GSP1289

Overview

Concerned about AI hallucinations? While AI can be a valuable resource, it sometimes generates inaccurate, outdated, or overly general responses - a phenomenon known as "hallucination." This lab teaches you how to implement a Retrieval Augmented Generation (RAG) pipeline to address this issue. RAG improves large language models (LLMs) like Gemini by grounding their output in contextually relevant information from a specific dataset.

Assume you are helping Coffee-on-Wheels, a pioneering mobile coffee vendor, analyze customer feedback on its services. Without access to the latest data, Gemini's responses might be inaccurate. To solve this problem, you decide to build a RAG pipeline that includes three steps:

1. Generate embeddings: Convert customer feedback text into vector embeddings, which are numerical representations of data that capture semantic meaning.
2. Search vector space: Create an index of these vectors, search for similar items, and retrieve them.
3. Generate improved answers: Augment Gemini with the retrieved information to produce more accurate and relevant responses.

BigQuery allows seamless connection to remote generative AI models on Vertex AI. It also provides various functions for embeddings, vector search, and text generation directly through SQL queries or Python notebooks.

For a deeper dive, check out the course Create Embeddings, Vector Search, and RAG with BigQuery on Google Cloud Skills Boost.

What you'll learn

• Create a source connection and grant IAM permissions.
• Generate embeddings and convert text data to vector embeddings.
• Search the vector space and retrieve similar items.
• Generate an improved answer by augmenting Gemini with the search results.

Prerequisites

To complete this lab, you should be familiar with BigQuery and SQL coding.

Setup and requirements

Before you click the Start Lab button

Read these instructions. Labs are timed and you cannot pause them. The timer, which starts when you click Start Lab, shows how long Google Cloud resources are made available to you.

This hands-on lab lets you do the lab activities in a real cloud environment, not in a simulation or demo environment. It does so by giving you new, temporary credentials you use to sign in and access Google Cloud for the duration of the lab.

To complete this lab, you need:

• Access to a standard internet browser (Chrome browser recommended).

• Time to complete the lab—remember, once you start, you cannot pause a lab.

How to start your lab and sign in to the Google Cloud console

1. Click the Start Lab button. If you need to pay for the lab, a dialog opens for you to select your payment method. On the left is the Lab Details pane with the following:

   • The Open Google Cloud console button
   • Time remaining
   • The temporary credentials that you must use for this lab
   • Other information, if needed, to step through this lab

2. Click Open Google Cloud console (or right-click and select Open Link in Incognito Window if you are running the Chrome browser).

   The lab spins up resources, and then opens another tab that shows the Sign in page.

   Tip: Arrange the tabs in separate windows, side-by-side.

   Note: If you see the Choose an account dialog, click Use Another Account.

3. If necessary, copy the Username below and paste it into the Sign in dialog.

   {{{user_0.username | "Username"}}}

   You can also find the Username in the Lab Details pane.

4. Click Next.

5. Copy the Password below and paste it into the Welcome dialog.

   {{{user_0.password | "Password"}}}

   You can also find the Password in the Lab Details pane.

6. Click Next.

   Important: You must use the credentials the lab provides you. Do not use your Google Cloud account credentials. Note: Using your own Google Cloud account for this lab may incur extra charges.

7. Click through the subsequent pages:

   • Accept the terms and conditions.
   • Do not add recovery options or two-factor authentication (because this is a temporary account).
   • Do not sign up for free trials.

After a few moments, the Google Cloud console opens in this tab.

Note: To access Google Cloud products and services, click the Navigation menu or type the service or product name in the Search field.

Task 1. Create a source connection and grant IAM permissions

Create a source connection

To use remote generative AI models on Vertex AI in BigQuery, like Gemini and an embedding model, create a new external source connection.

1. In the Google Cloud console, on the Navigation menu (), click BigQuery.

2. Switch to Explorer, click + Add data, and then use the search bar for data sources to search for Vertex AI. Click on the result for Vertex AI > BigQuery Federation.

3. In the Connection type dropdown, select Vertex AI remote models, remote functions, BigLake and Spanner (Cloud Resource).

4. In the Connection ID field, enter embedding_conn.

5. Click Create connection.

6. Once the connection is created, click on Go to connection in the pop-up confirmation to navigate to the connection and copy the Service account id value. You need it later to assign permissions to this account.

Grant IAM permissions

To use BigQuery data and Vertex AI resources, grant the service account the necessary IAM permissions.

1. Next, you need to grant permissions via IAM. Perform the steps that follow:
   • In the Google Cloud console, on the Navigation menu (), navigate to IAM & Admin > IAM.
   • Click on Grant access.
   • In the Add principals section:
     • In the New principals text field, paste the Service account id value that you copied earlier.
     • Under Assign Role, select the following roles (search for them if you need to):
       • BigQuery Data Owner
       • Vertex AI User
2. Click Save to apply the changes.
3. Navigate to APIs and Services from the Navigation menu (), click + Enable APIs and services, search Vertex AI API, click the Enable button.

Click Check my progress to verify the objective. Create a source connection and grant IAM permissions.

Task 2. Generate embeddings

1. In the Google Cloud console, on the Navigation menu (), navigate to BigQuery.

2. In Explorer, navigate to the three dots besides the project, click Create dataset. For Dataset ID, enter CustomerReview. Keep the other option by default, and click Create dataset.

3. To connect to the embedding model, run the following SQL query in the query editor:

4. To upload the dataset from a CSV file, run the following SQL query:

5. (optional) To check the uploaded data in the table, click Go to table. Find the schema of the table and preview the data.

6. To generate embeddings from recent customer feedback and store them in a table, run the following SQL query in the query editor:

7. (Optional) To examine the embedding results, click Go to table. Find the schema of the table and preview the data. Note that the embedding results are floating-point numbers and may not be immediately interpretable.

Click Check my progress to verify the objective. Generate embeddings.

Task 3. Search the vector space and retrieve the similar items

1. To create an index of the vector search space, run the following SQL query:

2. To search the vector space and retrieve the similar items, run the following SQL query:

3. (Optional) To check the query results, click Go to table. Find the schema of the table and preview the data.

Click Check my progress to verify the objective. Search the vector space and retrieve the similar items.

Task 4. Generate an improved answer

1. To connect to the Gemini model, run the following SQL query:

   CREATE OR REPLACE MODEL `CustomerReview.Gemini` REMOTE WITH CONNECTION `us.embedding_conn` OPTIONS (ENDPOINT = '{{{project_0.startup_script.gemini_flash_model_id | Gemini Model ID | disablehighlight}}}');

2. To enhance Gemini's responses, provide it with relevant and recent data retrieved from the vector search by running the following query:

3. Check the Gemini-generated results in the Query results section below the query editor.

Click Check my progress to verify the objective. Generate the enhanced response with data retrieved from the vector search.

Congratulations!

To help Coffee-on-Wheels gain insights from customer feedback on its services, you successfully implemented a RAG pipeline in BigQuery, providing Gemini with relevant and up-to-date information. You connected to remote generative AI models, including an embedding model and Gemini, and followed three steps: creating embeddings, searching a vector space, and generating an improved answer. The goal is to enable you to apply this same approach to address your own AI hallucination challenges.

Google Cloud training and certification

...helps you make the most of Google Cloud technologies. Our classes include technical skills and best practices to help you get up to speed quickly and continue your learning journey. We offer fundamental to advanced level training, with on-demand, live, and virtual options to suit your busy schedule. Certifications help you validate and prove your skill and expertise in Google Cloud technologies.

Manual last updated November 20, 2025

Lab last tested November 20, 2025

Copyright 2025 Google LLC. All rights reserved. Google and the Google logo are trademarks of Google LLC. All other company and product names may be trademarks of the respective companies with which they are associated.