678APIdocs
Troubleshooting
Troubleshooting
Use this checklist when a client cannot call the API, returns errors, or consumes credits unexpectedly.
Check Base URL
Use https://www.678api.com/v1 and avoid duplicate /v1/v1 or /chat/completions.
Check API key
Make sure the key is active and has no extra spaces.
Check credits
Your wallet or subscription must have available credits.
Check model name
Use a model listed in your console.
Common errors and fixes
| Error | Likely cause | Fix |
|---|---|---|
| 401 / unauthorized | Wrong key, disabled key, or hidden whitespace | Copy the key again, create a test key, and avoid duplicating the Bearer prefix. |
| 403 / no permission | Your group or route cannot access the model | Use a model available in your console or ask the administrator to confirm group permissions. |
| 404 / model not found | Wrong model name | Copy the exact model name, including case, hyphens, and version suffixes. |
| 429 / rate limited | Concurrency, frequency, or upstream route limit | Lower concurrency, reduce context, wait and retry, or use separate keys per client. |
| Credits exist but request fails | Pre-charge, subscription quota, request limit, or model permission issue | Check wallet, subscription, and usage logs; retry with a shorter prompt. |
| Timeout | Network, proxy, slow model, or long context | Test with a short prompt, check local proxy, reduce max tokens, and retry. |
| Streaming interrupted | Client streaming compatibility or connection interruption | Disable streaming for a test, upgrade the client, reduce output length, or switch model. |
| Old configuration remains active | Client cache, stale environment variables, or CC Switch did not apply | Restart the client and terminal; inspect environment variables; recreate the provider if needed. |
| Unexpectedly high usage | Long context, repeated history, Agent mode, or tool calls | Clear conversation history, reduce context, disable broad codebase indexing, and inspect usage logs. |
| Model list cannot refresh | Client does not call model list or expects a different URL format | Fill model names manually; use https://www.678api.com/v1, or https://www.678api.com if the client asks for a root URL. |
Quick self-check
curl https://www.678api.com/v1/models -H "Authorization: Bearer YOUR_API_KEY"If the model list returns, your network and API key are basically working. Then focus on client Base URL, model name, and local cache.