[HouMinXi]

Bug

On the provider detail page, bulk-deleting API keys (moonshot-native in this case) shows "deleted N connections" and then immediately "network error during batch delete". The keys are actually gone. The failure toast is a lie.

Fixes: dashboard bulk-delete success+failure toast pair.

Root Cause

ConfirmModal binds onConfirm to <Button onClick={onConfirm}>. A native click passes a MouseEvent.

handleBatchDeleteConfirm(onAfter?) treated any truthy first argument as a post-delete callback. After DELETE 200 it toasted success, then await onAfter() invoked the MouseEvent object, which throws, and the catch toasted batchDeleteNetworkError.

API side is fine: DELETE /api/providers with { ids } returns 200 and deletes the rows. The second toast is UI-only.

Measured 2026-09-04 against handleBatchDeleteConfirm + ConfirmModal wiring. Reproduction: pass a { type: "click" } stand-in as the first argument after a 200; success fires, then the network-error toast.

Fix

• Run onAfter only when typeof onAfter === "function".
• ConfirmModal confirm click: onClick={() => void onConfirm()} so the event never reaches the handler.

How to Verify

1. Open a provider with several API keys (moonshot-native / any openai-compatible node).
2. Select more than one key. Confirm bulk delete.
3. Expect one success toast. No follow-up failure toast. Keys gone from the list.

Test Plan

• tests/unit/ui/batch-delete-click-event.test.tsx — click-event stand-in after 200 does not toast error; a real onAfter still runs
• Injection: revert typeof check → the click-event case goes red; restore → green
• tests/unit/ui/confirm-delete-connection-7361.test.tsx still 3/3
• tests/unit/providers-route-managed-catalog.test.ts still 6/6 (includes batch DELETE)

Risk

Low. ConfirmModal still calls onConfirm() with no args, which matches its type (() => void | Promise<void>). onAfter is unused in production callers today; the typeof guard is the belt, the ConfirmModal wrap is the suspenders.

Not in this PR

No chunking of DELETE past the existing 100-id API cap. Activate/deactivate already chunk; delete does not. Separate if you bulk-delete >100 at once.

Also in this commit (base hygiene that blocked CI)

• Prefix unused useProviderConnections arg as _isSearchProvider (eslint @typescript-eslint/no-unused-vars on the file this PR already touches).
• changelog.d/fixes/reset-aware-model-family.md now starts with - (fragment gate).
• Register tests/unit/reset-aware-request-scope-12600.test.ts in stryker.conf.json tap.testFiles (mutation-test-coverage --strict).
• open-sse/executors/glm.ts: drop the extra 16th arg to createSSETransformStreamWithLogger (TS2554; the 65536 never reached TransformStream).

Not in this PR: HuggingChat/stream public-error unit failures on release/v3.8.51, and local-only forge daemon_state (.code-forge/ is gitignored).

[HouMinXi]

Technical review (LOCAL, review-default, baseline c3945a724) on 6aac03b83.

HOLD, 0 CONFIRMED, 2 UNCERTAIN. Not restarting rounds.

Adjudication:

• typeof onAfter === "function" in handleBatchDeleteConfirm: this is the bug. ConfirmModal used to bind onClick={onConfirm}, so React passed a MouseEvent into onAfter. The truthiness check then called event(), which threw and the catch toasted batchDeleteNetworkError after a successful DELETE. tests/unit/ui/batch-delete-click-event.test.tsx covers that path. Other ConfirmModal callers either wrap with () => void handler() or pass a zero-arg confirm; only this handler accepted an optional callback and treated a truthy argument as callable.
• onClick={() => void onConfirm()} on ConfirmModal: not cosmetic. ConfirmModalProps.onConfirm is () => void | Promise<void> — no event. The wrapper stops React from injecting the MouseEvent. Together with the typeof guard, the click event never reaches the batch-delete handler.

glm.ts dropping the extra positional is the TS2554 fix the other open PRs inherit from this branch. Unused KeyHealth / translateJsonResponse were already on the tip.

CI remaining: merge-integrity (cli-tunnel skill dry-run generate + tip changelog fragment without - ), unit shards. Same fingerprint as #12732. Typecheck / FQG on this PR are green.

Contributor only — not merging.

[HouMinXi]

Rebased onto upstream/release/v3.8.51 @ 9d1a896c6 (#12670).

range-diff every unique commit =. All rewritten commits GPG G. behind=0 vs tip.

Own tests re-run on the new base. --force-with-lease to the previous fork SHA.

⚠️ base-red inherited: #12732 (FQG / unit shards / changelog-integrity / glm arity until #12770 lands). Files this PR does not touch.

[HouMinXi]

Rebased onto release/v3.8.51 tip f9a1cc8a9 (#12682 / #12691 / #12834).

cb6a70c37 → 45c645244. range-diff 2 commits =. GPG G. No file overlap with the three landed commits.

[HouMinXi]

Rebased onto upstream/release/v3.8.51 b345c7f6c (#12870 OpenCode v2 plugin).

• 45c645244b3 → b3503bc4194
• range-diff identity (eq=2); all commits GPG G
• 2 unique commit(s) ahead, 0 behind
• --force-with-lease to HouMinXi:fix/batch-delete-click-event