Parse PDFs in a retrieval-augmented generation pipeline

This tutorial guides you through the process of creating a retrieval-augmented generation (RAG) pipeline based on parsed PDF content.

PDF files, such as financial documents, can be challenging to use in RAG pipelines because of their complex structure and mix of text, figures, and tables. This tutorial shows you how to use BigQuery ML capabilities in combination with Document AI's Layout Parser to build a RAG pipeline based on key information extracted from a PDF file.

You can alternatively perform this tutorial by using a Colab Enterprise notebook.

Objectives

This tutorial covers the following tasks:

• Creating a Cloud Storage bucket and uploading a sample PDF file.
• Creating a Cloud resource connection so that you can connect to Cloud Storage and Vertex AI from BigQuery.
• Creating an object table over the PDF file to make the PDF file available in BigQuery.
• Creating a Document AI processor that you can use to parse the PDF file.
• Creating a remote model that lets you use the Document AI API to access the document processor from BigQuery.
• Using the remote model with the ML.PROCESS_DOCUMENT function to parse the PDF contents into chunks and then write that content to a BigQuery table.
• Extracting PDF content from the JSON data returned by the ML.PROCESS_DOCUMENT function, and then writing that content to a BigQuery table.
• Creating a remote model that lets you use the Vertex AI text-embedding-004 embedding generation model from BigQuery.
• Using the remote model with the ML.GENERATE_EMBEDDING function to generate embeddings from the parsed PDF content, and then writing those embeddings to a BigQuery table. Embeddings are numerical representations of the PDF content that enable you to perform semantic search and retrieval on the PDF content.
• Using the VECTOR_SEARCH function on the embeddings to identify semantically similar PDF content.
• Creating a remote model that lets you use the Vertex AI gemini-1.5-flash text generation model from BigQuery.
• Perform retrieval-augmented generation (RAG) by using the remote model with the ML.GENERATE_TEXT function to generate text, using vector search results to augment the prompt input and improve results.

Costs

In this document, you use the following billable components of Google Cloud:

• BigQuery: You incur costs for the data that you process in BigQuery.
• Vertex AI: You incur costs for calls to Vertex AI models.
• Document AI: You incur costs for calls to the Document AI API.
• Cloud Storage: You incur costs for object storage in Cloud Storage.

To generate a cost estimate based on your projected usage, use the pricing calculator. New Google Cloud users might be eligible for a free trial.

For more information, see the following pricing pages:

• BigQuery pricing
• Vertex AI pricing
• Document AI pricing
• Cloud Storage pricing

Before you begin

1. In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

   Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

   Go to project selector

2. Make sure that billing is enabled for your Google Cloud project.

3. Enable the BigQuery, BigQuery Connection, Vertex AI, Document AI, and Cloud Storage APIs.

   Enable the APIs

Required roles

• To create Cloud Storage buckets and objects, you need membership in the roles/storage.storageAdmin role.

• To create a Document AI processor, you need membership in the roles/documentai.editor role.

• To create a connection, you need membership in the roles/bigquery.connectionAdmin role.

• To grant permissions to the connection's service account, you need membership in the roles/resourcemanager.projectIamAdmin role.

• The IAM permissions needed in this tutorial for the remaining BigQuery operations are included in the following two roles:

  • BigQuery Data Editor (roles/bigquery.dataEditor) to create models, tables, and indexes.
  • BigQuery User (roles/bigquery.user) to run BigQuery jobs.

Create a dataset

Create a BigQuery dataset to store your ML model:

1. In the Google Cloud console, go to the BigQuery page.

   Go to the BigQuery page

2. In the Explorer pane, click your project name.

3. Click View actions > Create dataset.

   

4. On the Create dataset page, do the following:

   • For Dataset ID, enter bqml_tutorial.

   • For Location type, select Multi-region, and then select US (multiple regions in United States).

     The public datasets are stored in the US multi-region. For simplicity, store your dataset in the same location.

   • Leave the remaining default settings as they are, and click Create dataset.

     

Create a connection

Create a Cloud resource connection and get the connection's service account. Create the connection in the same location.

Select one of the following options:

Console

1. Go to the BigQuery page.

   Go to BigQuery

2. To create a connection, click Add, and then click Connections to external data sources.

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

4. In the Connection ID field, enter a name for your connection.

5. Click Create connection.

6. Click Go to connection.

7. In the Connection info pane, copy the service account ID for use in a later step.

bq

1. In a command-line environment, create a connection:

   bq mk --connection --location=REGION --project_id=PROJECT_ID \
       --connection_type=CLOUD_RESOURCE CONNECTION_ID

   The --project_id parameter overrides the default project.

   Replace the following:

   • REGION: your connection region
   • PROJECT_ID: your Google Cloud project ID
   • CONNECTION_ID: an ID for your connection

   When you create a connection resource, BigQuery creates a unique system service account and associates it with the connection.

   Troubleshooting: If you get the following connection error, update the Google Cloud SDK:

   Flags parsing error: flag --connection_type=CLOUD_RESOURCE: value should be one of...
   

2. Retrieve and copy the service account ID for use in a later step:

   bq show --connection PROJECT_ID.REGION.CONNECTION_ID

   The output is similar to the following:

   name                          properties
   1234.REGION.CONNECTION_ID     {"serviceAccountId": "[e-mail]"}
   

Terraform

Use the google_bigquery_connection resource.

Note: To create BigQuery objects using Terraform, you must enable the Cloud Resource Manager API.

To authenticate to BigQuery, set up Application Default Credentials. For more information, see Set up authentication for client libraries.

The following example creates a Cloud resource connection named my_cloud_resource_connection in the US region:

# This queries the provider for project information.
data "google_project" "default" {}

# This creates a cloud resource connection in the US region named my_cloud_resource_connection.
# Note: The cloud resource nested object has only one output field - serviceAccountId.
resource "google_bigquery_connection" "default" {
  connection_id = "my_cloud_resource_connection"
  project       = data.google_project.default.project_id
  location      = "US"
  cloud_resource {}
}

To apply your Terraform configuration in a Google Cloud project, complete the steps in the following sections.

Prepare Cloud Shell

1. Launch Cloud Shell.

2. Set the default Google Cloud project where you want to apply your Terraform configurations.

   You only need to run this command once per project, and you can run it in any directory.

   export GOOGLE_CLOUD_PROJECT=PROJECT_ID

   Environment variables are overridden if you set explicit values in the Terraform configuration file.

Prepare the directory

Each Terraform configuration file must have its own directory (also called a root module).

1. In Cloud Shell, create a directory and a new file within that directory. The filename must have the .tf extension—for example main.tf. In this tutorial, the file is referred to as main.tf.

   mkdir DIRECTORY && cd DIRECTORY && touch main.tf

2. If you are following a tutorial, you can copy the sample code in each section or step.

   Copy the sample code into the newly created main.tf.

   Optionally, copy the code from GitHub. This is recommended when the Terraform snippet is part of an end-to-end solution.

3. Review and modify the sample parameters to apply to your environment.
4. Save your changes.
5. Initialize Terraform. You only need to do this once per directory.

   terraform init

   Optionally, to use the latest Google provider version, include the -upgrade option:

   terraform init -upgrade

Apply the changes

1. Review the configuration and verify that the resources that Terraform is going to create or update match your expectations:

   terraform plan

   Make corrections to the configuration as necessary.

2. Apply the Terraform configuration by running the following command and entering yes at the prompt:

   terraform apply

   Wait until Terraform displays the "Apply complete!" message.

3. Open your Google Cloud project to view the results. In the Google Cloud console, navigate to your resources in the UI to make sure that Terraform has created or updated them.

Note: Terraform samples typically assume that the required APIs are enabled in your Google Cloud project.

Grant access to the service account

Select one of the following options:

Console

1. Go to the IAM & Admin page.

   Go to IAM & Admin

2. Click Grant Access.

   The Add principals dialog opens.

3. In the New principals field, enter the service account ID that you copied earlier.

4. In the Select a role field, select Document AI, and then select Document AI Viewer.

5. Click Add another role.

6. In the Select a role field, select Cloud Storage, and then select Storage Object Viewer.

7. Click Add another role.

8. In the Select a role field, select Vertex AI, and then select Vertex AI User.

9. Click Save.

gcloud

Use the gcloud projects add-iam-policy-binding command:

gcloud projects add-iam-policy-binding 'PROJECT_NUMBER' --member='serviceAccount:MEMBER' --role='roles/documentai.viewer' --condition=None
gcloud projects add-iam-policy-binding 'PROJECT_NUMBER' --member='serviceAccount:MEMBER' --role='roles/storage.objectViewer' --condition=None
gcloud projects add-iam-policy-binding 'PROJECT_NUMBER' --member='serviceAccount:MEMBER' --role='roles/aiplatform.user' --condition=None
 

Replace the following:

• PROJECT_NUMBER: your project number.
• MEMBER: the service account ID that you copied earlier.

Upload the sample PDF to Cloud Storage

To upload the sample PDF to Cloud Storage, follow these steps:

1. Download the scf23.pdf sample PDF by going to https://www.federalreserve.gov/publications/files/scf23.pdf and clicking download .
2. Create a Cloud Storage bucket.
3. Upload the scf23.pdf file to the bucket.

Create an object table

Create an object table over the PDF file in Cloud Storage:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, run the following statement:

   CREATE OR REPLACE EXTERNAL TABLE `bqml_tutorial.pdf`
   WITH CONNECTION `LOCATION.CONNECTION_ID`
   OPTIONS(
     object_metadata = 'SIMPLE',
     uris = ['gs://BUCKET/scf23.pdf']);

   Replace the following:

   • LOCATION: the connection location.
   • CONNECTION_ID: the ID of your BigQuery connection.

     When you view the connection details in the Google Cloud console, the CONNECTION_ID is the value in the last section of the fully qualified connection ID that is shown in Connection ID, for example projects/myproject/locations/connection_location/connections/myconnection.

   • BUCKET: the Cloud Storage bucket containing the scf23.pdffile. The full uri option value should look similar to ['gs://mybucket/scf23.pdf'].

Create a document processor

Create a document processor based on the Layout Parser processor in the us multi-region.

Create the remote model for the document processor

Create a remote model to access the Document AI processor:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, run the following statement:

   CREATE OR REPLACE MODEL `bqml_tutorial.parser_model`
   REMOTE WITH CONNECTION `LOCATION.CONNECTION_ID`
     OPTIONS(remote_service_type = 'CLOUD_AI_DOCUMENT_V1', document_processor = 'PROCESSOR_ID');

   Replace the following:

   • LOCATION: the connection location.
   • CONNECTION_ID: the ID of your BigQuery connection.

     When you view the connection details in the Google Cloud console, the CONNECTION_ID is the value in the last section of the fully qualified connection ID that is shown in Connection ID, for example projects/myproject/locations/connection_location/connections/myconnection.

   • PROCESSOR_ID: the document processor ID. To find this value, view the processor details, and then look at the ID row in the Basic Information section.

Parse the PDF file into chunks

Use the document processor with the ML.PROCESS_DOCUMENT function to parse the PDF file into chunks, and then write that content to a table. The ML.PROCESS_DOCUMENT function returns the PDF chunks in JSON format.

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, run the following statement:

   CREATE or REPLACE TABLE bqml_tutorial.chunked_pdf AS (
     SELECT * FROM ML.PROCESS_DOCUMENT(
     MODEL bqml_tutorial.parser_model,
     TABLE bqml_tutorial.pdf,
     PROCESS_OPTIONS => (JSON '{"layout_config": {"chunking_config": {"chunk_size": 250}}}')
     )
   );

Parse the PDF chunk data into separate columns

Extract the PDF content and metadata information from the JSON data returned by the ML.PROCESS_DOCUMENT function, and then write that content to a table:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, run the following statement to parse the PDF content:

   CREATE OR REPLACE TABLE bqml_tutorial.parsed_pdf AS (
   SELECT
     uri,
     JSON_EXTRACT_SCALAR(json , '$.chunkId') AS id,
     JSON_EXTRACT_SCALAR(json , '$.content') AS content,
     JSON_EXTRACT_SCALAR(json , '$.pageFooters[0].text') AS page_footers_text,
     JSON_EXTRACT_SCALAR(json , '$.pageSpan.pageStart') AS page_span_start,
     JSON_EXTRACT_SCALAR(json , '$.pageSpan.pageEnd') AS page_span_end
   FROM bqml_tutorial.chunked_pdf, UNNEST(JSON_EXTRACT_ARRAY(ml_process_document_result.chunkedDocument.chunks, '$')) json
   );

3. In the query editor, run the following statement to view a subset of the parsed PDF content:

   SELECT *
   FROM `bqml_tutorial.parsed_pdf`
   ORDER BY id
   LIMIT 5;

   The output is similar to the following:

   +-----------------------------------+------+------------------------------------------------------------------------------------------------------+-------------------+-----------------+---------------+
   |                uri                |  id  |                                                 content                                              | page_footers_text | page_span_start | page_span_end |
   +-----------------------------------+------+------------------------------------------------------------------------------------------------------+-------------------+-----------------+---------------+
   | gs://mybucket/scf23.pdf           | c1   | •BOARD OF OF FEDERAL GOVERN NOR RESERVE SYSTEM RESEARCH & ANALYSIS                                   | NULL              | 1               | 1             |
   | gs://mybucket/scf23.pdf           | c10  | • In 2022, 20 percent of all families, 14 percent of families in the bottom half of the usual ...    | NULL              | 8               | 9             |
   | gs://mybucket/scf23.pdf           | c100 | The SCF asks multiple questions intended to capture whether families are credit constrained, ...     | NULL              | 48              | 48            |
   | gs://mybucket/scf23.pdf           | c101 | Bankruptcy behavior over the past five years is based on a series of retrospective questions ...     | NULL              | 48              | 48            |
   | gs://mybucket/scf23.pdf           | c102 | # Percentiles of the Distributions of Income and Net Worth                                           | NULL              | 48              | 49            |
   +-----------------------------------+------+------------------------------------------------------------------------------------------------------+-------------------+-----------------+---------------+
    

Create the remote model for embedding generation

Create a remote model that represents a hosted Vertex AI text embedding generation model:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, run the following statement:

   CREATE OR REPLACE MODEL `bqml_tutorial.embedding_model`
     REMOTE WITH CONNECTION `LOCATION.CONNECTION_ID`
     OPTIONS (ENDPOINT = 'text-embedding-004');

   Replace the following:

   • LOCATION: the connection location.
   • CONNECTION_ID: the ID of your BigQuery connection.

     When you view the connection details in the Google Cloud console, the CONNECTION_ID is the value in the last section of the fully qualified connection ID that is shown in Connection ID, for example projects/myproject/locations/connection_location/connections/myconnection.

Generate embeddings

Generate embeddings for the parsed PDF content and then write them to a table:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, run the following statement:

   CREATE OR REPLACE TABLE `bqml_tutorial.embeddings` AS
   SELECT * FROM ML.GENERATE_EMBEDDING(
     MODEL `bqml_tutorial.embedding_model`,
     TABLE `bqml_tutorial.parsed_pdf`
   );

Run a vector search

Run a vector search against the parsed PDF content.

The following query takes text input, creates an embedding for that input using the ML.GENERATE_EMBEDDING function, and then uses the VECTOR_SEARCH function to match the input embedding with the most similar PDF content embeddings. The results are the top ten PDF chunks that are most semantically similar to the input.

1. Go to the BigQuery page.

   Go to BigQuery

2. In the query editor, run the following SQL statement:

   SELECT query.query, base.id AS pdf_chunk_id, base.content, distance
   FROM
     VECTOR_SEARCH( TABLE `bqml_tutorial.embeddings`,
       'ml_generate_embedding_result',
       (
       SELECT
         ml_generate_embedding_result,
         content AS query
       FROM
         ML.GENERATE_EMBEDDING( MODEL `bqml_tutorial.embedding_model`,
           ( SELECT 'Did the typical family net worth increase? If so, by how much?' AS content)
         )
       ),
       top_k => 10,
       OPTIONS => '{"fraction_lists_to_search": 0.01}')
   ORDER BY distance DESC;

   The output is similar to the following:

   +-------------------------------------------------+--------------+------------------------------------------------------------------------------------------------------+---------------------+
   |                query                            | pdf_chunk_id |                                                 content                                              | distance            |
   +-------------------------------------------------+--------------+------------------------------------------------------------------------------------------------------+---------------------+
   | Did the typical family net worth increase? ,... | c9           | ## Assets                                                                                            | 0.31113668174119469 |
   |                                                 |              |                                                                                                      |                     |
   |                                                 |              | The homeownership rate increased slightly between 2019 and 2022, to 66.1 percent. For ...            |                     |
   +-------------------------------------------------+--------------+------------------------------------------------------------------------------------------------------+---------------------+
   | Did the typical family net worth increase? ,... | c50          | # Box 3. Net Housing Wealth and Housing Affordability                                                | 0.30973592073929113 |
   |                                                 |              |                                                                                                      |                     |
   |                                                 |              | For families that own their primary residence ...                                                    |                     |
   +-------------------------------------------------+--------------+------------------------------------------------------------------------------------------------------+---------------------+
   | Did the typical family net worth increase? ,... | c50          | 3 In the 2019 SCF, a small portion of the data collection overlapped with early months of            | 0.29270064592817646 |
   |                                                 |              | the COVID- ...                                                                                       |                     |
   +-------------------------------------------------+--------------+------------------------------------------------------------------------------------------------------+---------------------+
    

Create the remote model for text generation

Create a remote model that represents a hosted Vertex AI text generation model:

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, run the following statement:

   CREATE OR REPLACE MODEL `bqml_tutorial.text_model`
     REMOTE WITH CONNECTION `LOCATION.CONNECTION_ID`
     OPTIONS (ENDPOINT = 'gemini-1.5-flash-002');

   Replace the following:

   • LOCATION: the connection location.
   • CONNECTION_ID: the ID of your BigQuery connection.

     When you view the connection details in the Google Cloud console, the CONNECTION_ID is the value in the last section of the fully qualified connection ID that is shown in Connection ID, for example projects/myproject/locations/connection_location/connections/myconnection.

Generate text augmented by vector search results

Perform a vector search on the embeddings to identify semantically similar PDF content, and then use the ML.GENERATE_TEXT function with the vector search results to augment the prompt input and improve the text generation results. In this case, the query uses information from the PDF chunks to answer a question about the change in family net worth over the past decade.

1. In the Google Cloud console, go to the BigQuery page.

   Go to BigQuery

2. In the query editor, run the following statement:

   SELECT
     ml_generate_text_llm_result AS generated
     FROM
     ML.GENERATE_TEXT( MODEL `bqml_tutorial.text_model`,
       (
       SELECT
       CONCAT( 'Did the typical family net worth change? How does this compare the SCF survey a decade earlier? Be concise and use the following context:',
       STRING_AGG(FORMAT("context: %s and reference: %s", base.content, base.uri), ',\n')) AS prompt,
       FROM
         VECTOR_SEARCH( TABLE
           `bqml_tutorial.embeddings`,
           'ml_generate_embedding_result',
           (
           SELECT
             ml_generate_embedding_result,
             content AS query
           FROM
             ML.GENERATE_EMBEDDING( MODEL `bqml_tutorial.embedding_model`,
               (
               SELECT
                 'Did the typical family net worth change? How does this compare the SCF survey a decade earlier?' AS content
               )
             )
           ),
           top_k => 10,
           OPTIONS => '{"fraction_lists_to_search": 0.01}')
         ),
         STRUCT(512 AS max_output_tokens, TRUE AS flatten_json_output)
     );

   The output is similar to the following:

   +-------------------------------------------------------------------------------+
   |               generated                                                       |
   +-------------------------------------------------------------------------------+
   | Between the 2019 and 2022 Survey of Consumer Finances (SCF), real median      |
   | family net worth surged 37 percent to $192,900, and real mean net worth       |
   | increased 23 percent to $1,063,700.  This represents the largest three-year   |
   | increase in median net worth in the history of the modern SCF, exceeding the  |
   | next largest by more than double.  In contrast, between 2010 and 2013, real   |
   | median net worth decreased 2 percent, and real mean net worth remained        |
   | unchanged.                                                                    |
   +-------------------------------------------------------------------------------+
    

Clean up

Caution: Deleting a project has the following effects:

• Everything in the project is deleted. If you used an existing project for the tasks in this document, when you delete it, you also delete any other work you've done in the project.
• Custom project IDs are lost. When you created this project, you might have created a custom project ID that you want to use in the future. To preserve the URLs that use the project ID, such as an appspot.com URL, delete selected resources inside the project instead of deleting the whole project.

If you plan to explore multiple architectures, tutorials, or quickstarts, reusing projects can help you avoid exceeding project quota limits.

1. In the Google Cloud console, go to the Manage resources page.

   Go to Manage resources

2. In the project list, select the project that you want to delete, and then click Delete.
3. In the dialog, type the project ID, and then click Shut down to delete the project.