Skip to content

Latest commit

 

History

History

google-search

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
images
images
 
 
README.md
README.md
 
 
main.py
main.py
 
 
requirements.txt
requirements.txt
 
 

README.md

Grounding with public data with Google Search

In this sample, you'll learn how to ground LLM with Vertex AI’s Google Search. More specifically, you'll:

  1. Ask the LLM questions that require recent data. For example, "How's the weather in London today?" and see that it cannot answer questions that require recent data.
  2. Ground the LLM with Vertex AI's Google Search.
  3. Ask the LLM the same questions again and see that you get answers back.

Without grounding

First, let's ask a question to the LLM about the weather without any grounding. You can do this in Vertex AI section of Google Cloud Console against any model.

Prompt:

What was the weather like in London yesterday?

Response:

I do not have access to real-time information, including weather data. To get the weather in London yesterday, I recommend checking a weather website or app.

Without grounding

As you can see, the model cannot answer questions about the weather yesterday, as it's not trained on that data.

There's also a main.py sample that you can run without grounding:

python main.py --project_id your-project-id

You get a similar response:

Prompt: What was the weather like in London yesterday?
Response text: I do not have access to real-time information, including weather data. To get the weather information for London yesterday, I recommend checking a reliable weather website or app.

Setup grounding with Google Search

Let's setup grounding with Google Search now.

Go back to Vertex AI section of Google Cloud Console and in the Advanced section, select Enable grounding:

Enable grounding

Customize grounding and point to Google Search:

Customize grounding

With grounding

Finally, we're ready to ask questions about the weather with grounding enabled.

Prompt:

What was the weather like in London yesterday?

Response:

Yesterday, June 17th, 2024, London had a high temperature of 73 degrees Fahrenheit and a low of 52 degrees Fahrenheit.

Also run main.py Python sample with grounding:

python main.py --project_id your-project-id --google_search_grounding

You get a similar response:

Prompt: What was the weather like in London yesterday?
Response text: Yesterday in London, the temperature ranged from a low of 52°F to a high of 70°F. The weather was mostly sunny with some passing clouds.

Yay, it works!

References