Map a low-level fetch failure (server unreachable, connection reset,
the request timeout firing as a non-Abort transport error) to a
transient error. Talking to a local OpenAI-compatible server that is
down or mid-generation is retryable — a retry after backoff can
succeed once the server is up / a slow generation completes.
Map a low-level
fetchfailure (server unreachable, connection reset, the request timeout firing as a non-Abort transport error) to a transient error. Talking to a local OpenAI-compatible server that is down or mid-generation is retryable — a retry after backoff can succeed once the server is up / a slow generation completes.