Elasticsearch CosineSimilarity类型转换异常求助

在Elasticsearch中使用CosineSimilarity对name_vector2和description_vector2两个向量字段进行相似度检索时，遇到类型转换错误。API传入的$query_vector参数有效，且能正常索引带有向量的文档。

请求语句 #

{
    "size":10,
    "query": {
        "script_score": {
            "query": {"match_all": {}},
            "script": {
                "source": "cosineSimilarity(params.query_vector, doc['name_vector2']) + cosineSimilarity(params.query_vector, doc['description_vector2']) + 2.0",
                "params": {
                    "query_vector": $query_vector
                }
            }
        }
    },
    "_source": {"includes": ["name", "description"]}
}

错误信息 #

{
    "error": {
        "root_cause": [{
            "type": "script_exception",
            "reason": "runtime error",
            "script_stack": ["cosineSimilarity(params.query_vector, doc['name_vector2']) + cosineSimilarity(params.query_vector, doc['description_vector2']) + 2.0", " ^---- HERE"],
            "script": "cosineSimilarity(params.query_vector, doc['name_vector2']) + cosineSimilarity(params.query_vector, doc['description_vector2']) + 2.0",
            "lang": "painless",
            "position": {
                "offset": 41,
                "start": 0,
                "end": 132
            }
        }],
        "type": "search_phase_execution_exception",
        "reason": "all shards failed",
        "phase": "query",
        "grouped": true,
        "failed_shards": [{
            "shard": 0,
            "index": "demandes_internes",
            "node": "4_1L4y7rTd-VrfOktvr-eQ",
            "reason": {
                "type": "script_exception",
                "reason": "runtime error",
                "script_stack": ["cosineSimilarity(params.query_vector, doc['name_vector2']) + cosineSimilarity(params.query_vector, doc['description_vector2']) + 2.0", " ^---- HERE"],
                "script": "cosineSimilarity(params.query_vector, doc['name_vector2']) + cosineSimilarity(params.query_vector, doc['description_vector2']) + 2.0",
                "lang": "painless",
                "position": {
                    "offset": 41,
                    "start": 0,
                    "end": 132
                },
                "caused_by": {
                    "type": "class_cast_exception",
                    "reason": "cannot implicitly cast def [org.elasticsearch.index.mapper.vectors.DenseVectorScriptDocValues] to java.lang.String"
                }
            }
        }]
    },
    "status": 400
}

集群健康状态 #

{
  "cluster_name" : "ent",
  "status" : "yellow",
  "timed_out" : false,
  "number_of_nodes" : 1,
  "number_of_data_nodes" : 1,
  "active_primary_shards" : 8,
  "active_shards" : 8,
  "relocating_shards" : 0,
  "initializing_shards" : 0,
  "unassigned_shards" : 1,
  "delayed_unassigned_shards" : 0,
  "number_of_pending_tasks" : 0,
  "number_of_in_flight_fetch" : 0,
  "task_max_waiting_in_queue_millis" : 0,
  "active_shards_percent_as_number" : 88.88888888888889
}

原因分析 #

错误核心是class_cast_exception，提示无法将DenseVectorScriptDocValues类型转为String。这是因为在Painless脚本中直接使用doc['name_vector2']获取向量字段时，返回的是向量文档值对象，而非cosineSimilarity函数所需的原始向量数组。

解决方案 #

修改脚本中向量字段的获取方式，使用.vector属性提取原始向量数组，同时添加字段存在性判断避免空指针错误：

修改后的请求语句 #

{
    "size":10,
    "query": {
        "script_score": {
            "query": {"match_all": {}},
            "script": {
                "source": """
                    double score = 2.0;
                    if (doc.containsKey('name_vector2')) {
                        score += cosineSimilarity(params.query_vector, doc['name_vector2'].vector);
                    }
                    if (doc.containsKey('description_vector2')) {
                        score += cosineSimilarity(params.query_vector, doc['description_vector2'].vector);
                    }
                    return score;
                """,
                "params": {
                    "query_vector": $query_vector
                }
            }
        }
    },
    "_source": {"includes": ["name", "description"]}
}

内容的提问来源于stack exchange，提问作者Xbattlax