anomalyco/opencode
1
0
Fork 0
mirror of https://github.com/anomalyco/opencode.git synced 2026-07-22 02:05:26 -04:00
Code Issues 5.9k Packages Projects Releases 19 Wiki Activity

Implement exponential back-off when hitting rate limits #1172

New Issue
Open
opened 2026-02-16 17:29:51 -05:00 by yindo · 12 comments
yindo commented 2026-02-16 17:29:51 -05:00
Owner
Copy Link
Copy Source

Originally created by @BigMitchGit on GitHub (Aug 8, 2025).

Originally assigned to: @thdxr on GitHub.

OpenAI has some brutal RPM rate limits on tier 1, which will often abruptly stop opencode and requires me to manually resume by sending 'continue' message.

An exponential back-off with auto-retry would keep the agent loop running, it'd just take a bit longer to complete when rate limits are hit.

Originally created by @BigMitchGit on GitHub (Aug 8, 2025). Originally assigned to: @thdxr on GitHub. OpenAI has some brutal RPM rate limits on tier 1, which will often abruptly stop opencode and requires me to manually resume by sending 'continue' message. An exponential back-off with auto-retry would keep the agent loop running, it'd just take a bit longer to complete when rate limits are hit.
yindo commented 2026-02-16 17:29:51 -05:00
Author
Owner
Copy Link
Copy Source

@daliusd commented on GitHub (Aug 8, 2025):

If that helps here is error you get when you hit rate limit:

{"type":"error","sequence_number":2,"error":{"type":"tokens","code":"rate_limit_exceeded","message":"Rate limit reached
for gpt-5-mini in organization org-XXXXXXXXXXXXXX on tokens per min (TPM): Limit 200000, Used 124035,
Requested 79017. Please try again in 915ms. Visit https://platform.openai.com/account/rate-limits to learn
more.","param":null}}
@daliusd commented on GitHub (Aug 8, 2025): If that helps here is error you get when you hit rate limit: ``` {"type":"error","sequence_number":2,"error":{"type":"tokens","code":"rate_limit_exceeded","message":"Rate limit reached for gpt-5-mini in organization org-XXXXXXXXXXXXXX on tokens per min (TPM): Limit 200000, Used 124035, Requested 79017. Please try again in 915ms. Visit https://platform.openai.com/account/rate-limits to learn more.","param":null}} ```
yindo commented 2026-02-16 17:29:51 -05:00
Author
Owner
Copy Link
Copy Source

@AlexeyYar commented on GitHub (Aug 8, 2025):

Same here:
{"type":"error","sequence_number":2,"error":{"type":"tokens","code":"rate_limit_exceeded","message":"Rate limit
reached for gpt-5 in organization org-************* on tokens per min (TPM): Limit 30000, Used 14582,
Requested 21045. Please try again in 11.254s. Visit https://platform.openai.com/account/rate-limits to learn
more.","param":null}}

@AlexeyYar commented on GitHub (Aug 8, 2025): Same here: {"type":"error","sequence_number":2,"error":{"type":"tokens","code":"rate_limit_exceeded","message":"Rate limit reached for gpt-5 in organization org-************* on tokens per min (TPM): Limit 30000, Used 14582, Requested 21045. Please try again in 11.254s. Visit https://platform.openai.com/account/rate-limits to learn more.","param":null}}
yindo commented 2026-02-16 17:29:51 -05:00
Author
Owner
Copy Link
Copy Source

@WilliamAGH commented on GitHub (Aug 11, 2025):

Yeah, Cerebras and Groq are currently unusable with Opencode right now still because of this.

Constant errors like:

AI_RetryError: Failed after 4 attempts. Last error: Too Many Requests

@WilliamAGH commented on GitHub (Aug 11, 2025): Yeah, Cerebras and Groq are currently unusable with Opencode right now still because of this. Constant errors like: > AI_RetryError: Failed after 4 attempts. Last error: Too Many Requests
yindo commented 2026-02-16 17:29:51 -05:00
Author
Owner
Copy Link
Copy Source

@JC1738 commented on GitHub (Aug 14, 2025):

Same, AI_RetryError: Failed after 4 attempts. Last error: Requests per second limit exceeded - too many requests sent.

cerebras with Qwen 3 Coder 480B

I can't get through the /init without the being blocked by this error, and unfortunately you can't resume the init.

@JC1738 commented on GitHub (Aug 14, 2025): Same, AI_RetryError: Failed after 4 attempts. Last error: Requests per second limit exceeded - too many requests sent. cerebras with Qwen 3 Coder 480B I can't get through the /init without the being blocked by this error, and unfortunately you can't resume the init.
yindo commented 2026-02-16 17:29:51 -05:00
Author
Owner
Copy Link
Copy Source

@jsugg-qventus commented on GitHub (Sep 11, 2025):

Same here. On each message.

@jsugg-qventus commented on GitHub (Sep 11, 2025): Same here. On each message.
yindo commented 2026-02-16 17:29:51 -05:00
Author
Owner
Copy Link
Copy Source

@sra commented on GitHub (Dec 9, 2025):

After these rate limit stops, what do you type to make it continue? I've not had great luck.. Cerebras is awesome until the brick wall per minute limit

@sra commented on GitHub (Dec 9, 2025): After these rate limit stops, what do you type to make it continue? I've not had great luck.. Cerebras is awesome until the brick wall per minute limit
yindo commented 2026-02-16 17:29:51 -05:00
Author
Owner
Copy Link
Copy Source

@rekram1-node commented on GitHub (Dec 9, 2025):

We have backoff for ratelimits, and we respect the retry-after type headers so once it is ready to retry it should automatically go through

@rekram1-node commented on GitHub (Dec 9, 2025): We have backoff for ratelimits, and we respect the retry-after type headers so once it is ready to retry it should automatically go through
yindo commented 2026-02-16 17:29:51 -05:00
Author
Owner
Copy Link
Copy Source

@jcgottlieb commented on GitHub (Jan 9, 2026):

My experience as of today with opencode latest /connected to the OpenAI API @ Tier 1 with model 5.1-codex-mini (200K TPM rate limit), even VERY SIMPLE requests are rate limited and fail every time, making the OpenAI API unusable without throwing money at a higher tier. Anyone solved this without $++ yet?

@jcgottlieb commented on GitHub (Jan 9, 2026): My experience as of today with opencode latest /connected to the OpenAI API @ Tier 1 with model 5.1-codex-mini (200K TPM rate limit), even `VERY SIMPLE` requests are rate limited and fail every time, making the OpenAI API unusable without throwing money at a higher tier. Anyone solved this without $++ yet?
yindo commented 2026-02-16 17:29:52 -05:00
Author
Owner
Copy Link
Copy Source

@tgrushka commented on GitHub (Jan 14, 2026):

Exponential backoff seems way overkill.

To state the obvious: exponential backoff means OpenCode waits TWICE THE LENGTH OF TIME it waited the last time on every retry.

This means, after only 7 retries, you wait over a minute.
On the 8th retry, you wait 2 minutes.

I think it should be linear, such as retry after 1 second, then 2 seconds, then 3 seconds, ... up to some kind of user configurable maximum such as 10 or 15 seconds.

"Exponential backoff" seems to be the go-to solution for everything nowadays. Even AI suggests it strongly to me for anything and everything network-related, indiscriminately. I don't know where that came from ... perhaps the 1980s, when networks were really unreliable, and retrying a network backup or something made sense to eventually take hours to make a retry attempt. But nowadays? In 2026? Exponential? Maybe in some low priority cloud services such as backups (even then, why exponential and not linear?), but it really hits the user hard in terms of experience in user-facing apps.

@tgrushka commented on GitHub (Jan 14, 2026): Exponential backoff seems way overkill. To state the obvious: exponential backoff means OpenCode waits TWICE THE LENGTH OF TIME it waited the last time on every retry. This means, after only 7 retries, you wait over a minute. On the 8th retry, you wait 2 minutes. I think it should be linear, such as retry after 1 second, then 2 seconds, then 3 seconds, ... up to some kind of user configurable maximum such as 10 or 15 seconds. "Exponential backoff" seems to be the go-to solution for everything nowadays. Even AI suggests it **strongly** to me for anything and everything network-related, indiscriminately. I don't know where that came from ... perhaps the 1980s, when networks were really unreliable, and retrying a network backup or something made sense to eventually take hours to make a retry attempt. But nowadays? In 2026? Exponential? Maybe in some low priority cloud services such as backups (even then, why exponential and not linear?), but it really hits the user hard in terms of experience in user-facing apps.
yindo commented 2026-02-16 17:29:52 -05:00
Author
Owner
Copy Link
Copy Source

@shaneqld commented on GitHub (Jan 27, 2026):

We have backoff for ratelimits, and we respect the retry-after type headers so once it is ready to retry it should automatically go through

Is it possible that this behaviour isn't consistent across providers? I've run into rate limits errors fairly frequently (multiple times per day) using OpenAI models and the agent loop completely stops (ie no retries).

In my case, it happens primarily using gpt-5.2-codex, either directly via OpenAI or via OpenCode Zen.

Example error when using OpenAI provider:

{"type":"error","sequence_number":2,"error":{"type":"tokens","code":"rate_limit_exceeded","message":"Rate limit reached for gpt-5.2-codex in organization org-xyz on tokens per min (TPM): Limit 500000, Used 404104, Requested 107839. Please try again in 1.433s. Visit https://platform.openai.com/account/rate-limits to learn more.","param":null}}

Example error when using OpenCode Zen provider:

{"type":"error","sequence_number":2,"error":{"type":"server_error","code":"rate_limit_exceeded","message":"The model was unable to complete inference due to an internal error.","param":null}}
@shaneqld commented on GitHub (Jan 27, 2026): > We have backoff for ratelimits, and we respect the retry-after type headers so once it is ready to retry it should automatically go through Is it possible that this behaviour isn't consistent across providers? I've run into rate limits errors fairly frequently (multiple times per day) using OpenAI models and the agent loop completely stops (ie no retries). In my case, it happens primarily using `gpt-5.2-codex`, either directly via OpenAI or via OpenCode Zen. Example error when using OpenAI provider: ```json {"type":"error","sequence_number":2,"error":{"type":"tokens","code":"rate_limit_exceeded","message":"Rate limit reached for gpt-5.2-codex in organization org-xyz on tokens per min (TPM): Limit 500000, Used 404104, Requested 107839. Please try again in 1.433s. Visit https://platform.openai.com/account/rate-limits to learn more.","param":null}} ``` Example error when using OpenCode Zen provider: ```json {"type":"error","sequence_number":2,"error":{"type":"server_error","code":"rate_limit_exceeded","message":"The model was unable to complete inference due to an internal error.","param":null}} ```
yindo commented 2026-02-16 17:29:52 -05:00
Author
Owner
Copy Link
Copy Source

@ownself commented on GitHub (Feb 1, 2026):

Is there a way to explicitly configurate RPM limit for certain model in OpenCode.json? Our API Provider lower the some Open sourced models that I want to use.
But right now it keeps hitting Request limit, sign..

@ownself commented on GitHub (Feb 1, 2026): Is there a way to explicitly configurate RPM limit for certain model in OpenCode.json? Our API Provider lower the some Open sourced models that I want to use. But right now it keeps hitting Request limit, sign..
yindo commented 2026-02-16 17:29:52 -05:00
Author
Owner
Copy Link
Copy Source

@alainivars commented on GitHub (Feb 7, 2026):

Is there a way to explicitly configure RPM limit for certain model in OpenCode.json? Our API Provider lower the some Open sourced models that I want to use. But right now it keeps hitting Request limit, sign..

Yes, I totally agree, we should be able to set by model, in the opencode.json something like
"model_name" :
...
"rateLimit": {
"maxRequests": 20,
"timeWindow": "1m"
}
}

@alainivars commented on GitHub (Feb 7, 2026): > Is there a way to explicitly configure RPM limit for certain model in OpenCode.json? Our API Provider lower the some Open sourced models that I want to use. But right now it keeps hitting Request limit, sign.. Yes, I totally agree, we should be able to set by model, in the opencode.json something like "model_name" : ... "rateLimit": { "maxRequests": 20, "timeWindow": "1m" } }
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
acp
bad
bug
discussion
docs
documentation
enhancement
extension
good first issue
growth
help-wanted
mcp
model-problem
needs:compliance
nix
opentui
perf
pull-request
Mirrored from GitHub Pull Request
 tui
web
windows
zen
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#1172
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?