feat: support multiple rate limits by xl-openai · Pull Request #11260 · openai/codex

xl-openai · 2026-02-10T01:58:00Z

Added multi-limit support end-to-end by carrying limit_name in rate-limit snapshots and handling multiple buckets instead of only codex.
Extended /usage client parsing to consume additional_rate_limits
Updated TUI /status and in-memory state to store/render per-limit snapshots
Extended app-server rate-limit read response: kept rate_limits and added rate_limits_by_name.
Adjusted usage-limit error messaging for non-default codex limit buckets

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5a1de97fb3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-02-10T07:08:30Z

codex-rs/app-server/src/codex_message_processor.rs

Fall back when codex snapshot has no limit data

This selection always prefers the "codex" entry, but get_rate_limits_many now always includes a synthetic codex snapshot even when /usage returns rate_limit: null and only non-codex buckets have real window data. In that case rate_limits in GetAccountRateLimitsResponse is returned empty while rate_limits_by_name has populated limits, which regresses legacy clients that still read only rate_limits. Prefer codex only when it actually contains window/credit data, otherwise fall back to a populated snapshot.

Useful? React with 👍 / 👎.

rate_limits is intentionally the canonical codex bucket for legacy clients; falling back to additional buckets would change its semantics and misrepresent what legacy clients are actually able to interpret.

chatgpt-codex-connector · 2026-02-10T07:08:30Z

codex-rs/codex-api/src/rate_limits.rs

Use matching limit name in secondary-header parser test

This test sets x-codex-secondary-* headers but calls parse_rate_limit_for_limit(..., Some("codex_other")), so the parser looks for x-codex-other-* keys and cannot populate primary, making the subsequent expect("primary") fail. The limit name and header prefix need to match (for example codex_secondary with x-codex-secondary-*).

Useful? React with 👍 / 👎.

xl-openai · 2026-02-11T02:03:35Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: dd069802d6

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-02-11T02:10:33Z

codex-rs/tui/src/chatwidget.rs

+            self.rate_limit_snapshots_by_limit_id
+                .insert(limit_id, display);


Prune stale limit buckets when applying updates

This path only upserts a single bucket into rate_limit_snapshots_by_limit_id, and buckets are cleared only when snapshot is None. With the new multi-limit polling flow, snapshots are emitted one-by-one from /usage, so if a bucket disappears from a later response (for example, codex_other is no longer returned), there is no tombstone and the old entry remains in memory, causing /status to keep rendering stale limits indefinitely.

Useful? React with 👍 / 👎.

I think it is OK for now.

chatgpt-codex-connector · 2026-02-11T02:10:33Z

codex-rs/codex-api/src/rate_limits.rs

+    let suffix = "-primary-used-percent";
+    let prefix = header_name.strip_suffix(suffix)?;


Detect non-default limit ids beyond primary headers

parse_rate_limit_updates discovers additional buckets via header_name_to_limit_id, but this helper only matches headers ending in -primary-used-percent. If a non-default bucket is reported with only secondary-window headers, that bucket is never discovered, so no snapshot is emitted and consumers miss the update entirely. Limit-id discovery should include secondary header families (or use a broader prefix-based detection) to avoid dropping valid buckets.

Useful? React with 👍 / 👎.

This should no happen.

ccy-oai · 2026-02-11T02:27:57Z

codex-rs/app-server/src/bespoke_event_handling.rs

are the core rate limits name/idless?

ok i see you store them rate_limits

I can update the test with a real value of limit_id/name.

ccy-oai · 2026-02-11T02:33:10Z

codex-rs/app-server-protocol/schema/json/v1/ForkConversationResponse.json

do we still update v1 schemas?

Must be auto-generated from some command I run. Let me revert.

just write-app-server-schema actually it will still generate v1 (otherwise C1 will fail).

ccy-oai · 2026-02-11T02:47:56Z

codex-rs/core/src/error.rs

does it have to gate to pro or can it be more general?

xl-openai force-pushed the xl/skill-limit branch 3 times, most recently from f0abfb0 to 5a1de97 Compare February 10, 2026 06:57

xl-openai marked this pull request as ready for review February 10, 2026 06:57

chatgpt-codex-connector bot reviewed Feb 10, 2026

View reviewed changes

xl-openai force-pushed the xl/skill-limit branch 6 times, most recently from 9f74d29 to dd06980 Compare February 11, 2026 01:52

chatgpt-codex-connector bot reviewed Feb 11, 2026

View reviewed changes

feat: support multiple rate limits

5d24c10

ccy-oai reviewed Feb 11, 2026

View reviewed changes

xl-openai force-pushed the xl/skill-limit branch from dd06980 to 7c81f7e Compare February 11, 2026 02:45

ccy-oai reviewed Feb 11, 2026

View reviewed changes

ccy-oai approved these changes Feb 11, 2026

View reviewed changes

aibrahim-oai approved these changes Feb 11, 2026

View reviewed changes

Address comments.

b494b5b

xl-openai force-pushed the xl/skill-limit branch from 7c81f7e to b494b5b Compare February 11, 2026 02:59

xl-openai merged commit fdd0cd1 into main Feb 11, 2026
36 of 38 checks passed

xl-openai deleted the xl/skill-limit branch February 11, 2026 04:09

github-actions bot locked and limited conversation to collaborators Feb 11, 2026

		self.rate_limit_snapshots_by_limit_id
		.insert(limit_id, display);

		let suffix = "-primary-used-percent";
		let prefix = header_name.strip_suffix(suffix)?;

Conversation

xl-openai commented Feb 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector bot Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

xl-openai commented Feb 11, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector bot Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Comments

xl-openai commented Feb 10, 2026 •

edited

Loading