fix(provider): accept OpenAI ultrafast service tier by rekram1-node · Pull Request #53875 · anomalyco/opencode

rekram1-node · 2026-10-08T05:42:59Z

Summary

Extend the existing @ai-sdk/[email protected] patch to accept serviceTier: "ultrafast" for Chat Completions and Responses, including matching type declarations.
Retain the existing explicit-tier passthrough patch. SDK versions, manifests, and lockfile are unchanged.
Add one offline regression test for GPT-6 Astra Responses streaming, checking that service_tier: "ultrafast" reaches the outgoing request.

Reproduced the validation failure before patching.
Provider suite passed before trimming the test matrix; the remaining focused regression test passes.
bun typecheck from packages/opencode
Prettier, oxlint, and git diff --check
Verified the patch applies to the original published 3.0.88 package.

No live-provider calls were made.

Nix hashes, Sonnet 5.5 in the Zen docs tables, and the OpenAI ultrafast service tier fix (anomalyco#53875), which grows the @ai-sdk/[email protected] patch without changing the version it is pinned to. No file this fork has touched is in the merge. Co-Authored-By: Claude Opus 5 <[email protected]> Claude-Session: https://claude.ai/code/session_017Gr2KVmmJbKQPntkpbsnP9

rekram1-node added 2 commits October 8, 2026 00:42

fix(provider): accept OpenAI ultrafast service tier

8672f3d

test(provider): focus ultrafast regression coverage

a5dc521

rekram1-node merged commit 5e2df97 into dev Oct 8, 2026
10 checks passed

rekram1-node deleted the openai-ultrafast branch October 8, 2026 05:52

pminevp mentioned this pull request Oct 12, 2026

Upstream sync: anomalyco/opencode dev into dev (2026-10-12) (conflicts) AxsionDev/Lunos#190

Closed