在 2024 年演示日活动中，观看有关如何使用 Firebase 构建和运行 AI 赋能应用的演示。立即观看。

本頁面由 Cloud Translation API 翻譯而成。

使用向量嵌入項目搜尋

本頁面說明如何使用 Cloud Firestore 執行 K 近鄰 (KNN) 向量搜尋，並使用下列技巧：

儲存向量值
建立及管理 KNN 向量索引
使用支援的其中一個向量距離度量方式，建立 K 近鄰 (KNN) 查詢

儲存向量嵌入

您可以從 Cloud Firestore 資料建立向量值，例如文字嵌入，並將這些值儲存在 Cloud Firestore 文件中。

使用向量嵌入功能的寫入作業

以下範例說明如何在 Cloud Firestore 文件中儲存向量嵌入：

Python

from google.cloud import firestore
from google.cloud.firestore_v1.vector import Vector

firestore_client = firestore.Client()
collection = firestore_client.collection("coffee-beans")
doc = {
    "name": "Kahawa coffee beans",
    "description": "Information about the Kahawa coffee beans.",
    "embedding_field": Vector([0.18332680, 0.24160706, 0.3416704]),
}

collection.add(doc)vector_search.py

Node.js

import {
  Firestore,
  FieldValue,
} from "@google-cloud/firestore";

const db = new Firestore();
const coll = db.collection('coffee-beans');
await coll.add({
  name: "Kahawa coffee beans",
  description: "Information about the Kahawa coffee beans.",
  embedding_field: FieldValue.vector([1.0 , 2.0, 3.0])
});

Go

import (
	"context"
	"fmt"
	"io"

	"cloud.google.com/go/firestore"
)

type CoffeeBean struct {
	Name           string             `firestore:"name,omitempty"`
	Description    string             `firestore:"description,omitempty"`
	EmbeddingField firestore.Vector32 `firestore:"embedding_field,omitempty"`
	Color          string             `firestore:"color,omitempty"`
}

func storeVectors(w io.Writer, projectID string) error {
	ctx := context.Background()

	// Create client
	client, err := firestore.NewClient(ctx, projectID)
	if err != nil {
		return fmt.Errorf("firestore.NewClient: %w", err)
	}
	defer client.Close()

	// Vector can be represented by Vector32 or Vector64
	doc := CoffeeBean{
		Name:           "Kahawa coffee beans",
		Description:    "Information about the Kahawa coffee beans.",
		EmbeddingField: []float32{1.0, 2.0, 3.0},
		Color:          "red",
	}
	ref := client.Collection("coffee-beans").NewDoc()
	if _, err = ref.Set(ctx, doc); err != nil {
		fmt.Fprintf(w, "failed to upsert: %v", err)
		return err
	}

	return nil
}
vector_store.go

Java

import com.google.cloud.firestore.CollectionReference;
import com.google.cloud.firestore.DocumentReference;
import com.google.cloud.firestore.FieldValue;
import com.google.cloud.firestore.VectorQuery;

CollectionReference coll = firestore.collection("coffee-beans");

Map<String, Object> docData = new HashMap<>();
docData.put("name", "Kahawa coffee beans");
docData.put("description", "Information about the Kahawa coffee beans.");
docData.put("embedding_field", FieldValue.vector(new double[] {1.0, 2.0, 3.0}));

ApiFuture<DocumentReference> future = coll.add(docData);
DocumentReference documentReference = future.get();

使用 Cloud 函式計算向量嵌入

如要在建立或更新文件時計算及儲存向量嵌入資料，您可以設定 Cloud 函式：

Python

@functions_framework.cloud_event
def store_embedding(cloud_event) -> None:
  """Triggers by a change to a Firestore document.
  """
  firestore_payload = firestore.DocumentEventData()
  payload = firestore_payload._pb.ParseFromString(cloud_event.data)

  collection_id, doc_id = from_payload(payload)
  # Call a function to calculate the embedding
  embedding = calculate_embedding(payload)
  # Update the document
  doc = firestore_client.collection(collection_id).document(doc_id)
  doc.set({"embedding_field": embedding}, merge=True)

Node.js

/**
 * A vector embedding will be computed from the
 * value of the `content` field. The vector value
 * will be stored in the `embedding` field. The
 * field names `content` and `embedding` are arbitrary
 * field names chosen for this example.
 */
async function storeEmbedding(event: FirestoreEvent<any>): Promise<void> {
  // Get the previous value of the document's `content` field.
  const previousDocumentSnapshot = event.data.before as QueryDocumentSnapshot;
  const previousContent = previousDocumentSnapshot.get("content");

  // Get the current value of the document's `content` field.
  const currentDocumentSnapshot = event.data.after as QueryDocumentSnapshot;
  const currentContent = currentDocumentSnapshot.get("content");

  // Don't update the embedding if the content field did not change
  if (previousContent === currentContent) {
    return;
  }

  // Call a function to calculate the embedding for the value
  // of the `content` field.
  const embeddingVector = calculateEmbedding(currentContent);

  // Update the `embedding` field on the document.
  await currentDocumentSnapshot.ref.update({
    embedding: embeddingVector,
  });
}

Go

  // Not yet supported in the Go client library

Java

  // Not yet supported in the Java client library

建立及管理向量索引

您必須先建立對應的索引，才能使用向量嵌入資料執行最近鄰搜尋。以下範例說明如何使用 Google Cloud CLI 建立及管理向量索引。您也可以使用 Firebase CLI 和 Terraform 管理向量索引。

建立向量索引

建立向量索引之前，請先升級至 Google Cloud CLI 的最新版本：

gcloud components update

如要建立向量索引，請使用 gcloud firestore indexes composite create：

gcloud

gcloud firestore indexes composite create \
--collection-group=collection-group \
--query-scope=COLLECTION \
--field-config field-path=vector-field,vector-config='vector-configuration' \
--database=database-id

其中：

collection-group 是集合群組的 ID。
vector-field 是包含向量嵌入的欄位名稱。
database-id 是資料庫的 ID。
vector-configuration 包含向量 dimension 和索引類型。dimension 是整數，最多可達 2048。索引類型必須是 flat。設定索引格式如下：{"dimension":"DIMENSION", "flat": "{}"}。

以下範例會建立複合式索引，包括欄位 vector-field 的向量索引，以及欄位 color 的遞增索引。您可以使用這類索引，在最鄰近搜尋之前預先篩選資料。

gcloud

gcloud firestore indexes composite create \
--collection-group=collection-group \
--query-scope=COLLECTION \
--field-config=order=ASCENDING,field-path="color" \
--field-config field-path=vector-field,vector-config='{"dimension":"1024", "flat": "{}"}' \
--database=database-id

列出所有向量索引

gcloud

gcloud firestore indexes composite list --database=database-id

將 database-id 替換為資料庫的 ID。

刪除向量索引

gcloud

gcloud firestore indexes composite delete index-id --database=database-id

其中：

index-id 是要刪除的索引 ID。使用 indexes composite list 擷取索引 ID。
database-id 是資料庫的 ID。

說明向量索引

gcloud

gcloud firestore indexes composite describe index-id --database=database-id

其中：

index-id 是所要描述索引的 ID。使用 indexes composite list 擷取索引 ID。
database-id 是資料庫的 ID。

提出最鄰近查詢

您可以執行相似度搜尋，找出向量嵌入項目的最鄰近項目。相似度搜尋需要向量索引。如果不存在索引，Cloud Firestore 會建議使用 gcloud CLI 建立索引。

以下範例會找出查詢向量的 10 個最近鄰。

Python

from google.cloud.firestore_v1.base_vector_query import DistanceMeasure
from google.cloud.firestore_v1.vector import Vector

collection = db.collection("coffee-beans")

# Requires a single-field vector index
vector_query = collection.find_nearest(
    vector_field="embedding_field",
    query_vector=Vector([0.3416704, 0.18332680, 0.24160706]),
    distance_measure=DistanceMeasure.EUCLIDEAN,
    limit=5,
)vector_search.py

Node.js

import {
  Firestore,
  FieldValue,
  VectorQuery,
  VectorQuerySnapshot,
} from "@google-cloud/firestore";

// Requires a single-field vector index
const vectorQuery: VectorQuery = coll.findNearest({
  vectorField: 'embedding_field',
  queryVector: [3.0, 1.0, 2.0],
  limit: 10,
  distanceMeasure: 'EUCLIDEAN'
});

const vectorQuerySnapshot: VectorQuerySnapshot = await vectorQuery.get();

Go

import (
	"context"
	"fmt"
	"io"

	"cloud.google.com/go/firestore"
)

func vectorSearchBasic(w io.Writer, projectID string) error {
	ctx := context.Background()

	// Create client
	client, err := firestore.NewClient(ctx, projectID)
	if err != nil {
		return fmt.Errorf("firestore.NewClient: %w", err)
	}
	defer client.Close()

	collection := client.Collection("coffee-beans")

	// Requires a vector index
	// https://firebase.google.com/docs/firestore/vector-search#create_and_manage_vector_indexes
	vectorQuery := collection.FindNearest("embedding_field",
		[]float32{3.0, 1.0, 2.0},
		5,
		// More info: https://firebase.google.com/docs/firestore/vector-search#vector_distances
		firestore.DistanceMeasureEuclidean,
		nil)

	docs, err := vectorQuery.Documents(ctx).GetAll()
	if err != nil {
		fmt.Fprintf(w, "failed to get vector query results: %v", err)
		return err
	}

	for _, doc := range docs {
		fmt.Fprintln(w, doc.Data()["name"])
	}
	return nil
}
vector_search_basic.go

Java

import com.google.cloud.firestore.VectorQuery;
import com.google.cloud.firestore.VectorQuerySnapshot;

VectorQuery vectorQuery = coll.findNearest(
        "embedding_field",
        new double[] {3.0, 1.0, 2.0},
        /* limit */ 10,
        VectorQuery.DistanceMeasure.EUCLIDEAN);

ApiFuture<VectorQuerySnapshot> future = vectorQuery.get();
VectorQuerySnapshot vectorQuerySnapshot = future.get();

向量距離

最鄰近查詢支援下列向量距離選項：

EUCLIDEAN：測量向量之間的歐氏距離。詳情請參閱 Euclidean。
COSINE：根據向量之間的夾角比較向量，讓您測量不以向量大小為依據的相似度。建議您使用 DOT_PRODUCT 搭配單位正規化向量，而不要使用餘弦距離，因為在數學上，這兩者是等價的，且 DOT_PRODUCT 的效能更佳。如要進一步瞭解，請參閱「餘弦相似度」一文。
DOT_PRODUCT：類似 COSINE，但會受到向量大小的影響。詳情請參閱「內積」。

選擇距離測量單位

視所有向量嵌入項目是否已正規化而定，您可以決定要使用哪種距離測量方法來找出距離測量值。經過規格化的向量嵌入值大小 (長度) 會精確為 1.0。

此外，如果您知道模型是使用哪種距離測量值進行訓練，請使用該距離測量值來計算向量嵌入值之間的距離。

規一化資料

如果資料集中的所有向量嵌入值都已標準化，則所有三種距離評估方法都會提供相同的語意搜尋結果。從本質上來說，雖然每個距離測量值都會傳回不同的值，但這些值的排序方式相同。當嵌入值經過標準化後，DOT_PRODUCT 通常是最具運算效率的做法，但在大多數情況下，差異不大。不過，如果您的應用程式對效能極為敏感，DOT_PRODUCT 可能有助於調整效能。

非標準化資料

如果向量嵌入未經過標準化，則使用 DOT_PRODUCT 做為距離測量值，在數學上並不正確，因為點積乘積無法測量距離。視嵌入資料的產生方式和偏好的搜尋類型而定，COSINE 或 EUCLIDEAN 距離評估值產生的搜尋結果，在主觀上會比其他距離評估值更優。您可能需要實驗 COSINE 或 EUCLIDEAN，才能判斷哪一個最適合您的用途。

不確定資料是否已標準化

如果您不確定資料是否已正規化，但仍想使用 DOT_PRODUCT，建議改用 COSINE。COSINE 就像內建規範化的 DOT_PRODUCT。使用 COSINE 測量的距離範圍從 0 到 2。如果結果接近 0，表示向量非常相似。

預先篩選文件

如要在尋找最近鄰點之前預先篩選文件，可以將相似度搜尋與其他查詢運算子結合。支援 and 和 or 複合篩選器。如要進一步瞭解支援的欄位篩選器，請參閱「查詢運算子」。

Python

from google.cloud.firestore_v1.base_vector_query import DistanceMeasure
from google.cloud.firestore_v1.vector import Vector

collection = db.collection("coffee-beans")

# Similarity search with pre-filter
# Requires a composite vector index
vector_query = collection.where("color", "==", "red").find_nearest(
    vector_field="embedding_field",
    query_vector=Vector([0.3416704, 0.18332680, 0.24160706]),
    distance_measure=DistanceMeasure.EUCLIDEAN,
    limit=5,
)vector_search.py

Node.js

// Similarity search with pre-filter
// Requires composite vector index
const preFilteredVectorQuery: VectorQuery = coll
    .where("color", "==", "red")
    .findNearest({
      vectorField: "embedding_field",
      queryVector: [3.0, 1.0, 2.0],
      limit: 5,
      distanceMeasure: "EUCLIDEAN",
    });

const vectorQueryResults = await preFilteredVectorQuery.get();

Go

import (
	"context"
	"fmt"
	"io"

	"cloud.google.com/go/firestore"
)

func vectorSearchPrefilter(w io.Writer, projectID string) error {
	ctx := context.Background()

	// Create client
	client, err := firestore.NewClient(ctx, projectID)
	if err != nil {
		return fmt.Errorf("firestore.NewClient: %w", err)
	}
	defer client.Close()

	collection := client.Collection("coffee-beans")

	// Similarity search with pre-filter
	// Requires a composite vector index
	vectorQuery := collection.Where("color", "==", "red").
		FindNearest("embedding_field",
			[]float32{3.0, 1.0, 2.0},
			5,
			// More info: https://firebase.google.com/docs/firestore/vector-search#vector_distances
			firestore.DistanceMeasureEuclidean,
			nil)

	docs, err := vectorQuery.Documents(ctx).GetAll()
	if err != nil {
		fmt.Fprintf(w, "failed to get vector query results: %v", err)
		return err
	}

	for _, doc := range docs {
		fmt.Fprintln(w, doc.Data()["name"])
	}
	return nil
}
vector_search_prefilter.go

Java

import com.google.cloud.firestore.VectorQuery;
import com.google.cloud.firestore.VectorQuerySnapshot;

VectorQuery preFilteredVectorQuery = coll
        .whereEqualTo("color", "red")
        .findNearest(
                "embedding_field",
                new double[] {3.0, 1.0, 2.0},
                /* limit */ 10,
                VectorQuery.DistanceMeasure.EUCLIDEAN);

ApiFuture<VectorQuerySnapshot> future = preFilteredVectorQuery.get();
VectorQuerySnapshot vectorQuerySnapshot = future.get();

擷取計算的向量距離

您可以在 FindNearest 查詢中指派 distance_result_field 輸出屬性名稱，藉此擷取計算的向量距離，如以下範例所示：

Python

from google.cloud.firestore_v1.base_vector_query import DistanceMeasure
from google.cloud.firestore_v1.vector import Vector

collection = db.collection("coffee-beans")

vector_query = collection.find_nearest(
    vector_field="embedding_field",
    query_vector=Vector([0.3416704, 0.18332680, 0.24160706]),
    distance_measure=DistanceMeasure.EUCLIDEAN,
    limit=10,
    distance_result_field="vector_distance",
)

docs = vector_query.stream()

for doc in docs:
    print(f"{doc.id}, Distance: {doc.get('vector_distance')}")vector_search.py

Node.js

const vectorQuery: VectorQuery = coll.findNearest(
    {
      vectorField: 'embedding_field',
      queryVector: [3.0, 1.0, 2.0],
      limit: 10,
      distanceMeasure: 'EUCLIDEAN',
      distanceResultField: 'vector_distance'
    });

const snapshot: VectorQuerySnapshot = await vectorQuery.get();

snapshot.forEach((doc) => {
  console.log(doc.id, ' Distance: ', doc.get('vector_distance'));
});

Go

import (
	"context"
	"fmt"
	"io"

	"cloud.google.com/go/firestore"
)

func vectorSearchDistanceResultField(w io.Writer, projectID string) error {
	ctx := context.Background()

	client, err := firestore.NewClient(ctx, projectID)
	if err != nil {
		return fmt.Errorf("firestore.NewClient: %w", err)
	}
	defer client.Close()

	collection := client.Collection("coffee-beans")

	// Requires a vector index
	// https://firebase.google.com/docs/firestore/vector-search#create_and_manage_vector_indexes
	vectorQuery := collection.FindNearest("embedding_field",
		[]float32{3.0, 1.0, 2.0},
		10,
		firestore.DistanceMeasureEuclidean,
		&firestore.FindNearestOptions{
			DistanceResultField: "vector_distance",
		})

	docs, err := vectorQuery.Documents(ctx).GetAll()
	if err != nil {
		fmt.Fprintf(w, "failed to get vector query results: %v", err)
		return err
	}

	for _, doc := range docs {
		fmt.Fprintf(w, "%v, Distance: %v\n", doc.Data()["name"], doc.Data()["vector_distance"])
	}
	return nil
}
vector_search_result_field.go

Java

import com.google.cloud.firestore.VectorQuery;
import com.google.cloud.firestore.VectorQueryOptions;
import com.google.cloud.firestore.VectorQuerySnapshot;

VectorQuery vectorQuery = coll.findNearest(
        "embedding_field",
        new double[] {3.0, 1.0, 2.0},
        /* limit */ 10,
        VectorQuery.DistanceMeasure.EUCLIDEAN,
        VectorQueryOptions.newBuilder().setDistanceResultField("vector_distance").build());

ApiFuture<VectorQuerySnapshot> future = vectorQuery.get();
VectorQuerySnapshot vectorQuerySnapshot = future.get();

for (DocumentSnapshot document : vectorQuerySnapshot.getDocuments()) {
    System.out.println(document.getId() + " Distance: " + document.get("vector_distance"));
}

如果您想使用欄位遮罩來傳回部分文件欄位和 distanceResultField，則必須在欄位遮罩中加入 distanceResultField 的值，如以下範例所示：

Python

vector_query = collection.select(["color", "vector_distance"]).find_nearest(
    vector_field="embedding_field",
    query_vector=Vector([0.3416704, 0.18332680, 0.24160706]),
    distance_measure=DistanceMeasure.EUCLIDEAN,
    limit=10,
    distance_result_field="vector_distance",
)vector_search.py

Node.js

const vectorQuery: VectorQuery = coll
    .select('name', 'description', 'vector_distance')
    .findNearest({
      vectorField: 'embedding_field',
      queryVector: [3.0, 1.0, 2.0],
      limit: 10,
      distanceMeasure: 'EUCLIDEAN',
      distanceResultField: 'vector_distance'
    });

Go

import (
	"context"
	"fmt"
	"io"

	"cloud.google.com/go/firestore"
)

func vectorSearchDistanceResultFieldMasked(w io.Writer, projectID string) error {
	ctx := context.Background()

	client, err := firestore.NewClient(ctx, projectID)
	if err != nil {
		return fmt.Errorf("firestore.NewClient: %w", err)
	}
	defer client.Close()

	collection := client.Collection("coffee-beans")

	// Requires a vector index
	// https://firebase.google.com/docs/firestore/vector-search#create_and_manage_vector_indexes
	vectorQuery := collection.Select("color", "vector_distance").
		FindNearest("embedding_field",
			[]float32{3.0, 1.0, 2.0},
			10,
			firestore.DistanceMeasureEuclidean,
			&firestore.FindNearestOptions{
				DistanceResultField: "vector_distance",
			})

	docs, err := vectorQuery.Documents(ctx).GetAll()
	if err != nil {
		fmt.Fprintf(w, "failed to get vector query results: %v", err)
		return err
	}

	for _, doc := range docs {
		fmt.Fprintf(w, "%v, Distance: %v\n", doc.Data()["color"], doc.Data()["vector_distance"])
	}
	return nil
}
vector_search_result_field_masked.go

Java

import com.google.cloud.firestore.VectorQuery;
import com.google.cloud.firestore.VectorQueryOptions;
import com.google.cloud.firestore.VectorQuerySnapshot;

VectorQuery vectorQuery = coll
        .select("name", "description", "vector_distance")
        .findNearest(
          "embedding_field",
          new double[] {3.0, 1.0, 2.0},
          /* limit */ 10,
          VectorQuery.DistanceMeasure.EUCLIDEAN,
          VectorQueryOptions.newBuilder()
            .setDistanceResultField("vector_distance")
            .build());

ApiFuture<VectorQuerySnapshot> future = vectorQuery.get();
VectorQuerySnapshot vectorQuerySnapshot = future.get();

for (DocumentSnapshot document : vectorQuerySnapshot.getDocuments()) {
    System.out.println(document.getId() + " Distance: " + document.get("vector_distance"));
}

指定距離門檻

您可以指定相似度門檻，只傳回門檻內的文件。門檻欄位的行為取決於您選擇的距離測量單位：

EUCLIDEAN 和 COSINE 距離會將閾值限制在距離小於或等於指定閾值的文件。這些距離測量值會隨著向量相似度降低。
DOT_PRODUCT distance 會將閾值限制在距離大於或等於指定閾值的文件。點積距離會隨著向量相似度的增加而增加。

以下範例說明如何指定距離門檻，以便使用 EUCLIDEAN 距離指標，傳回最多 10 個距離為 4.5 單位以內的最近文件：

Python

from google.cloud.firestore_v1.base_vector_query import DistanceMeasure
from google.cloud.firestore_v1.vector import Vector

collection = db.collection("coffee-beans")

vector_query = collection.find_nearest(
    vector_field="embedding_field",
    query_vector=Vector([0.3416704, 0.18332680, 0.24160706]),
    distance_measure=DistanceMeasure.EUCLIDEAN,
    limit=10,
    distance_threshold=4.5,
)

docs = vector_query.stream()

for doc in docs:
    print(f"{doc.id}")vector_search.py

Node.js

const vectorQuery: VectorQuery = coll.findNearest({
  vectorField: 'embedding_field',
  queryVector: [3.0, 1.0, 2.0],
  limit: 10,
  distanceMeasure: 'EUCLIDEAN',
  distanceThreshold: 4.5
});

const snapshot: VectorQuerySnapshot = await vectorQuery.get();

snapshot.forEach((doc) => {
  console.log(doc.id);
});

Go

import (
	"context"
	"fmt"
	"io"

	"cloud.google.com/go/firestore"
)

func vectorSearchDistanceThreshold(w io.Writer, projectID string) error {
	ctx := context.Background()

	client, err := firestore.NewClient(ctx, projectID)
	if err != nil {
		return fmt.Errorf("firestore.NewClient: %w", err)
	}
	defer client.Close()

	collection := client.Collection("coffee-beans")

	// Requires a vector index
	// https://firebase.google.com/docs/firestore/vector-search#create_and_manage_vector_indexes
	vectorQuery := collection.FindNearest("embedding_field",
		[]float32{3.0, 1.0, 2.0},
		10,
		firestore.DistanceMeasureEuclidean,
		&firestore.FindNearestOptions{
			DistanceThreshold: firestore.Ptr[float64](4.5),
		})

	docs, err := vectorQuery.Documents(ctx).GetAll()
	if err != nil {
		fmt.Fprintf(w, "failed to get vector query results: %v", err)
		return err
	}

	for _, doc := range docs {
		fmt.Fprintln(w, doc.Data()["name"])
	}
	return nil
}
vector_search_distance_threshold.go

Java

import com.google.cloud.firestore.VectorQuery;
import com.google.cloud.firestore.VectorQueryOptions;
import com.google.cloud.firestore.VectorQuerySnapshot;

VectorQuery vectorQuery = coll.findNearest(
        "embedding_field",
        new double[] {3.0, 1.0, 2.0},
        /* limit */ 10,
        VectorQuery.DistanceMeasure.EUCLIDEAN,
        VectorQueryOptions.newBuilder()
          .setDistanceThreshold(4.5)
          .build());

ApiFuture<VectorQuerySnapshot> future = vectorQuery.get();
VectorQuerySnapshot vectorQuerySnapshot = future.get();

for (DocumentSnapshot document : vectorQuerySnapshot.getDocuments()) {
    System.out.println(document.getId());
}

限制

使用向量嵌入功能時，請注意下列限制：

支援的嵌入維度上限為 2048。如要儲存較大的索引，請使用降維。
最接近查詢傳回的文件數量上限為 1000 個。
向量搜尋不支援即時快照事件監聽器。
只有 Python、Node.js、Go 和 Java 用戶端程式庫支援向量搜尋。

後續步驟

請參閱 Cloud Firestore 的最佳做法。
瞭解大量讀取和寫入。