建立受管知識庫

以下是使用 API 搭配 AWS CLI 或支援的 SDK 建立受管知識庫和設定資料來源的範例，例如 Python。呼叫 CreateKnowledgeBase 之後，您可以呼叫 CreateDataSource，在 dataSourceConfiguration 中使用連線資訊建立資料來源。

若要了解您可以透過包含選用 vectorIngestionConfiguration 欄位套用至擷取的自訂項目，請參閱自訂資料來源的擷取。

AWS Command Line Interface

步驟 1：建立知識庫

使用受管內嵌模型 （預設）：

aws bedrock-agent create-knowledge-base \
 --name "my-managed-kb" \
 --role-arn "arn:aws:iam::123456789012:role/BedrockKBRole" \
 --description "My managed knowledge base" \
 --knowledge-base-configuration file://kb-config.json

kb-config.json
{
    "type": "MANAGED",
    "managedKnowledgeBaseConfiguration": {
        "embeddingModelType": "MANAGED"
    }
}

使用自訂內嵌模型 （客戶提供的 Bedrock 模型）：

aws bedrock-agent create-knowledge-base \
 --name "my-custom-embed-kb" \
 --role-arn "arn:aws:iam::123456789012:role/BedrockKBRole" \
 --description "My managed knowledge base with custom embedding" \
 --knowledge-base-configuration file://kb-config.json

kb-config.json
{
    "type": "MANAGED",
    "managedKnowledgeBaseConfiguration": {
        "embeddingModelType": "CUSTOM",
        "embeddingModelArn": "arn:aws:bedrock:us-west-2::foundation-model/amazon.titan-embed-text-v2:0",
        "embeddingModelConfiguration": {
            "bedrockEmbeddingModelConfiguration": {
                "dimensions": 1024
            }
        }
    }
}

使用自訂多模式內嵌模型 (TwelveLabs Marengo Embed 3.0)：

多模態內嵌模型使用CUSTOMembeddingModelType與文字內嵌模型相同的 。差別在於 bedrockEmbeddingModelConfiguration接受您傳遞模型特定設定的額外modelConfiguration欄位，而且您必須supplementalDataStorageConfiguration為多模型儲存目的地指定 。

aws bedrock-agent create-knowledge-base \
 --name "my-multimodal-embed-kb" \
 --role-arn "arn:aws:iam::123456789012:role/BedrockKBRole" \
 --description "My managed knowledge base with multimodal embedding" \
 --knowledge-base-configuration file://kb-config.json

kb-config.json
{
    "type": "MANAGED",
    "managedKnowledgeBaseConfiguration": {
        "embeddingModelType": "CUSTOM",
        "embeddingModelArn": "arn:aws:bedrock:us-east-1::foundation-model/twelvelabs.marengo-embed-3-0-v1:0",
        "embeddingModelConfiguration": {
            "bedrockEmbeddingModelConfiguration": {
                "embeddingDataType": "FLOAT",
                "modelConfiguration": {
                    "version": "1",
                    "audio": {
                        "segmentation": {
                            "method": "dynamic",
                            "dynamic": {
                                "minDurationSec": 4
                            }
                        }
                    },
                    "video": {
                        "segmentation": {
                            "method": "fixed",
                            "fixed": {
                                "durationSec": 6
                            }
                        }
                    }
                }
            }
        },
        "supplementalDataStorageConfiguration": {
            "storageLocations": [
                {
                    "s3Location": {
                        "uri": "s3://amzn-s3-demo-bucket"
                    },
                    "type": "S3"
                }
            ]
        }
    }
}

modelConfiguration 欄位是 JSON 物件，其內容專屬於您選擇的內嵌模型。version 欄位為必要欄位，且必須設定為 1。對於每個 video audio和 模態，您可以設定 segmentationmethod的 dynamic，其接受minDurationSec值，或 fixed，其接受durationSec值。如需 TwelveLabs Marengo Embed 3.0 接受的設定，請參閱 TwelveLabs Marengo Embed 3.0。

步驟 2：建立資料來源

aws bedrock-agent create-data-source \
 --name "S3-connector" \
 --description "S3 data source connector for Amazon Bedrock to use content in S3" \
 --knowledge-base-id "your-knowledge-base-id" \
 --data-source-configuration file://bedrock-s3-managed-connector-configuration.json \
 --data-deletion-policy "DELETE" \
 --vector-ingestion-configuration '{"parsingConfiguration":{"parsingStrategy":"SMART_PARSING"}}'

bedrock-s3-managed-connector-configuration.json
{
    "type": "MANAGED_KNOWLEDGE_BASE_CONNECTOR",
    "managedKnowledgeBaseConnectorConfiguration": {
        "mediaExtractionConfiguration": {
            "imageExtractionConfiguration": {
                "imageExtractionStatus": "ENABLED"
            }
        },
        "connectorParameters": {
            "type": "S3",
            "version": "1",
            "connectionConfiguration": {
                "bucketName": "your-test-s3-bucket",
                "bucketOwnerAccountId": "123456789012"
            },
            "deletionProtectionConfiguration": {
                "enableDeletionProtection": false
            }
        }
    }
}