Google Vertex¶

In this quickstart you will learn how to run evaluation functions using models from google Vertex.

In [ ]:

# !pip install trulens trulens-providers-google

# !pip install trulens trulens-providers-google

In [ ]:

from google import genai

PROJECT_ID = "gemini-experiments-467318"  # 👈 Replace with your real project ID if different
LOCATION = "us-central1"

vertex_client = genai.Client(
    vertexai=True, project=PROJECT_ID, location=LOCATION
)

from google import genai PROJECT_ID = "gemini-experiments-467318" # 👈 Replace with your real project ID if different LOCATION = "us-central1" vertex_client = genai.Client( vertexai=True, project=PROJECT_ID, location=LOCATION )

In [ ]:

response = vertex_client.models.generate_content(
    model="gemini-2.0-flash",
    contents="Tell me about trulens in a few words",
)
print(response.text)

response = vertex_client.models.generate_content( model="gemini-2.0-flash", contents="Tell me about trulens in a few words", ) print(response.text)

Create Simple LLM Application¶

In [ ]:

from trulens.apps.app import instrument
from trulens.core import TruSession


session = TruSession()
session.reset_database()

from trulens.apps.app import instrument from trulens.core import TruSession session = TruSession() session.reset_database()

In [ ]:

class PromptProcessor:
    @instrument
    def generate_completion(self, user_prompt:str) -> str:
        """
        Generate answer from context.
        """
        prompt = f"""Provide a helpful response with relevant background information for the following: {user_prompt}"""

        resp = vertex_client.models.generate_content(
            model="gemini-2.0-flash",
            contents=prompt,
        )

        return resp.text


simple_application = PromptProcessor()

class PromptProcessor: @instrument def generate_completion(self, user_prompt:str) -> str: """ Generate answer from context. """ prompt = f"""Provide a helpful response with relevant background information for the following: {user_prompt}""" resp = vertex_client.models.generate_content( model="gemini-2.0-flash", contents=prompt, ) return resp.text simple_application = PromptProcessor()

Initialize Feedback Function(s)¶

In [ ]:

from trulens.core import Feedback
from trulens.providers.google import Google

google_vertex_client = Google(vertexai=True, project=PROJECT_ID, location=LOCATION)
relevance = Feedback(google_vertex_client.relevance).on_input_output()
# By default this will check relevance on the main app input and main app
# output.

from trulens.core import Feedback from trulens.providers.google import Google google_vertex_client = Google(vertexai=True, project=PROJECT_ID, location=LOCATION) relevance = Feedback(google_vertex_client.relevance).on_input_output() # By default this will check relevance on the main app input and main app # output.

Construct the app¶

Wrap the custom application with TruApp, add list of feedbacks for eval

In [ ]:

from trulens.apps.app import TruApp

tru_app = TruApp(
    simple_application,
    app_name="RAG",
    app_version="v1",
    feedbacks=[
        relevance
    ],
)

from trulens.apps.app import TruApp tru_app = TruApp( simple_application, app_name="RAG", app_version="v1", feedbacks=[ relevance ], )

Run the app¶

Use tru_app as a context manager for the custom PromptProcessor app.

In [ ]:

with tru_app as recording:
    resp = simple_application.generate_completion("What is a good name for a store that sells colorful socks?")

with tru_app as recording: resp = simple_application.generate_completion("What is a good name for a store that sells colorful socks?")

In [ ]:

print(resp)

print(resp)

Explore in a Dashboard¶

In [ ]:

from trulens.dashboard import run_dashboard

run_dashboard(session)

from trulens.dashboard import run_dashboard run_dashboard(session)