Automatic function calling rejects `datetime` and `UUID` arguments

Aberta
  • Algo errado?
#3,331 1 comentário 0 reações 1 responsável Ver no GitHub

Mantenedores costumam responder em até 1 dia

@Venkaiahbabuneelam já está trabalhando nisso.

Desde 11/11/2026.

  • #3333 de @copybara-service — aberto

Avaliação

Dificuldade
3/5
Tempo estimado
1-2 dias
Facilidade para iniciantes
75/100
Tipo de issue
Bug
Clareza
Claramente especificada
Status de atividade
Ativa
Stack de tecnologia
python
Domínio
backend-api-design

Direção de pesquisa

Start by reading convert_if_exist_pydantic_model, which the issue identifies as the point rejecting JSON strings for datetime and UUID, and compare its validation with the pydantic schema generation. Check the automatic function calling tests for argument conversion and add coverage for both types. Done when valid schema-compatible strings reach the tool as converted arguments and the tests pass.

Escrita pelo modelo de indexação a partir do texto da issue.

Descrição

priority: p2 type: bug
Environment details
  • Programming language: Python
  • OS: macOS
  • Language runtime version: Python 3.12.13
  • Package version: google-genai 3.0.0 (same on 2.24.0 and main)
Steps to reproduce

With automatic function calling, a tool that takes a datetime or a UUID never runs. The declaration tells the model to send a date-time (or uuid) string, the model does, and google-genai rejects it.

This fakes the Gemini API, so no key is needed:

import datetime, json, uuid

import httpx
from google import genai
from google.genai import types

ran, sent = [], []


def schedule(when: datetime.datetime, job_id: uuid.UUID) -> str:
    """Schedule a job."""
    ran.append(True)
    return "ok"


def fake_gemini(request):
    sent.append(json.loads(request.content))
    args = {"when": "2026-11-10T09:30:00Z", "job_id": "12345678-1234-5678-1234-567812345678"}
    part = {"functionCall": {"name": "schedule", "args": args}} if len(sent) == 1 else {"text": "done"}
    return httpx.Response(200, json={"candidates": [{"content": {"role": "model", "parts": [part]}}]})


client = genai.Client(
    api_key="fake",
    http_options=types.HttpOptions(httpx_client=httpx.Client(transport=httpx.MockTransport(fake_gemini))),
)
client.models.generate_content(
    model="gemini-2.5-flash", contents="Schedule it.", config=types.GenerateContentConfig(tools=[schedule])
)

print("declared:", sent[0]["tools"][0]["functionDeclarations"][0]["parameters_json_schema"]["properties"])
print("ran:", bool(ran))
print("sent back:", sent[1]["contents"][-1]["parts"][0]["functionResponse"]["response"])

uv run --no-project --python 3.12 --with "google-genai[httpx]==3.0.0" python repro.py prints:

declared: {'when': {'format': 'date-time', 'type': 'string'}, 'job_id': {'format': 'uuid', 'type': 'string'}}
ran: False
sent back: {'error': "Failed to parse parameter when for function schedule from function call part because function call argument value 2026-11-10T09:30:00Z is not compatible with parameter annotation <class 'datetime.datetime'>."}

job_id fails the same way once when is gone. The call still completes, so nothing says the tool never ran.

The cause: the schema comes from pydantic, but convert_if_exist_pydantic_model ends with isinstance(value, annotation), and a JSON string is never a datetime or UUID. Validating with pydantic too, e.g. pydantic.TypeAdapter(annotation).validate_python(value), would make them agree.

Related, for other types: #2013 (set[str] and other generics, fix in #3295) and #2199 (IntEnum, fix in #3288). From their diffs, neither PR covers datetime or UUID.

Thanks!

Linguagem predominante
Python
Estrelas
4.1k
Forks
1.1k
Merge médio
1d 20h
PRs com merge (30d)
90

Preparar o ambiente

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de googleapis/python-genai

Todas as issues de googleapis/python-genai

Issues semelhantes

Mais issues de Python

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.