fix(failover): classify Moonshot balance 429 as billing by leno23 · Pull Request #83079 · openclaw/openclaw

leno23 · 2026-05-17T11:21:59Z

Summary

Fixes Moonshot/Kimi exhausted-balance responses that arrive as HTTP 429 from being classified as generic rate_limit.
Keeps the override narrow by checking high-confidence billing text only for Moonshot/Kimi before the generic HTTP 429 status rule returns rate_limit.
Keeps the existing regression coverage for Moonshot/Kimi balance-shaped 429 payloads, ordinary Moonshot 429 rate limits, and non-Moonshot providers.

Test Plan

node --import tsx ./tmp-openclaw-pr-83079-green-probe.mjs
git diff --check
Attempted: OPENCLAW_VITEST_MAX_WORKERS=1 OPENCLAW_VITEST_NO_OUTPUT_TIMEOUT_MS=180000 node scripts/run-vitest.mjs run src/agents/failover-error.test.ts -t "lets Moonshot/Kimi billing-shaped 429 payloads win over generic rate limit status" (local runner stalled before test output; see Real behavior proof caveat)

Real behavior proof

Behavior addressed: a Moonshot/Kimi response with HTTP 429 and an Insufficient account balance / Please recharge payload should surface as billing, not rate_limit, so OpenClaw can show billing/recharge guidance instead of rate-limit behavior.

Real environment tested: local OpenClaw source checkout using Node.js v22.18.0 after applying commit 14b33d7160f964a2f5606d5e37a8973bef216577 on this PR branch. No live Moonshot credentials were used.

Exact steps or command run after this patch: created a temporary repo-root probe that imports src/agents/failover-error.ts and src/agents/pi-embedded-helpers/errors.ts, then ran node --import tsx ./tmp-openclaw-pr-83079-green-probe.mjs; also ran git diff --check.

Evidence after fix:

$ node --import tsx ./tmp-openclaw-pr-83079-green-probe.mjs
{"cases":[["moonshot balance 429","billing","billing"],["kimi provider hint balance 429","billing","billing"],["moonshot ordinary 429","rate_limit","rate_limit"],["openai balance-shaped 429","rate_limit","rate_limit"]],"signal":{"kind":"reason","reason":"billing"}}

$ git diff --check
(passed with no output)

Observed result after fix: Moonshot and Kimi billing-shaped 429 payloads resolve to billing; an ordinary Moonshot 429 remains rate_limit; the same balance-shaped 429 payload from openai remains rate_limit; classifyFailoverSignal() returns { kind: "reason", reason: "billing" } for the Moonshot balance-shaped 429 signal.

What was not tested: live Moonshot/Kimi depleted-account behavior was not tested because this run has no provider credentials. The focused Vitest command stalled locally before producing test output (no output for 180000ms; terminating stalled Vitest process group), so CI is expected to provide the official runner result.

Risk / Notes

Scope is intentionally narrow: only Moonshot/Kimi billing-shaped 429 responses are allowed to override the generic 429 → rate_limit rule.
The fix addresses the matcher-order review finding by checking isBillingErrorMessage(message) directly inside the provider-specific 429 branch, instead of depending on the earlier message classification.

Fixes #43447

clawsweeper · 2026-05-17T11:22:54Z

Codex review: needs maintainer review before merge.

Workflow note: Future ClawSweeper reviews update this same comment in place.

How this review workflow works

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by maintainer comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors can comment @clawsweeper re-review or @clawsweeper re-run on their own open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

Summary
The PR threads provider context into failover HTTP-status classification, classifies Moonshot/Kimi billing-shaped HTTP 429 payloads as billing before the generic 429 rate-limit rule, adds regression coverage, and updates the changelog.

Reproducibility: yes. at source level. Current main returns rate_limit for explicit HTTP 429 before the Moonshot/Kimi exhausted-balance payload can be preserved as billing, and the PR adds focused cases for that path.

Real behavior proof
Sufficient (terminal): The PR body includes after-fix terminal output from a local source-checkout probe exercising the exact classifier path and showing the expected Moonshot/Kimi billing results.

Next step before merge
No repair lane is needed because the PR already contains the focused fix; the remaining action is maintainer landing review after the pending checks complete.

Security
Cleared: The diff only changes failover classification logic, focused tests, and changelog text; I found no concrete security or supply-chain concern.

Review details

Best possible solution:

Land the narrow provider-aware 429 classifier override with the regression test once the remaining checks finish green, then let the linked source issue close from the merge.

Do we have a high-confidence way to reproduce the issue?

Yes, at source level. Current main returns rate_limit for explicit HTTP 429 before the Moonshot/Kimi exhausted-balance payload can be preserved as billing, and the PR adds focused cases for that path.

Is this the best way to solve the issue?

Yes. The PR uses a narrow provider-scoped 429 exception and avoids reordering the shared message classifier, which is the lower-risk fix for this bug.

Acceptance criteria:

Wait for the latest GitHub check runs on f061e7d to complete.
Optional focused proof: node scripts/run-vitest.mjs run src/agents/failover-error.test.ts -t "lets Moonshot/Kimi billing-shaped 429 payloads win over generic rate limit status"

What I checked:

Current main generic 429 behavior: Current main maps explicit HTTP 429 directly to rate_limit, before any later billing text can win for the Moonshot/Kimi exhausted-balance payload. (src/agents/pi-embedded-helpers/errors.ts:646, 543518bd43d7)
Current main matcher ordering: The message classifier checks rate-limit patterns before billing patterns, and failover-matches includes rate[_ ]limit as a rate-limit signal while insufficient balance is a billing signal. (src/agents/pi-embedded-helpers/errors.ts:827, 543518bd43d7)
PR fix shape: The PR passes provider into classifyFailoverClassificationFromHttpStatus() and returns billing for Moonshot/Kimi HTTP 429 only when isBillingErrorMessage(message) matches, then keeps the generic 429 to rate_limit fallback. (src/agents/pi-embedded-helpers/errors.ts:653, f061e7d08c97)
Regression coverage: The added test covers Moonshot billing 429, Kimi provider-hint billing 429, ordinary Moonshot 429 rate limits, non-Moonshot balance-shaped 429 payloads, and classifyFailoverSignal(). (src/agents/failover-error.test.ts:299, f061e7d08c97)
Review discussion incorporated: The PR discussion identified that messageReason could not become billing because rate_limit_reached matched first; the current diff uses the recommended direct isBillingErrorMessage(message) check in the provider-specific 429 branch. (src/agents/pi-embedded-helpers/errors.ts:653, f061e7d08c97)
Latest head and changed files: The live PR head is f061e7d, changing only CHANGELOG.md, src/agents/failover-error.test.ts, and src/agents/pi-embedded-helpers/errors.ts. (f061e7d08c97)

Likely related people:

steipete: Local blame and GitHub file history show recent work on the central failover classifier and related failover handling paths. (role: recent area contributor; confidence: medium; commits: 4ccd07718d2f, 936c02e22c98; files: src/agents/pi-embedded-helpers/errors.ts, src/agents/failover-error.test.ts, src/agents/pi-embedded-helpers/failover-matches.ts)
vincentkoc: Recent main history for errors.ts includes auth/failover classification work, adjacent to the status and message-classification surface this PR changes. (role: recent adjacent contributor; confidence: medium; commits: 3485a907d13f; files: src/agents/pi-embedded-helpers/errors.ts)
hclsys: Recent main history for failover-matches.ts includes failover matcher changes and verification on related agent failover behavior. (role: recent adjacent contributor; confidence: medium; commits: 398dd6e0b091; files: src/agents/pi-embedded-helpers/failover-matches.ts)
altaywtf: The live PR is assigned to altaywtf, and the latest head commit only adds the changelog entry after the focused classifier fix. (role: current PR handler; confidence: medium; commits: f061e7d08c97; files: CHANGELOG.md)

Remaining risk / open question:

The latest head still had queued or in-progress check runs during review, including network-runtime-boundary and Socket Security checks.
The proof uses the issue-backed payload against the classifier path; no live depleted Moonshot/Kimi account was exercised.

Codex review notes: model gpt-5.5, reasoning high; reviewed against 543518bd43d7.

MantisCartography · 2026-05-17T12:07:01Z

Verified ClawSweeper's P1 finding locally — the fix as written cannot work because of matcher ordering.

What I verified:

I inspected the branch source at 701a3a3d3f53 and traced the full classification path:

classifyFailoverClassificationFromMessage() in errors.ts checks isRateLimitErrorMessage() before isBillingErrorMessage() (lines ~848 vs ~855+).
The Moonshot fixture MOONSHOT_INSUFFICIENT_BALANCE_429_PAYLOAD contains "rate_limit_reached", which matches the rate-limit pattern /rate[_ ]limit/ in failover-matches.ts.
So messageClassification resolves to "rate_limit", not "billing".
The new 429 guard at errors.ts:651 checks messageReason === "billing", which is never true for this payload.
The generic return toReasonClassification("rate_limit") at line ~657 fires instead.

Result: The test lets Moonshot/Kimi billing-shaped 429 payloads win over generic rate limit status should fail because resolveFailoverReasonFromError returns "rate_limit" for the Moonshot fixture.

Suggested fix approaches:

Option A: In the 429 handler, check isBillingErrorMessage(message) for known billing-sensitive providers (Moonshot/Kimi) directly, without relying on messageReason. E.g.:

if (status === 429) {
  if (
    (isProvider(provider, "moonshot") || isProvider(provider, "kimi")) &&
    isBillingErrorMessage(message)
  ) {
    return toReasonClassification("billing");
  }
  return toReasonClassification("rate_limit");
}

Option B: Reorder classifyFailoverClassificationFromMessage() to check billing patterns before rate-limit patterns when a provider hint indicates a billing-sensitive provider, though this is broader and more risky.

Option A is narrower and avoids changing the shared message classifier ordering. It also means the Moonshot payload's "Insufficient account balance. Please recharge" text will match the existing insufficient_balance billing pattern in failover-matches.ts, correctly classifying as billing even though "rate_limit_reached" also appears in the payload.

Has anyone run the test suite on this branch? I expect src/agents/failover-error.test.ts to fail on the Moonshot/Kimi 429 billing test case.

leno23 · 2026-05-17T13:31:42Z

Addressed the matcher-order finding in 14b33d7160f964a2f5606d5e37a8973bef216577: the 429 branch now checks Moonshot/Kimi billing text directly with isBillingErrorMessage(message) before falling back to generic rate_limit, so the rate_limit_reached token no longer prevents exhausted-balance payloads from reaching billing.

Verification added to the PR body: source-level RED/GREEN probe for Moonshot, Kimi provider hint, ordinary Moonshot 429, non-Moonshot 429, plus git diff --check. The local Vitest runner stalled before output, so I left the PR as draft and documented that CI still needs to provide the official runner result.

Copilot

Pull request overview

Narrow fix to ensure Moonshot/Kimi providers that return HTTP 429 with an "Insufficient account balance" payload are classified as billing rather than the generic rate_limit, so the UI can show recharge guidance and failover behavior is correct (fixes #43447).

Changes:

Thread provider into classifyFailoverClassificationFromHttpStatus and, in the 429 branch, return billing for Moonshot/Kimi when the message is a billing-shaped error.
Add a regression test covering Moonshot/Kimi billing 429s, ordinary Moonshot 429s, and non-Moonshot 429 payloads.
Add a CHANGELOG entry.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
src/agents/pi-embedded-helpers/errors.ts	Adds provider-aware billing override before the 429 → rate_limit fallback.
src/agents/failover-error.test.ts	Adds coverage for the new Moonshot/Kimi billing-429 behavior.
CHANGELOG.md	Documents the fix.

Preserve billing classification for Moonshot/Kimi HTTP 429 payloads that report insufficient account balance, while leaving ordinary 429 responses as rate_limit.

altaywtf · 2026-05-17T15:06:14Z

Merged via squash.

Prepared head SHA: 9f70bf593544cc709a45aebcf3d1e0fe0be91f67
Merge commit: 019dbcc7490e0dbfce4f9554d2113c8c4faf48b9

Thanks @leno23!

openclaw-barnacle Bot added agents Agent runtime and tooling size: S proof: supplied External PR includes structured after-fix real behavior proof. labels May 17, 2026

clawsweeper Bot added P1 High-priority user-facing bug, regression, or broken workflow. impact:session-state Session, memory, transcript, context, or agent state can drift or corrupt. impact:auth-provider Auth, provider routing, model choice, or SecretRef resolution may break. labels May 17, 2026

clawsweeper Bot mentioned this pull request May 17, 2026

Kimi/Moonshot 'Rate Limit' error masks insufficient funds, causes UI lockout #43447

Closed

clawsweeper Bot added the proof: sufficient ClawSweeper judged the real behavior proof convincing. label May 17, 2026

altaywtf self-assigned this May 17, 2026

clawsweeper Bot added the proof: sufficient ClawSweeper judged the real behavior proof convincing. label May 17, 2026

altaywtf marked this pull request as ready for review May 17, 2026 14:37

Copilot AI review requested due to automatic review settings May 17, 2026 14:37

Copilot started reviewing on behalf of altaywtf May 17, 2026 14:37 View session

Copilot AI reviewed May 17, 2026

View reviewed changes

clawsweeper Bot removed the proof: sufficient ClawSweeper judged the real behavior proof convincing. label May 17, 2026

altaywtf force-pushed the hermes-auto/issue-43447-moonshot-429-billing branch from 8377b4b to dd05301 Compare May 17, 2026 14:51

clawsweeper Bot added the proof: sufficient ClawSweeper judged the real behavior proof convincing. label May 17, 2026

altaywtf force-pushed the hermes-auto/issue-43447-moonshot-429-billing branch from dd05301 to f061e7d Compare May 17, 2026 14:58

openclaw-barnacle Bot removed the proof: sufficient ClawSweeper judged the real behavior proof convincing. label May 17, 2026

clawsweeper Bot added the proof: sufficient ClawSweeper judged the real behavior proof convincing. label May 17, 2026

altaywtf force-pushed the hermes-auto/issue-43447-moonshot-429-billing branch from f061e7d to 5856a6a Compare May 17, 2026 15:05

openclaw-barnacle Bot removed the proof: sufficient ClawSweeper judged the real behavior proof convincing. label May 17, 2026

leno23 and others added 3 commits May 17, 2026 18:05

fix(failover): classify moonshot exhausted balance as billing

a6c9cd5

Preserve billing classification for Moonshot/Kimi HTTP 429 payloads that report insufficient account balance, while leaving ordinary 429 responses as rate_limit.

fix(failover): preserve Moonshot balance 429 billing

c0d38c6

fix(failover): document Moonshot billing 429 handling

9f70bf5

altaywtf force-pushed the hermes-auto/issue-43447-moonshot-429-billing branch from 5856a6a to 9f70bf5 Compare May 17, 2026 15:05

altaywtf merged commit 019dbcc into openclaw:main May 17, 2026
14 checks passed

github-actions Bot mentioned this pull request May 17, 2026

📡 Upstream Digest — 2026-05-17 16:48 UTC curtismercier/openclaw-mods#884

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(failover): classify Moonshot balance 429 as billing#83079

fix(failover): classify Moonshot balance 429 as billing#83079
altaywtf merged 3 commits into
openclaw:mainfrom
leno23:hermes-auto/issue-43447-moonshot-429-billing

leno23 commented May 17, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented May 17, 2026 •

edited

Loading

Uh oh!

MantisCartography commented May 17, 2026

Uh oh!

leno23 commented May 17, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

altaywtf commented May 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Conversation

leno23 commented May 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test Plan

Real behavior proof

Risk / Notes

Uh oh!

clawsweeper Bot commented May 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

MantisCartography commented May 17, 2026

Uh oh!

leno23 commented May 17, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

altaywtf commented May 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

leno23 commented May 17, 2026 •

edited

Loading

clawsweeper Bot commented May 17, 2026 •

edited

Loading