AttributeError: google.generativeai.models has no attribute 'Type' on macOS

[ripred]

Description of the bug:

Describe the bug

I'm encountering an AttributeError when trying to use the google-generativeai library on macOS. The error occurs even in a clean virtual environment and with the latest version of the library. The library seems unable to find the Type enum within the models module.

Steps to reproduce

1. Create a clean Python virtual environment:

   python3.12 -m venv .venv312  # Or use your preferred Python version
   source .venv312/bin/activate

2. Install the google-generativeai library:

   pip install google-generativeai

3. Create a simple Python script (test_genai.py) with the following content:

   #!/usr/bin/env python3
   import google.generativeai.models as glm
   
   print(glm.Type.OBJECT)
   print(glm.Type.STRING)

4. Run the script:

   python test_genai.py

Expected behavior

The script should print the enum values:

<Type.OBJECT: 'OBJECT'>
<Type.STRING: 'STRING'>

Actual behavior

The script raises an AttributeError:

Traceback (most recent call last):
File "/Users/trent/dev/gemini/arduino_project_manager/test_genai.py", line 4, in
print(glm.Type.OBJECT)
^^^^^^^^
AttributeError: module 'google.generativeai.models' has no attribute 'Type'

Environment

• Operating System: macOS [Your macOS Version - e.g., Sonoma 14.3.1]
• Python Version: [Your Python Version - e.g., 3.12.2] (Get this from python3.12 --version inside the venv). Also mention you tried 3.10.
• google-generativeai Version: [Version from pip show google-generativeai inside the venv - e.g., 0.8.4]
• Virtual Environment: Yes (and I tried multiple Python versions)

Additional context

I have tried the following troubleshooting steps:

• Uninstalled and reinstalled google-generativeai multiple times.
• Created a completely clean virtual environment.
• Tried different Python versions (3.10, 3.12).
• Checked for conflicting google-* packages in the virtual environment (none found).
• The same error occurred using a fresh virtual environment with Python 3.12

A workaround is to downgrade to google-generativeai==0.4.0 and use string literals ("object", "string") instead of glm.Type.OBJECT in the tool definitions. However, this is not ideal, as it prevents using newer features of the library.

Actual vs expected behavior:

No response

Any other information you'd like to share?

No response

[asyncmacro]

The error also occurs on Linux. I am not sure whether the developers did intentionally not export Type but It is also available on glm.protos.Type. for example: you can use glm.protos.Type.STRING.

@pamorgan can I make a change exposing Type?

[KSruthiVel]

@ripred I modified the code to:

import google.generativeai.models as glm

print(f"<Type.OBJECT: '{glm.protos.Type.OBJECT.name}'>")
print(f"<Type.STRING: '{glm.protos.Type.STRING.name}'>")

This produced the expected output:

<Type.OBJECT: 'OBJECT'>
<Type.STRING: 'STRING'>

The Type enum is only accessible via glm.protos.Type, and by default, glm.protos.Type.OBJECT and glm.protos.Type.STRING return integer values (6 and 1, respectively). Using .name allows retrieving the symbolic representation as expected.