NoxzRCW b8c7ee599d fix(translator): keep upstream usage from trailing empty-choices chunks (#11883) ...

openaiToClaudeResponse() returned early on !chunk.choices?.[0], dropping the trailing usage-only chunk many OpenAI-compatible upstreams send when stream_options.include_usage is set (confirmed on Fireworks kimi-k3) — state.usage stayed undefined and billing fell back to an uncached token estimate. 154/154 focused assertions across the fix + regression suite. Thanks for tracking down the billing impact!

2026-08-28 15:49:14 -03:00

..

_helpers

…

_setup

feat(combo): choose sort method for combo models (manual/provider/score/name) (#11812)

2026-08-28 12:37:58 -03:00

benchmarks

…

boundary

…

e2e

chore(release): merge release/v3.8.50 tip into release/v3.8.51 — sync-back step 1/2

2026-08-28 14:07:39 -03:00

fixtures

fix(responses-continuation): recover a real id/output for passthrough and translate-mode replies (#11434)

2026-08-24 19:57:12 -03:00

golden-set

…

helpers

…

homolog

…

integration

test(security): drop the guard-target set left unused by the sync

2026-08-28 15:19:29 -03:00

live

…

llm-security

…

load

…

manual

…

security

…

snapshots

fix(providers): retire common ChatGPT Web provider (#11754)

2026-08-28 06:52:46 -03:00

translator

…

unit

fix(translator): keep upstream usage from trailing empty-choices chunks (#11883)

2026-08-28 15:49:14 -03:00

_cp_mock_hook.mts

…

_cp_mock_module.mts

…

_run_dns_guard_test.sh

…

e2e-tool-calls.sh

…

theoldllm-stress.test.ts

…