fix: recognize 'hit your limit' as retryable error pattern

[mauriciozaffari]

Summary

Anthropic's rate limit error "You've hit your limit · resets 5pm (UTC)" is not recognized as retryable by either the model-fallback or runtime-fallback error classifiers, preventing automatic fallback to alternative providers.

Changes

• Add "hit your limit" and "hit the limit" to RETRYABLE_MESSAGE_PATTERNS in src/shared/model-error-classifier.ts (used by model-fallback)
• Add /hit.{0,10}(?:your|the)?.{0,5}limit/i regex to RETRYABLE_ERROR_PATTERNS in src/hooks/runtime-fallback/constants.ts (used by runtime-fallback)

Context

When Anthropic returns this error, the log shows:

[background-agent] Session error - no retry: {
  "errorName":"APIError",
  "errorMessage":"Claude Code returned an error result: You've hit your limit · resets 5pm (UTC)",
  "hasFallbackChain":true,
  "canRetry":false
}

The fallback chain exists (hasFallbackChain: true) but canRetry is false because neither classifier matches the "hit your limit" phrasing. The existing patterns cover "over limit", "rate limit", and "rate_limit" but miss this specific Anthropic wording.

Testing

• bun run typecheck passes
• bun run build passes

---

Summary by cubic

Recognize Anthropic “You’ve hit your limit”, Gemini “spending cap”, and generic “usage limit” messages so cross‑provider fallback triggers reliably. Decouples error handling so quota/billing/usage STOP errors trigger fallback, while same‑provider retries stay unchanged.

• Bug Fixes
  • Add “hit your/the limit” to RETRYABLE_MESSAGE_PATTERNS; add runtime regex for hit.*limit, usage.*limit, and spending.*cap.
  • Add isStopModelError; update shouldRetryError to return true for retryable and STOP errors; classify “spending cap” and “usage limit has been reached” as STOP.
  • Expand tests for new patterns, STOP vs retryable behavior, and fallback selection.

^{Written for commit 00cb572. Summary will update on new commits.}

[github-actions]

All contributors have signed the CLA. Thank you! ✅
_{Posted by the CLA Assistant Lite bot.}

[cubic-dev-ai]

No issues found across 2 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Auto-approved: Safe addition of specific Anthropic rate limit error patterns to retryable classifiers, enabling intended fallback behavior without risk of regression.}

[mauriciozaffari]

I have read the CLA Document and I hereby sign the CLA

[cubic-dev-ai]

1 issue found across 2 files (changes from recent commits).

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="src/shared/model-error-classifier.ts">

<violation number="1" location="src/shared/model-error-classifier.ts:53">
P2: `spending cap` is a quota/billing exhaustion signal, so it should not be classified as retryable.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

0 issues found across 3 files (changes from recent commits).

_{Requires human review: Auto-approval blocked by 1 unresolved issue from previous reviews.}

[cubic-dev-ai]

0 issues found across 2 files (changes from recent commits).

_{Auto-approved: Recognizes Anthropic and Gemini-specific quota errors as retryable for cross-provider fallback. The logic correctly separates transient retries from provider-exhaustion fallbacks.}