatProxy URLSuited to
OpenAIhttps://qd.zsopc.com/v1every text model
Anthropichttps://qd.zsopc.com (no /v1)the Claude family
Googlehttps://qd.zsopc.com (no /v1beta)the Gemini family

The OpenAI request format alone reaches Claude and Gemini models too — the gateway translates the protocol for you, see the compatibility matrix.

Self-hosting: configure with environment variables#

When you self-host, you can preset the provider on the server so users don't each enter a key:

OPENAI_API_KEY=sk-gpushare-xxx
OPENAI_PROXY_URL=https://qd.zsopc.com/v1
OPENAI_MODEL_LIST=-all,+claude-sonnet-4-6,+gpt-6.1-sol,+deepseek-v4.1-flash

# Optional: Claude and Gemini over their native protocols
ANTHROPIC_API_KEY=sk-gpushare-xxx
ANTHROPIC_PROXY_URL=https://qd.zsopc.com
GOOGLE_API_KEY=sk-gpushare-xxx
GOOGLE_PROXY_URL=https://qd.zsopc.com

In OPENAI_MODEL_LIST, -all hides the built-in models and +id adds one. Configure the database, auth, object storage and the other variables per the official Docker deployment guide (image lobehub/lobehub, port 3210).

Things to know#

  • Client request mode sends requests from the browser straight to the API, and this platform's API does not allow cross-origin calls from third-party web pages, so they fail. Leave it off and let the LobeHub server forward the requests.
  • Use Responses API Specification switches to /v1/responses, which only some models support (see the compatibility matrix); leave it off when you mix vendors.
  • Fetch models returns every model your key can call, including image and video models that don't work in a chat.
  • If replies come back empty, first check that the OpenAI proxy URL ends in /v1.

Troubleshooting#

SymptomWhat to check
401 invalid_api_keyWhether the key was copied in full (76 characters); re-reveal it on the key detail page
Check fails / empty repliesThe OpenAI proxy URL should be https://qd.zsopc.com/v1; the Anthropic and Google formats take no version
400 model_not_foundModel ID spelling — it must match Models exactly
402 quota_exceededBalance exhausted. Every key shares one account balance — top up on the Top up page
CORS or network errors in the browserTurn off client request mode

Full error reference: Error codes.

Other clients#