在檢索增強生成管道中剖析 PDF

建立資料集

建立 BigQuery 資料集來儲存機器學習模型。

{
  "datasetReference": {
     "datasetId": "bqml_tutorial"
  }
}

建立連線

建立Cloud 資源連線，並取得連線的服務帳戶。在相同位置建立連線。

如果已設定預設連線，或具備 BigQuery 管理員角色，可以略過這個步驟。

import google.api_core.exceptions
from google.cloud import bigquery_connection_v1

client = bigquery_connection_v1.ConnectionServiceClient()


def create_connection(
    project_id: str,
    location: str,
    connection_id: str,
):
    """Creates a BigQuery connection to a Cloud Resource.

    Cloud Resource connection creates a service account which can then be
    granted access to other Google Cloud resources for federated queries.

    Args:
        project_id: The Google Cloud project ID.
        location: The location of the connection (for example, "us-central1").
        connection_id: The ID of the connection to create.
    """

    parent = client.common_location_path(project_id, location)

    connection = bigquery_connection_v1.Connection(
        friendly_name="Example Connection",
        description="A sample connection for a Cloud Resource.",
        cloud_resource=bigquery_connection_v1.CloudResourceProperties(),
    )

    try:
        created_connection = client.create_connection(
            parent=parent, connection_id=connection_id, connection=connection
        )
        print(f"Successfully created connection: {created_connection.name}")
        print(f"Friendly name: {created_connection.friendly_name}")
        print(
            f"Service Account: {created_connection.cloud_resource.service_account_id}"
        )

    except google.api_core.exceptions.AlreadyExists:
        print(f"Connection with ID '{connection_id}' already exists.")
        print("Please use a different connection ID.")
    except Exception as e:
        print(f"An unexpected error occurred while creating the connection: {e}")

const {ConnectionServiceClient} =
  require('@google-cloud/bigquery-connection').v1;
const {status} = require('@grpc/grpc-js');

const client = new ConnectionServiceClient();

/**
 * Creates a new BigQuery connection to a Cloud Resource.
 *
 * A Cloud Resource connection creates a service account that can be granted access
 * to other Google Cloud resources.
 *
 * @param {string} projectId The Google Cloud project ID. for example, 'example-project-id'
 * @param {string} location The location of the project to create the connection in. for example, 'us-central1'
 * @param {string} connectionId The ID of the connection to create. for example, 'example-connection-id'
 */
async function createConnection(projectId, location, connectionId) {
  const parent = client.locationPath(projectId, location);

  const connection = {
    friendlyName: 'Example Connection',
    description: 'A sample connection for a Cloud Resource',
    // The service account for this cloudResource will be created by the API.
    // Its ID will be available in the response.
    cloudResource: {},
  };

  const request = {
    parent,
    connectionId,
    connection,
  };

  try {
    const [response] = await client.createConnection(request);

    console.log(`Successfully created connection: ${response.name}`);
    console.log(`Friendly name: ${response.friendlyName}`);

    console.log(`Service Account: ${response.cloudResource.serviceAccountId}`);
  } catch (err) {
    if (err.code === status.ALREADY_EXISTS) {
      console.log(`Connection '${connectionId}' already exists.`);
    } else {
      console.error(`Error creating connection: ${err.message}`);
    }
  }
}

# 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 {}
}

將存取權授予服務帳戶

選取下列選項之一：

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
 

將範例 PDF 上傳至 Cloud Storage

如要將範例 PDF 上傳至 Cloud Storage，請按照下列步驟操作：

1. 前往 https://www.federalreserve.gov/publications/files/scf23.pdf，然後按一下下載圖示 download，即可下載 scf23.pdf 範例 PDF。
2. 建立 Cloud Storage bucket。
3. 上傳 scf23.pdf 檔案至 bucket。

建立物件資料表

在 Cloud Storage 中的 PDF 檔案上建立物件資料表：

1. 前往 Cloud de Confiance 控制台的「BigQuery」頁面。

   前往「BigQuery」

2. 在查詢編輯器中執行下列陳述式：

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

   更改下列內容：

   • LOCATION：連線位置。
   • CONNECTION_ID：BigQuery 連線的 ID。

     在 Cloud de Confiance 控制台中查看連線詳細資料時，CONNECTION_ID 是「連線 ID」中顯示的完整連線 ID 最後一個區段的值，例如 projects/myproject/locations/connection_location/connections/myconnection。

   • ：包含 scf23.pdf 檔案的 Cloud Storage bucket。完整的 uri 選項值應類似 ['gs://mybucket/scf23.pdf']。BUCKET

建立文件處理器

建立文件處理器，以 us 多區域的版面配置剖析器處理器為基礎。從「處理器詳細資料」頁面複製預測端點，以供下一節使用。

為文件處理器建立遠端模型

建立遠端模型，存取 Document AI 處理器：

1. 前往 Cloud de Confiance 控制台的「BigQuery」頁面。

   前往「BigQuery」

2. 在查詢編輯器中執行下列陳述式：

   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');

   更改下列內容：

   • LOCATION：連線位置。
   • CONNECTION_ID：BigQuery 連線的 ID。

     在 Cloud de Confiance 控制台中查看連線詳細資料時，CONNECTION_ID 是「連線 ID」中顯示的完整連線 ID 最後一個區段的值，例如 projects/myproject/locations/connection_location/connections/myconnection。

   • PROCESSOR_ID：文件處理器 ID。如要找出這個值，請查看處理器詳細資料，然後查看「基本資訊」部分中的「ID」列。

將 PDF 檔案剖析為區塊

使用 ML.PROCESS_DOCUMENT 函式搭配文件處理器，將 PDF 檔案剖析為多個分塊，然後將內容寫入表格。ML.PROCESS_DOCUMENT 函式會以 JSON 格式傳回 PDF 區塊。

1. 前往 Cloud de Confiance 控制台的「BigQuery」頁面。

   前往「BigQuery」

2. 在查詢編輯器中執行下列陳述式：

   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}}}')
     )
   );

將 PDF 區塊資料剖析為不同的資料欄

從 ML.PROCESS_DOCUMENT 函式傳回的 JSON 資料中擷取 PDF 內容和中繼資料資訊，然後將該內容寫入表格：

1. 前往 Cloud de Confiance 控制台的「BigQuery」頁面。

   前往「BigQuery」

2. 在查詢編輯器中執行下列陳述式，剖析 PDF 內容：

   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. 在查詢編輯器中執行下列陳述式，即可查看已剖析 PDF 內容的子集：

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

   輸出結果會與下列內容相似：

   +-----------------------------------+------+------------------------------------------------------------------------------------------------------+-------------------+-----------------+---------------+
   |                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            |
   +-----------------------------------+------+------------------------------------------------------------------------------------------------------+-------------------+-----------------+---------------+
    

生成嵌入項目

為剖析的 PDF 內容生成嵌入項目，然後寫入資料表：

1. 前往 Cloud de Confiance 控制台的「BigQuery」頁面。

   前往「BigQuery」

2. 在查詢編輯器中執行下列陳述式：

   CREATE OR REPLACE TABLE `bqml_tutorial.embeddings` AS (
     SELECT *, AI.EMBED(content, endpoint => 'text-embedding-005').result AS embedding
     FROM bqml_tutorial.parsed_pdf
   );

執行向量搜尋

對剖析的 PDF 內容執行向量搜尋。

下列查詢會接收文字輸入內容、使用 AI.EMBED 函式為該輸入內容建立嵌入，然後使用 VECTOR_SEARCH 函式，將輸入嵌入與最相似的 PDF 內容嵌入進行比對。結果會顯示與家庭淨值變化最相關的十個 PDF 區塊。

1. 前往「BigQuery」頁面

   前往「BigQuery」

2. 在查詢編輯器中，執行下列 SQL 陳述式：

   SELECT
     distance,
     base.id AS chunk_id,
     base.page_span_start AS start_page,
     base.page_span_end AS end_page,
     base.content
   FROM
     VECTOR_SEARCH(
       TABLE `bqml_tutorial.embeddings`,
       'embedding',
       query_value =>
         AI.EMBED(
           'Did the typical family net worth increase? If so, by how much?',
           endpoint => 'text-embedding-005').result,
       top_k => 3,
       OPTIONS => '{"fraction_lists_to_search": 0.01}')
   ORDER BY distance DESC;

   輸出結果會與下列內容相似：

   +----------+----------+------------+----------+-----------------------------------+
   | distance | chunk_id | start_page | end_page | content                           |
   +----------+----------+------------+----------+-----------------------------------+
   | 0.645685 | 26       | 17         | 18       | 18 Between the first quarter of   |
   |          |          |            |          | 2019 and the first quarter of...  |
   +----------+----------+------------+----------+-----------------------------------+
   | 0.602665 | 30       | 19         | 21       | ## Net Worth by Family            |
   |          |          |            |          | Characteristics...                |
   +----------+----------+------------+----------+-----------------------------------+
   | 0.599438 | 24       | 17         | 21       | # Net Worth                       |
   |          |          |            |          | The net improvements in...        |
   +----------+----------+------------+----------+-----------------------------------+
   

根據向量搜尋結果生成文字

對嵌入執行向量搜尋，找出語意相似的 PDF 內容，然後搭配向量搜尋結果使用 AI.GENERATE 函式，擴增提示輸入內容並提升文字生成結果。在本例中，查詢會使用 PDF 區塊中的資訊，回答有關過去十年家庭淨值變化的問題。

1. 前往 Cloud de Confiance 控制台的「BigQuery」頁面。

   前往「BigQuery」

2. 在查詢編輯器中執行下列陳述式：

   SELECT
     AI.GENERATE(
       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", base.content), ',\n')
       ),
       endpoint => 'gemini-2.5-pro'
     ).result AS response
   FROM
     VECTOR_SEARCH(
       TABLE `bqml_tutorial.embeddings`,
       'embedding',
       query_value =>
         AI.EMBED(
           'Did the typical family net worth increase? If so, by how much?',
           endpoint => 'text-embedding-005').result,
       top_k => 3,
       OPTIONS => '{"fraction_lists_to_search": 0.01}')

   輸出結果會與下列內容相似：

   +-------------------------------------------------------------------------+
   | response                                                                |
   +-------------------------------------------------------------------------+
   | Yes, the typical family net worth changed significantly.                |
   |                                                                         |
   | Real median net worth surged 37% between the 2019 and 2022 SCF surveys. |
   | This contrasts sharply with a decade earlier (2010-2013), when real     |
   | median net worth decreased 2%.                                          |
   +-------------------------------------------------------------------------+