Description:**
A user tries to send a standard message in chat, but the assistant does not respond successfully. Instead, the conversation returns an error card and a retry option. The visible backend/provider-related error suggests that the configured model endpoint or API credential is not usable for inference because the connected account has no credits or the wrong account/org key is being used.
This is not just a provider error exposure issue; it also impacts core chat usability because the system does not gracefully recover, route to another valid model, or provide a cleaner actionable message.
Steps to Reproduce:**
Open the Chat page.
Select the current active assistant/model configuration.
Enter a normal prompt, for example: please tell me about yourself
Send the message.
Observe the response area.
Expected Result:**
The system should generate a valid AI response successfully.
If the selected model/provider is unavailable or misconfigured, the platform should either:
route the request to a valid configured model, or
fail over to a fallback model, or
show a clean user-friendly message without exposing raw provider/internal error details.
Chat should remain usable as a core conversation feature.
Actual Result:**
The response fails.
UI shows:
“I was unable to complete your request. Please try again.”
“Something went wrong”
“There was an error generating a response. Please try again.”
Detailed error exposed in UI:
“Uncaught Error: Insufficient credits. This account never purchased credits. Make sure your key is on the correct account or org...”
User cannot continue normal conversation with that configuration.
Description:**
A user tries to send a standard message in chat, but the assistant does not respond successfully. Instead, the conversation returns an error card and a retry option. The visible backend/provider-related error suggests that the configured model endpoint or API credential is not usable for inference because the connected account has no credits or the wrong account/org key is being used.
This is not just a provider error exposure issue; it also impacts core chat usability because the system does not gracefully recover, route to another valid model, or provide a cleaner actionable message.
Steps to Reproduce:**
Open the Chat page.
Select the current active assistant/model configuration.
Enter a normal prompt, for example: please tell me about yourself
Send the message.
Observe the response area.
Expected Result:**
The system should generate a valid AI response successfully.
If the selected model/provider is unavailable or misconfigured, the platform should either:
route the request to a valid configured model, or
fail over to a fallback model, or
show a clean user-friendly message without exposing raw provider/internal error details.
Chat should remain usable as a core conversation feature.
Actual Result:**
The response fails.
UI shows:
“I was unable to complete your request. Please try again.”
“Something went wrong”
“There was an error generating a response. Please try again.”
Detailed error exposed in UI:
“Uncaught Error: Insufficient credits. This account never purchased credits. Make sure your key is on the correct account or org...”
User cannot continue normal conversation with that configuration.