Quota Limit Exceeded Error is not handled properly.

[arhen]

Description

When using opencode with Github Copilot provider quota is reached max limit, there is no anything happened neither notification errors.
It stuck on Generating . . . loading:

Logs:

OpenCode version

v0.15.23

Steps to reproduce

1. Login using GitHub copilot with reached max limit token
2. Prompt whatever

Screenshot and/or share link

No response

Operating System

macOS 26.x

Terminal

Ghostty

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• Opencode Silent on Reaching Provider Quota #3106: Exact same issue - OpenCode gets stuck on "generating" when provider quota is reached without showing error notifications (this was already closed, suggesting a fix was implemented)
• Feature Request: Show Error Message When Claude Usage Limit Is Reached #877: Same symptoms but for Claude provider - stuck at "working..." when usage limit reached (also closed)
• More resilient retry policy for Amazon Bedrock rate limiting #452: Similar quota/rate limiting issues with AWS Bedrock, requests more resilient retry policy

Feel free to ignore if none of these address your specific case.

[berenar]

Yes I've seen this happen to me too. While #768 would allow the user to see what happened, it would not fix it

[CoolCold]

I've just hit the same issue, namely with Gemini and 429 / rate limit exceeded isssue
This type of things are handled quite well by Aider ( https://github.com/Aider-AI/aider ) and somewhat by Crush ( charmbracelet/crush#536 (comment) )

• Aider does exponential backoff - retrying in 2, 4, 8..seconds with clear visible indication
• Crush currently just reports the issue and stops - which is yet better than infinite retry
• From what I see, Gemini API replies with the time interval when you can repeat your query back again - can be taken as a hint I guess

Opencode keeps going for hours, like:

• 
• 

Note, that ratelimits can happen on both free and paid subscription on Gemini/Google

❯ opencode --version
1.0.45

My naive understanding of recent changes, in Zen ratelimiter - 30b1ae5#diff-8ab8205b374f3d1907779abb668ee07eb06a6d8ba82bd5ce4991005375d01d6cR102 says that currently the way is to implement it per provider , not as a generic phase of request handling and I don't see the retry with backoff - I may be totally wrong here of course.

[rekram1-node]

@CoolCold very valid points and both of these will be fixed soon

the visual part is something we have been missing wayy too long

and the other part will be fixed first thing tmr

[burka]

I hit those limits with my Cerebras.ai plan which will return a 429 until midnight each time and opencode just stalls. That feels really awkward :-/

[rekram1-node]

Dax has this fixed should ship soon I think

[burka]

When I am out of quota it just stalled yesterday, but today I saw this beautiful message :-) <3

[rekram1-node]

Yay!

[burka]

Love it! Don't love the message, but love the way opencode shows it! Thanks a lot! <3

[rekram1-node]

I think the message comes from whatever provider u are using