Error Handling

Artifact Chat uses a conversational approach to error handling, displaying errors as part of the natural conversation flow rather than disruptive alerts. This design pattern, often called the “Claude pattern,” is used by leading AI assistants like ChatGPT, Gemini, and Claude.

​

Error Display Philosophy

When an AI streaming error occurs, Artifact Chat displays it as an assistant message bubble in the conversation. This approach:

• Feels Conversational: The error is part of the dialogue, not a system interruption
• Preserves Context: You see your query and the error response together
• Less Alarming: Errors feel like “I couldn’t help with that” vs. “SYSTEM ERROR”
• Natural Retry UX: The retry button appears inline, like continuing the conversation

​

Error Types

Artifact Chat categorizes errors into four main types:

​

Network Errors

Connection issues, server unreachable, or timeouts. Common Causes:

• Internet connection issues
• Server temporarily unavailable
• Request timeout

User Message: “I wasn’t able to complete this response. Please check your connection and try again.”

​

Rate Limit Errors

Too many requests or quota exceeded. Common Causes:

• Exceeded API rate limits
• Usage quota reached
• Too many concurrent requests

User Message: “I’m receiving too many requests right now. Please wait a moment and try again.”

​

Safety Errors

Content policy violations or safety filter triggers. Common Causes:

• Content policy violation
• Safety filter triggered
• Inappropriate request

User Message: “I wasn’t able to respond to that request. Please rephrase your question.”

​

General Errors

Unknown or uncategorized errors. Common Causes:

• Unexpected server error
• Unknown issue

User Message: “Something went wrong. Please try again.”

​

Retry Logic

Artifact Chat automatically determines whether an error is retryable:

​

Retryable Errors

These errors can typically be resolved by trying again:

• Network connectivity issues
• Timeout errors
• Server 5xx errors
• Temporary service disruptions

When an error is retryable, a Retry button appears in the error message.

​

Non-Retryable Errors

These errors require different action:

• Rate limits (wait before retrying)
• Authentication errors (check credentials)
• Content policy violations (modify request)

For non-retryable errors, no retry button is shown. Instead, guidance is provided on what action to take.

​

User-Friendly Messages

All error messages shown to users are non-technical and helpful:

​

Handling Errors as a User

When you encounter an error:

​

For Network Errors

1. Check your internet connection
2. Wait a few seconds
3. Click the Retry button

​

For Rate Limit Errors

1. Wait for the suggested cooldown period
2. Reduce request frequency
3. Try again after waiting

​

For Safety Errors

1. Review your request
2. Rephrase to avoid potentially problematic content
3. Submit a new request

​

For General Errors

1. Try the request again
2. If it persists, try a different approach
3. Report the issue if it continues

​

Error Recovery Tips

​

Best Practices

1. Don’t Panic: Most errors are temporary and resolve quickly
2. Read the Message: Error messages provide guidance on what to do
3. Use Retry When Available: The retry button handles everything automatically
4. Wait for Rate Limits: Respect rate limit cooldowns to avoid extended delays
5. Rephrase if Needed: Safety errors often just need a different approach

​

Next Steps