You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
fix(server): OpenCode ChatGPT usage limits keep their reset time - #26264
When OpenCode runs an OpenAI model through a ChatGPT subscription and the Codex usage window is exhausted, the turn now fails as usage_limit (since #18315) but without resetAt. The thread shows "Reset time unavailable; retry manually", the resume-at-reset and snooze actions are hidden, and auto-resume never arms.
ChatGPT already says when the limit lifts. Its 429 body is:
{"error":{"type":"usage_limit_reached","message":"The usage limit has been reached","plan_type":"plus","resets_at":1788753614,"resets_in_seconds":57361}}
OpenCode 2 forwards that body on session.execution.failed as error.response.body, but the adapter ignored it.
Change
When the OpenCode 2 adapter classifies a failure as a usage limit, it decodes error.resets_at (Unix seconds) from response.body and passes it as resetAt to makeProviderFailure, which already validates it and keeps it only for usage_limit. Claude limits keep the message-based reset from #29301, which takes precedence. Other bodies without that field (OpenCode Go, Zen, other providers) still produce no reset time, as before. No contract, client, or orchestration changes; the existing reset UI and auto-resume worker pick it up.
Follow-up to #16366, which was closed by #18315 with the note that "OpenCode doesn't send a reset time in these events, so none is shown yet". For ChatGPT-backed models it does send one, in the response body. This is a small, focused fix in one adapter branch: it fills a field the failure contract already has, for the one provider shape that reports it. Separate from the Codex-provider fix in #18721 / #14169.
Verification
Added a ChatGPT 429 row to the existing classifies $name as $failureClass table in OpenCode2AdapterV2.test.ts, expecting resetAt: "2026-10-27T15:25:41.000Z". With the adapter change stashed it fails on exactly that row (missing resetAt); the other rows pass. With the change, the file passes (89/89 after merging current main).
vp run check:changed: format, lint, knip, typecheck (36 packages), and the changed test file all pass.
Payload shape: taken from a real ChatGPT Plus usage-limit 429 recorded in my local OpenCode store (responseBody on an OpenCode 1.x message). I confirmed that OpenCode 2.1.3's execution-failure mapper includes response: { body } when the HTTP error has a body, and the @opencode/schema 2.0.28 Session.StructuredError used here declares response.body.
Not checked: a live end-to-end ChatGPT quota refusal through T3. My quota had already reset when I wrote this, and I couldn't get OpenCode 2 to route to a local mock 429 endpoint. Server-only change, so there are no screenshots.
Implementation, tests, and PR: Claude Opus 5.5 via Claude Code in T3 Code.
ChatGPT-backed OpenAI models answer usage limits with a 429 whose body
carries error.resets_at. Read it from the failure's response body so the
usage-limit failure gets a resetAt, enabling resume and snooze at reset.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Macroscope's review found this PR approvable — Preserves provider-reported usage-limit reset times for existing display and recovery flows. Invalid timestamps are ignored, and automatic recovery remains controlled by existing preferences, disabled by default.
You can add or adjust custom eligibility rules. Learn more.
No actionable comments were generated in the recent review.
Walkthrough
The OpenCode 2 adapter parses a valid resets_at timestamp from an error response and uses it when no Claude-specific reset time is available. A test covers a ChatGPT quota error with HTTP 429 and verifies the expected reset time.
Changes: OpenCode 2 reset-time handling
Layer / File(s)
Summary
Parse and attach reset time packages/provider-opencode/src/server/v2/adapter.ts, apps/server/src/orchestration-v2/Adapters/OpenCode2AdapterV2.test.ts
The adapter converts a finite Unix-seconds resets_at value to an ISO timestamp. It returns null if decoding or date conversion fails, and uses the parsed timestamp when a Claude-specific reset time is unavailable. The test checks the ChatGPT quota-error case and conditionally expects resetAt.
pingdotgg/t3code#28375: Both changes affect usage-limit handling in the OpenCode 2 adapter. This PR preserves the reset time for a ChatGPT quota failure.
pingdotgg/t3code#29301: Both changes affect OpenCode 2 failure handling and resetAt. This PR adds ChatGPT reset extraction while retaining the Claude reset time as a fallback.
Suggested reviewers:juliusmarminge, fixfon
Merge Risk: Minimal · up to 51345
The change is mergeable; it preserves existing failure behavior while attaching the quota reset time for supported usage-limit responses.
Pre-merge checks: 4
Check name
Status
Explanation
Title check
The title clearly and concisely describes the main change: preserving the reset time for ChatGPT usage limits in the OpenCode adapter.
Description check
The description includes complete Problem, Change, Scope and approval, and Verification sections. It explains the affected behavior, implementation, scope rationale, issue reference, tests, validation…
Linked Issues check
Issue #26265 requires OpenCode 2 ChatGPT usage-limit failures to carry the reset time from the 429 body. The whole-PR diff (merge base to head) changes `packages/provider-opencode/src/server/v2/adapte…
Out of Scope Changes check
The whole-PR diff touches two files: packages/provider-opencode/src/server/v2/adapter.ts and apps/server/src/orchestration-v2/Adapters/OpenCode2AdapterV2.test.ts. The adapter change adds the `rese…
Full details: Description check
Explanation
The description includes complete Problem, Change, Scope and approval, and Verification sections. It explains the affected behavior, implementation, scope rationale, issue reference, tests, validation results, and untested live scenario.
Full details: Linked Issues check
Explanation
Issue #26265 requires OpenCode 2 ChatGPT usage-limit failures to carry the reset time from the 429 body. The whole-PR diff (merge base to head) changes packages/provider-opencode/src/server/v2/adapter.ts. A new usageLimitResetAt helper decodes error.resets_at (Unix seconds) from response.body with Schema.decodeUnknownOption. It converts the value to ISO with DateTime.make and DateTime.formatIso, and returns null when the body or field is absent or invalid. The failure record now sets resetAt: claudeLimit?.resetAt ?? usageLimitResetAt(response?.body), so the Claude message-based reset keeps precedence, as the PR describes. The test diff adds a ChatGPT provider.quota 429 row with usage_limit_reached body to the existing classifies $name as $failureClass table, expecting resetAt: "2026-10-27T15:25:41.000Z". I checked the fixture value: 1793114741 Unix seconds equals 2026-10-27T15:25:41Z. The test harness asserts resetAt only when the row defines it. The UI and auto-resume behavior named in the issue rely on existing code, which the PR states is unchanged. The adapter now supplies the field those paths read. The reviewed diff does not show the UI or worker code, so the end-to-end display and auto-resume behavior is not re-verified here. The coding requirement at the adapter level is met.
Full details: Out of Scope Changes check
Explanation
The whole-PR diff touches two files: packages/provider-opencode/src/server/v2/adapter.ts and apps/server/src/orchestration-v2/Adapters/OpenCode2AdapterV2.test.ts. The adapter change adds the resets_at extraction helper, the effect/Option import it uses, and the resetAt fallback in the usage-limit failure branch. The test change adds one ChatGPT row, and adjusts the table runner to pass resetAt only when a row defines it. All changes serve issue #26265. No contract, client, orchestration, or unrelated provider changes appear in the reviewed diff. The Option import and helper are needed by the new decode path.
Finishing Touches
Generate unit tests (beta)
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Warning
Some tools did not complete. Review the errors below.
ast-grep (0.51.1)
ast-grep could not parse rule config: /ast-grep-rules/ast-grep-essentials/rules/csharp/security/jwt-hardcoded-secret-csharp.yml
Tip
🎯 Customize your risk assessments
Guide how CodeRabbit assesses merge risk and security architecture risk. Highlight sensitive changes and explain your security boundaries using Merge Risk Instructions and Security Risk Instructions in your review settings. Security architecture risk assessments require the Advanced plan or above.
✅ Automatic approvals, with your rules
Let CodeRabbit approve pull requests once required threads are resolved, pre-merge checks pass, and your risk limits are met. Enable Automatic approvals in review settings, with overrides for teams or file paths.
CodeRabbit Verify
Run CodeRabbit Verify to test the changes in this pull request.
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 9 remain after this review.
Claude Code (Claude Opus 5.5) is posting this comment for the PR author.
I merged current main into the branch (29c81724). The diff against main is still the same two files, +31/−2.
The earlier Agent context budget failure (claudeSystemAppendBytes: 5135 -> 5258) came from the PR being 254 commits behind. main had shrunk agentContextBudget.json since this branch was created, and this PR doesn't touch agent context. On the new head, Agent Context Budget, CI, Mobile Fingerprint, PR Vouch and PR Size all pass, and the branch merges cleanly.
The pr:gate-failed label seems to be left over from that run. Could a maintainer clear it? The bug report is #26265.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
bugfixFix PR opened by an issue lead; enters the PR looppr:gate-failedA land gate failedsize:S10-29 changed lines (additions + deletions).vouch:unvouchedPR author is not yet trusted in the VOUCHED list.
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
When OpenCode runs an OpenAI model through a ChatGPT subscription and the Codex usage window is exhausted, the turn now fails as
usage_limit(since #18315) but withoutresetAt. The thread shows "Reset time unavailable; retry manually", the resume-at-reset and snooze actions are hidden, and auto-resume never arms.ChatGPT already says when the limit lifts. Its 429 body is:
{"error":{"type":"usage_limit_reached","message":"The usage limit has been reached","plan_type":"plus","resets_at":1788753614,"resets_in_seconds":57361}}OpenCode 2 forwards that body on
session.execution.failedaserror.response.body, but the adapter ignored it.Change
When the OpenCode 2 adapter classifies a failure as a usage limit, it decodes
error.resets_at(Unix seconds) fromresponse.bodyand passes it asresetAttomakeProviderFailure, which already validates it and keeps it only forusage_limit. Claude limits keep the message-based reset from #29301, which takes precedence. Other bodies without that field (OpenCode Go, Zen, other providers) still produce no reset time, as before. No contract, client, or orchestration changes; the existing reset UI and auto-resume worker pick it up.Scope and approval
Fixes #26265.
Follow-up to #16366, which was closed by #18315 with the note that "OpenCode doesn't send a reset time in these events, so none is shown yet". For ChatGPT-backed models it does send one, in the response body. This is a small, focused fix in one adapter branch: it fills a field the failure contract already has, for the one provider shape that reports it. Separate from the Codex-provider fix in #18721 / #14169.
Verification
classifies $name as $failureClasstable inOpenCode2AdapterV2.test.ts, expectingresetAt: "2026-10-27T15:25:41.000Z". With the adapter change stashed it fails on exactly that row (missingresetAt); the other rows pass. With the change, the file passes (89/89 after merging currentmain).vp run check:changed: format, lint, knip, typecheck (36 packages), and the changed test file all pass.responseBodyon an OpenCode 1.x message). I confirmed that OpenCode 2.1.3's execution-failure mapper includesresponse: { body }when the HTTP error has a body, and the@opencode/schema2.0.28Session.StructuredErrorused here declaresresponse.body.Implementation, tests, and PR: Claude Opus 5.5 via Claude Code in T3 Code.
🤖 Generated with Claude Code