Skip to content

Commit fea2e59

Browse files
androidandandroidand
committed
fix(provider): restore upstream OpenAI header timeout default
OPENAI_HEADER_TIMEOUT_DEFAULT had been changed from upstream's 300_000 to 10_000 with no comment explaining it — the adjacent fork comment is about LOCAL providers, which deliberately default to 180_000. A 10s header timeout on cloud OpenAI fails any response that is slow to send first headers, which is a live-traffic regression, not just a test one.
1 parent d8c7dc0 commit fea2e59

1 file changed

Lines changed: 1 addition & 1 deletion

File tree

‎packages/opencode/src/provider/provider.ts‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -41,7 +41,7 @@ import { ModelStatus } from "./model-status"
4141
import { RuntimeFlags } from "@/effect/runtime-flags"
4242
import { ProviderError } from "./error"
4343

44-
const OPENAI_HEADER_TIMEOUT_DEFAULT = 10_000
44+
const OPENAI_HEADER_TIMEOUT_DEFAULT = 300_000
4545
// fork: local/llama-skein providers (@ai-sdk/openai-compatible) got NO
4646
// default header timeout at all, so a backend that accepted a request and
4747
// then silently died (e.g. a model load crash) left the fetch waiting for

0 commit comments

Comments
 (0)