Troubleshooting
Start with your application's error, the request time and the model. Keep your API key out of anything you share.
Authentication errors
HTTP 401 means Divyam refused the key. Send the complete key, starting with divyam-sw-v1-, as Authorization: Bearer <key>. The email code and the key name don't work as keys. Check the key isn't revoked and your application loaded the latest one. If you lost the key, follow Rotation.
If you can't sign into the console, check your email address and request a new code.
Error codes
| Error | What to do |
|---|---|
| HTTP 401 | See Authentication errors. |
| HTTP 402 | Credit has run out, so retrying won't help. See Credits and billing. |
| HTTP 403 | Your plan doesn't allow this request or model. Email hello@divyam.ai. |
| Model not found | Copy the full model ID, including divyam-as/, from Models & Pricing. |
| HTTP 404 for the URL | Match the base path or request path to the API you use. See Protocol URLs. |
| HTTP 429 | Too many requests. Retry after the delay in the response. |
| HTTP 5xx | A model or service is unavailable. Read the message and retry later. |
| Old connection still used | Check for settings that override the base URL, key or model, then restart the application. |
Protocol URLs
Keep the inference host from your connection settings. An SDK takes its base path and appends an operation. A direct HTTP call takes the full request path for that API:
| API | SDK base path | Direct HTTP request path |
|---|---|---|
| OpenAI Chat Completions | /v1 | /v1/chat/completions |
| OpenAI Responses | /v1 | /v1/responses |
| Anthropic Messages | /anthropic | /anthropic/v1/messages |
| Gemini | /gemini | /gemini/v1beta/models/divyam-as/gpt-6-sol:generateContent |
Use the base path for your selected SDK, rather than adding an operation the SDK appends itself. For Claude Code, follow its connection guide.
Missing logs or usage
- Check the application got its answer or error from the Divyam base URL.
- Sign into the console with the account that owns the key.
- In Logs, choose All models, All statuses and a period covering the request (UTC). Select Load older requests if needed.
- Wait a few minutes and refresh, because Usage and Credits & Billing lag behind Logs.
Logs may not show requests that Divyam rejects before they reach a model, such as requests with a bad key. Keep the application's error for those.
Report an unresolved problem
If you're still stuck, email hello@divyam.ai with the request time and timezone, the model ID, the HTTP status, the request ID if you have one, and the error message with your API key removed.
Also say which application failed and what it was doing.