fix(openai-chat): let opted-out providers pin parallel_tool_calls:false by TooSpace · Pull Request #1662 · lidge-jun/opencodex

TooSpace · 2026-08-14T03:41:22Z

What

Add an opt-in provider flag pinParallelToolCallsFalse so an opted-out
openai-chat provider can actually send parallel_tool_calls: false on the
wire, instead of silently omitting it.

Fixes #1650.

Why

In src/adapters/openai-chat.ts, when provider.parallelToolCalls === false the
wire bit parallel_tool_calls: false is only emitted for the NVIDIA NIM
baseUrl. Every other opted-out provider omits the field entirely. For strict
OpenAI-compatible hosts that reject unknown fields this omit-by-default is
correct. But self-hosted gateways in the Kimi/GLM family (and vLLM-style
backends) do honor parallel_tool_calls and, without it, keep emitting
multiple concurrent tool calls in a single streamed turn -- which the Responses
translator then has to reject. parallelToolCalls: false looks like the opt-out
for this, but today it is a no-op for those providers.

Design / why opt-in

The omit-by-default behavior is deliberate and covered by existing tests
(tests/parallel-tool-calls-optin.test.ts: a false provider must NOT emit the
field). Sending parallel_tool_calls: false unconditionally would break those
strict hosts, so the wire bit stays gated:

NVIDIA NIM baseUrl: unchanged (still pinned).
Any other provider: pinned only when it sets pinParallelToolCallsFalse: true.
No effect unless parallelToolCalls === false.

Default behavior for every existing provider is unchanged.

Changes

src/types.ts: add documented optional pinParallelToolCallsFalse?: boolean to OcxProviderConfig.
src/adapters/openai-chat.ts: OR the new flag into the existing NVIDIA-baseUrl condition that pins the wire bit.
tests/parallel-tool-calls-optin.test.ts: cover pinned-false emits the bit, keeps emitting under a permissive request bit, and is a no-op without parallelToolCalls:false.

Testing

bun x tsc --noEmit clean on top of dev.
bun test for parallel-tool-calls-optin, openai-chat-hardening, openai-chat-parallel-stream, chat-completions-endpoint: 126 pass, 0 fail (includes 3 new cases). Existing omit-by-default assertions still pass.

Review readiness checklist

This PR stays in draft until every box below is ticked. Tick all four boxes once the requirements are met:

All CI tests are green on my local testing.
I pushed my PR to the latest dev commit.
I resolved all correct Codex and CodeRabbit findings.
My PR is ready for review.

Summary by CodeRabbit

New Features
- Added provider configuration to explicitly send parallel_tool_calls: false when parallel tool calls are disabled.
- Supports overriding permissive request-level settings when this option is enabled.
Bug Fixes
- Improved consistency of parallel tool-call settings across supported providers.
- Providers without the opt-in configuration remain unchanged.

When a provider sets parallelToolCalls:false, the wire bit parallel_tool_calls:false is only sent for the NVIDIA NIM baseUrl; every other opted-out openai-chat provider silently OMITS the field. That is the right default for strict OpenAI-compatible hosts that reject unknown knobs, but self-hosted gateways (Kimi/GLM-family, vLLM, etc.) that DO honor the field keep emitting concurrent tool calls in one streamed turn because they never receive it. Add an opt-in provider flag pinParallelToolCallsFalse that pins the wire bit for such gateways without hardcoding their URL. Default behavior is unchanged: the field is still omitted for every existing opted-out provider (the NVIDIA hardcode and the omit-by-default path are both preserved), and the flag has no effect unless parallelToolCalls === false. Fixes lidge-jun#1650

github-actions · 2026-08-14T03:41:33Z

✅ Deterministic PR hygiene checks passed.

coderabbitai · 2026-08-14T03:41:42Z

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 55b03e43-5e5a-4a04-8923-08df8a6e15e3

📥 Commits

Reviewing files that changed from the base of the PR and between a1e5192 and d31627f.

📒 Files selected for processing (3)

src/adapters/openai-chat.ts
src/types.ts
tests/parallel-tool-calls-optin.test.ts

📝 Walkthrough

Walkthrough

The change adds pinParallelToolCallsFalse to provider configuration. For opted-out openai-chat providers, request construction now sends parallel_tool_calls: false. Tests cover default emission, request-level overrides, and omitted provider configuration.

Changes

Parallel tool-call pinning

Layer / File(s)	Summary
Provider option, request behavior, and validation `src/types.ts`, `src/adapters/openai-chat.ts`, `tests/parallel-tool-calls-optin.test.ts`	`OcxProviderConfig` now supports `pinParallelToolCallsFalse`. When `parallelToolCalls` is `false`, `openai-chat` sends `parallel_tool_calls: false` for NVIDIA NIM or providers with this option enabled. Tests verify forced emission, precedence over request-level `true`, and omission without the provider opt-out.

Estimated code review effort: 2 (Simple) | ~10 minutes

Merge Risk: ⚪ Minimal · up to d3162

This localized change enables opted-in providers to send parallel_tool_calls:false while preserving existing defaults; no actionable merge-blocking risk remains after normal checks and review.

Possibly related PRs

lidge-jun/opencodex#1317: Both changes modify openai-chat.ts handling of the parallel_tool_calls request field.

Suggested labels: enhancement

Suggested reviewers: lidge-jun, olddonkey, ingwannu

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	The changes implement issue `#1650` through an explicit provider opt-in, preserve NVIDIA behavior, and retain omission for providers without the flag.
Out of Scope Changes check	✅ Passed	The provider type, request construction, and tests are directly related to issue `#1650` and the stated pull request objectives.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: allowing opted-out providers to pin parallel_tool_calls:false.

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

github-actions · 2026-08-14T03:41:43Z

✅ READY

all PR quality gates passed; the review readiness checklist is complete.

Review readiness checklist

✅ All CI tests are green on my local testing.
✅ I pushed my PR to the latest dev commit.
✅ I resolved all correct Codex and CodeRabbit findings.
✅ My PR is ready for review.

✅ 4/4 boxes ticked.

This pull request has been marked Ready for Review.
The review-ready label marks this PR as ready; review automation runs independently.
Maintainers notified: @lidge-jun @Ingwannu @Wibias

Wibias · 2026-08-14T06:57:47Z

Thanks for this. The provider-level opt-in is useful because it fixes a real compatibility gap without changing the safe default for every OpenAI-compatible backend. Gateways that need parallel_tool_calls: false can now receive it explicitly, while stricter providers that may reject the field remain untouched. That gives us the compatibility knob without introducing a broad regression risk. 🙏

…ls-pin-false fix(openai-chat): let opted-out providers pin parallel_tool_calls:false

TooSpace requested review from Ingwannu, Wibias and lidge-jun as code owners August 14, 2026 03:41

github-actions Bot added the bug Something isn't working label Aug 14, 2026

github-actions Bot marked this pull request as draft August 14, 2026 03:41

github-actions Bot added the review-ready label Aug 14, 2026

github-actions Bot marked this pull request as ready for review August 14, 2026 04:11

Wibias merged commit e480ce7 into lidge-jun:dev Aug 14, 2026
30 checks passed

Ingwannu mentioned this pull request Aug 14, 2026

[Bug]: parallelToolCalls=false only sends parallel_tool_calls:false for the NVIDIA NIM baseUrl; ignored for all other opted-out providers #1650

Closed

agentHits pushed a commit to agentHits/opencodex that referenced this pull request Sep 17, 2026

Merge pull request lidge-jun#1662 from TooSpace/fix/parallel-tool-cal…

fc6783f

…ls-pin-false fix(openai-chat): let opted-out providers pin parallel_tool_calls:false

Android-Login-Analysis mentioned this pull request Sep 19, 2026

tool choice policy is dropped on the Chat Completions path #5211

Closed

3 tasks

Navigation Menu

fix(openai-chat): let opted-out providers pin parallel_tool_calls:false - #1662

fix(openai-chat): let opted-out providers pin parallel_tool_calls:false#1662
Wibias merged 1 commit into
lidge-jun:devfrom
TooSpace:fix/parallel-tool-calls-pin-false

TooSpace commented Aug 14, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Aug 14, 2026

Uh oh!

coderabbitai Bot commented Aug 14, 2026 •

edited

Loading

Walkthrough

Changes

Uh oh!

github-actions Bot commented Aug 14, 2026 •

edited

Loading

Uh oh!

Wibias commented Aug 14, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

TooSpace commented Aug 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Why

Design / why opt-in

Changes

Testing

Review readiness checklist

Summary by CodeRabbit

Uh oh!

github-actions Bot commented Aug 14, 2026

Uh oh!

coderabbitai Bot commented Aug 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Uh oh!

github-actions Bot commented Aug 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ READY

Review readiness checklist

Uh oh!

Wibias commented Aug 14, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

TooSpace commented Aug 14, 2026 •

edited

Loading

coderabbitai Bot commented Aug 14, 2026 •

edited

Loading

github-actions Bot commented Aug 14, 2026 •

edited

Loading