Artificial Intelligence Recommendation：Java SDK 參考

介面

RecallEngineClient client = new RecallEngineClient(endpoint, username, password);

參數說明

召回引擎服務的訪問地址可以在PAI-Rec管控台 召回管理基本資料頁面查看。

可選配置方法

說明

• 用戶端預設連線逾時為200ms，讀寫逾時為500ms，串連池最大串連數為1000。

• 所有可選配置方法支援鏈式調用。

樣本

import com.aliyun.openservices.pairec.recallengine.RecallEngineClient;

// 通過環境變數擷取認證資訊
String endpoint = System.getenv("RECALL_ENGINE_SERVICE_ENDPOINT");
String username = System.getenv("RECALL_ENGINE_SERVICE_USERNAME");
String password = System.getenv("RECALL_ENGINE_SERVICE_PASSWORD");

// 建立用戶端
RecallEngineClient client = new RecallEngineClient(endpoint, username, password);

// 設定重試次數
client.withRetryTimes(2);

介面

public RecallResponse recall(RecallRequest request) throws RecallEngineException;

RecallRequest參數說明

RecallConf參數說明

RecallResponse說明

樣本

import com.aliyun.openservices.pairec.recallengine.*;

import java.util.HashMap;
import java.util.Map;

// 構造召回請求
RecallRequest request = new RecallRequest();
request.setInstanceId("your-instance-id");
request.setService("recall_test");
request.setVersion("V1");
request.setUid("123");

// 設定召回配置
Map<String, RecallConf> recalls = new HashMap<>();
recalls.put("u2i_recall", new RecallConf("123", 100));
request.setRecalls(recalls);

// 執行召回
RecallResponse resp = client.recall(request);
Record result = resp.getResult();

// 擷取結果資訊
System.out.println("Total records: " + result.size());
System.out.println("Field names: " + result.fieldNames());

介面

public WriteResponse write(String instanceId, String table, WriteRequest request) throws RecallEngineException;

參數說明

WriteRequest參數說明

WriteResponse說明

WriteResponse 繼承自 Response，包含以下欄位：

樣本

import com.aliyun.openservices.pairec.recallengine.*;

import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;

// 構造寫入請求
WriteRequest request = new WriteRequest();
request.setRequestId("write-req-123");

// 構造寫入資料
Map<String, Object> item = new HashMap<>();
item.put("user_id", "123");
item.put("item_id", "item_123");
item.put("score", 0.95);

List<Map<String, Object>> content = new ArrayList<>();
content.add(item);
request.setContent(content);

// 執行寫入
WriteResponse resp = client.write("your-instance-id", "u2i_table", request);

// 檢查寫入結果
System.out.println("Code: " + resp.getCode());       // 成功時返回 "OK"
System.out.println("Message: " + resp.getMessage());

介面

public Record sort(String name, boolean desc);
public Record retain(int count);
public Record filter(String name);
public Record filterByColumnValue(String name, Predicate<Object> predicate);
public Record filterByValues(Predicate<Map<String, Object>> predicate);
public Record random();
public List<Object> columnValues(String columnName);
public List<String> columnValuesString(String columnName);
public List<String> fieldNames();
public int size();

方法說明

說明

• sort、retain、filter、filterByColumnValue、filterByValues、random 方法均返回 this，支援鏈式調用。

• Record 不是安全執行緒的，請勿在多線程環境下並行作業同一個 Record 執行個體。

樣本

RecallResponse resp = client.recall(request);
Record result = resp.getResult();

// 按 score 列降序排序，去重後取前10條
Record processed = result
        .sort("score", true)
        .filter("item_id")
        .retain(10);

System.out.println("Processed records: " + processed.size());

// 擷取 item_id 列的所有值
List<String> itemIds = processed.columnValuesString("item_id");
System.out.println("Item IDs: " + itemIds);

// 按條件過濾：僅保留 score > 0.5 的記錄
Record filtered = result.filterByColumnValue("score", value -> {
    if (value instanceof Number) {
        return ((Number) value).doubleValue() > 0.5;
    }
    return false;
});

// 按全行條件過濾
Record filtered2 = result.filterByValues(row -> {
    Object category = row.get("category");
    return category != null && "video".equals(category.toString());
});