There was an error processing your request: Custom error: prompt is too long: 207379 tokens > 200000 maximum

[Ridoy302583]

Describe the bug

When project is too Big at that time provide this error. Is it possible to bypass this error? (Already package-lock.json removed but still same issue)

Link to the Bolt URL that caused the error

hh

Steps to reproduce

1. When project is too big
2. Provide this error.

Expected behavior

When project is too Big at that time it will works perfectly.

Screen Recording / Screenshot

ddd

Platform

• OS:Windows

Provider Used

antrophic

Model Used

No response

Additional context

No response

[Ridoy302583]

@xKevIsDev Please look into it.

[xKevIsDev]

this is purely an issue with the provider/model context limit. You can however try a workaround which is to export the project and start a new project using the exported zip (make sure to unzip first)

[Ridoy302583]

Target file and Lock file can bypass this error??? @xKevIsDev

[btothec]

I get a similar error when using Deepseek-R1 via DS or Nvidia API.
error details: custom error: the input (165236 tokens) is longer than the model's context length (163840 tokens).

This is on the initial prompt that I get this error. I can make the exact same request to a local model with 32768 context and everything is fine. I think it has something to do with the amount of thinking that is done with R1.

[xKevIsDev]

initial prompt should use no more than 10k tokens this is including all starter template files etc, can you pull the latest changes and let me know if the issue persists?

[Stijnus]

Can you test : #2001