fix(model-core): restore OpenAI server_error retryable patterns (regression from package extraction)

c85d2f9bc added 'server_error' and 'an error occurred while processing'
to the retryable message patterns to fix #3799 - the OpenAI streaming
server_error case where runtime fallback never fired. The package
layering refactor (2748009ff) moved model-error-classifier into
packages/model-core but dropped these two patterns during the move.

Restore both patterns at the matching position. packages/model-core
tests now pass on the server_error retryable assertion, and runtime
fallback once again retries the OpenAI streaming server_error envelope.
This commit is contained in:
YeonGyu-Kim
2026-05-22 00:06:03 +09:00
parent 9624914c25
commit 9c9e9885fe
@@ -81,6 +81,8 @@ const RETRYABLE_MESSAGE_PATTERNS = [
"请求过于频繁", // "too many requests"
"暂时不可用", // "temporarily unavailable"
"服务不可用", // "service unavailable"
"server_error",
"an error occurred while processing",
]
/**