AttributeError: 'NoneType' object has no attribute 'upper'

[vladislavkoz]

• [ ] This is actually a bug report.

What Model are you using?

• [ ] gemini-2.0-flash

instructor==1.7.8, google-genai==1.9.0

Describe the bug I use the following client self.client = instructor.from_genai( client=genai.Client( vertexai=True, project=app_config.gcp_project_id, location=app_config.gcp_vertex_region, ), mode=self.extractor_config.gemini.instructor_mode, # InstructorMode.GENAI_STRUCTURED_OUTPUTS )

Also, I use "self.client.chat.completions.create_with_completion" with response_model: Pydantic model. My response model contains Optional fields: field_1: Optional[str] = None

To Reproduce Call create_with_completion() func.

Screenshots

I assume it's because google doesn't support optional fields. Just want to confirm that it's the reason.

Side note, in addition to this, I had to add "jsonref" to my dependencies. Not sure whether it's instructor or "google-genai" thing.

[vladislavkoz]

BTW, it was possible to use Optional fields with "google-generativeai" client + InstructorMode.GEMINI_JSON

Now with "google-genai" only: "InstructorMode.GENAI_STRUCTURED_OUTPUTS" and "InstructorMode.GENAI_TOOLS" allowed and both have issues with optional fields

I assume that it's related to -> https://github.com/instructor-ai/instructor/issues/1442

[ivanleomk]

Hmm Gemini_JSON was parsing the raw chat completion, that's how we got the optionals. I've tried to avoid that and instead use just structured outputs and tool calling for the genai sdk.

Could you give this a try and see if it works for you now?

from pydantic import BaseModel
from typing import Optional
from google import genai
import instructor
import os
from google import generativeai


class User(BaseModel):
    name: str
    age: Optional[int]
    nationality: Optional[str]


client = instructor.from_genai(
    genai.Client(api_key=os.getenv("GOOGLE_API_KEY")),
    mode=instructor.Mode.GENAI_STRUCTURED_OUTPUTS,
)

resp = client.chat.completions.create(
    model="gemini-2.0-flash",
    messages=[{"role": "user", "content": "Ivan is a software engineer"}],
    response_model=User,
)

print(resp)

[vladislavkoz]

I didn't try it yet, will check a bit later, but I already see at least 1 diff. I use vertexAI(instead of AI studio with api key).

[vladislavkoz]

@ivanleomk Same error with the suggested solution.

    schema_type = schema_type.upper()
AttributeError: 'NoneType' object has no attribute 'upper'

[jxnl]

Recent fixes to Gemini streaming and schema handling (commits 61718fb6, 02f4b0bb, 61dce8f7) likely resolve this. Please verify with latest version.