发布

  • fix(sentry): suppress Anthropic quota and API timeout errors from Sentry (#1939)

    frostbyte_neo 发布于 2026-05-14 01:03:22 +00:00

    • fix(sentry): filter Anthropic quota and API timeout errors from Sentry

    Two classes of LLM provider errors were reaching Sentry as bugs even
    though they represent operator-actionable, non-code conditions:

    1. Anthropic HTTP 400 "usage limits" (issues #1883 and #1885):
      AnthropicBadRequestError messages containing "usage limits" now raise
      "Anthropic billing quota exceeded. Check your plan and billing details."
      — consistent with the existing OpenAI insufficient_quota handling.

    2. OpenAI/Openai-compatible API timeout (issue #1934):
      The "{provider} API request timed out." message wasn't matched by any
      _OPERATOR_ACTIONABLE_LLM_ERROR_PATTERNS entry, so it reached Sentry.

    Both cases are now suppressed via two new patterns added to
    _OPERATOR_ACTIONABLE_LLM_ERROR_PATTERNS in app/utils/sentry_sdk.py:

    • r"\bbilling quota exceeded\b" — catches quota exhaustion for all providers
    • r"\bapi request timed out\b" — catches transient endpoint timeouts

    Closes #1883
    Closes #1885
    Closes #1934

    https://claude.ai/code/session_0144hMeudYUNxuULDFvfhWn5

    • address greptile feedback

    Co-authored-by: Claude noreply@anthropic.com
    Co-authored-by: davincios davincios@users.noreply.github.com
    Co-authored-by: Cursor cursoragent@cursor.com

    下载附件