Frequently Asked Questions
Accounts, models, billing, troubleshooting — expand by topic.
Accounts & Authentication
How do I sign up?
At hub.lurus.cn, enter your email and password (or sign in with GitHub/Google) to automatically receive 5 Lubei + a free quota. All products share the same account.
Lost your API Key?
It is shown only once and cannot be recovered. In the console, delete the old one and create a new one. Store it in a password manager or environment variable — never hardcode it. Each account can create multiple Keys; assigning a separate Key per project is more secure.
How do I troubleshoot an invalid Key?
- Key is complete (starts with
sk-and has no missing characters) - Status is "enabled" (console → Token Management)
- Request header is
Authorization: Bearer sk-xxxx(one space after Bearer) - No extra spaces or line breaks (copy it again)
- Environment variable name is spelled correctly and is loaded
Still getting 401? Work through the checklist item by item — see Troubleshooting · invalid_api_key.
Models & Requests
Which models are supported?
OpenAI, Claude, Gemini, DeepSeek, and more — see the model list.
Returns "no available server"
Check the model name; confirm the Key has permission for that model; the model may currently have no available channel — contact an administrator.
How do I switch models?
Just change the model parameter; leave everything else unchanged.
How do I enable streaming responses?
Set "stream": true, and the response is returned chunk by chunk over SSE.
Billing & Quota
How do I check my usage?
In the console, see "Dashboard" or "Usage Logs".
Out of quota?
Contact an administrator to top up or upgrade your plan.
Where can I see model pricing?
See the pricing in the model list.
Technical Issues
What should I do about request timeouts?
- Check the network (
curl https://hub.lurus.cn/v1/models) - Reduce
max_tokens - Reasoning models (
deepseek-reasoner) take longer to think — this is normal - The SDK default timeout is about 60 seconds; you can increase
timeout - Persistent timeouts may indicate an upstream failure — switch models
429 error (Rate LimitRate LimitA cap on the number of requests allowed per unit of time, used to prevent abuse and overload. exceeded)
Lower the request rate + retry with exponential backoff (see Error Handling); Free defaults to 60 RPM, upgrade to Pro/Team to raise the limit; if you still hit it frequently on a paid plan, contact support@lurus.cn.
Context limit exceeded (e.g. deepseek-chat 64K, gemini-3-pro-preview 1M)
- Reduce input by trimming history
- Sliding window (keep the system message + the most recent N turns)
- Switch to a longer-context model
- Summarize very long documents before passing them in
Other Questions
Is my data secure?
HTTPS end to end; conversation content is not stored; only call metadata is recorded for billing.
Is there an SLA guarantee?
Enterprise customers can sign an SLA — contact sales.
What are the technical support channels?
support@lurus.cn / GitHub Issues.
Didn’t find your answer?
Send us your question and we’ll reply within a business day.