Feature: Added safety_settings parameter for gemini

[akumar-b6i]

Description

Currently, the Vertex API does not provide handling for safety_settings to Gemini. This update will enable users to configure safety_settings during the initialization of the Vertex module.

Vertex AI GenerativeMoel Reference : https://cloud.google.com/python/docs/reference/aiplatform/latest/vertexai.generative_models.GenerativeModel

Fixes # (issue)

New Package?

Did I fill in the tool.llamahub section in the pyproject.toml and provide a detailed README.md for my new integration or package?

• No

Version Bump?

Did I bump the version in the pyproject.toml file of the package I am updating? (Except for the llama-index-core package)

• No
  Have to confirm the right spot to update. Comment below.

Type of Change

Please delete options that are not relevant.

• New feature (non-breaking change which adds functionality)

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration

• I stared at the code and made sure it makes sense
• Ran the test file in the respective folder and it passed
• Ran below test code to confirm :

import pytest
from llama_index.llms.vertex import Vertex

def test_vertex_initialization():
    safety_settings = {
        'HarmCategory.HARM_CATEGORY_HATE_SPEECH': 'HarmBlockThreshold.BLOCK_LOW_AND_ABOVE',
        'HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT': 'HarmBlockThreshold.BLOCK_LOW_AND_ABOVE'
    }

    # Create an instance of the Vertex class
    llm = Vertex(
        model="gemini-pro",
        project='GCP_PROJECT_ID',
        safety_settings=safety_settings,
        location="GCP_PROJECT_LOCATION",
    )

    # Check if safety_settings is correctly set

    assert llm.safety_settings == safety_settings

    # # Check if safety_settings is of correct type
    assert isinstance(llm.safety_settings, dict)

    # # Check if safety_settings keys are correct
    assert set(llm.safety_settings.keys()) == set(['HarmCategory.HARM_CATEGORY_HATE_SPEECH', 'HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT'])

    # # Check if safety_settings values are correct
    assert set(llm.safety_settings.values()) == set(['HarmBlockThreshold.BLOCK_LOW_AND_ABOVE', 'HarmBlockThreshold.BLOCK_LOW_AND_ABOVE'])

Result: PASS

Suggested Checklist:

• I have performed a self-review of my own code

[akumar-b6i]

Feature Request : #10788

[logan-markewich]

This looks good to me. If you bump the version of the gemini llm integration package, it will auto publish on merge

[akumar-b6i]

This looks good to me. If you bump the version of the gemini llm integration package, it will auto publish on merge

I am a bit confused by the above wording, the change is in llama-index-llms-vertex so should the version update be here

authors = ["Your Name <you@example.com>"]
description = "llama-index llms vertex integration"
exclude = ["**/BUILD"]
license = "MIT"
name = "llama-index-llms-vertex"
readme = "README.md"
version = "0.1.5"

Or in the gemini folder here

[logan-markewich]

Whoops, in the vertex package (for some reason I assumed this change was in the gemini package lol)

[akumar-b6i]

Whoops, in the vertex package (for some reason I assumed this change was in the gemini package lol)

Done, let me know if it looks good.