Troubleshooting

Hit a problem? Start here

First locate where to go by symptom, then expand the high-frequency issues below to check them off. No duplicated content — this only points you to the authoritative page.

API errors (4xx / 5xx)

Complete error codes, response structure, and retry strategy — 401 / 402 / 404 / 429 / 5xx at a glance.

Account, Key & authentication

Sign-up, lost API Key, troubleshooting an invalid Key, and common issues with models and streaming calls.

Billing & quota

Free allowance, subscription plans, 鹿贝 deduction rules, and what to do after quota runs out.

Client can’t connect

Integration and troubleshooting for third-party clients such as Cherry Studio / Lobe Chat / OpenCat.

Product-specific issues

Platform, MemX, Lucrum, and other products each have their own FAQ — check the FAQ page in the corresponding product docs first.

Common symptoms

Troubleshoot by error

Expand the error you ran into and follow the checklist. For detailed error codes, see Error handling.

Returns 401 Unauthorized / invalid_api_key

json

{ "error": { "code": "invalid_api_key", "type": "authentication_error" } }

authentication_error means the Key is invalid or missing. Check each item:

The Key is complete, starts with sk-, and has no extra spaces or line breaks (copy it again)
Request header format Authorization: Bearer sk-xxxx (one space after Bearer)
The Key status is "Enabled" (Console → Token management)
The environment variable name is spelled correctly and has been loaded

Do not retry a 401 — fix the Key first, then send again. See Authentication and FAQ: How to troubleshoot an invalid Key.

Returns "no available server" / model_not_found (HTTP 404)

json

{ "error": { "code": "model_not_found", "message": "模型 xxx 无可用渠道", "type": "new_api_error" } }

Check the spelling of the model name (case-sensitive)
Confirm this Key has permission to access the model
This model may currently have no available channel
If you just created the Key, wait about 10 seconds and try again

For the list of available models, see Supported models.

Returns 429 Too Many Requests / rate_limit_exceeded

json

{ "error": { "code": "rate_limit_exceeded", "type": "rate_limit_error" } }

Rate limit exceeded. How to handle it:

Lower the request rate and retry with exponential backoff of 2 ** attempt seconds
Free defaults to 60 RPM; upgrade to Pro / Team to raise the limit
If you still hit it frequently after paying, contact support@lurus.cn

For retry code examples, see Error handling · Best practices.

Returns 402 / insufficient_quota (quota / balance insufficient)

json

{ "error": { "code": "insufficient_quota", "type": "billing_error" } }

First confirm whether you’ve used up today’s free quota (Free plan: 100 calls/day)
Check your 鹿贝 balance: identity.lurus.cn/wallet
Top up yourself or upgrade your plan; for the rules, see Billing

context_length_exceeded (context length exceeded)

json

{ "error": { "code": "context_length_exceeded", "type": "invalid_request_error" } }

For example deepseek-chat is 64K and gemini-3-pro-preview is 1M; when you exceed the model’s limit:

Reduce the input and trim history messages
Use a sliding window (keep system + the most recent N turns)
Switch to a model with a longer context

Request times out / no response for a long time

Check network connectivity: curl https://hub.lurus.cn/v1/models
Reduce max_tokens
Reasoning models (deepseek-reasoner) take a long time to think, which is normal
The SDK’s default timeout is about 60 seconds; you can increase timeout
Persistent timeouts may be an upstream failure — try another model

Didn’t find it? Contact support@lurus.cn

Please include: the full error message, the request ID (response header X-Request-ID), the time it occurred, and reproduction steps, so we can locate it quickly.

Hit a problem? Start here

Troubleshoot by error

Related docs