Error using ElasticSearch with filter expression

[HellerZhang]

源代码：

 List<String> documentIds = (List<String>) requestBody.getOrDefault("documentIds", new ArrayList<>());
 Filter.Expression filterExpression = null;
filterExpression = b.in("id", documentIds).build();

经过debug调试后，发现elastic的报错信息是：
ErrorCause@163 "ErrorCause: {"phase":"dfs","failed_shards":[{"shard":0,"index":"my-index","node":"5gGemd0KQeafjhS6eo5CZw","reason":{"type":"query_shard_exception","reason":"Failed to parse query [(metadata.id:[15, 25])]","index_uuid":"70QktWW0T7OhYORdpL1Lsg","index":"my-index","caused_by":{"type":"parse_exception","reason":"Cannot parse '(metadata.id:[15, 25])': Encountered " <RANGE_GOOP> "25 "" at line 1, column 18.\r\nWas expecting:\r\n "TO" ...\r\n ","caused_by":{"type":"parse_exception","reason":"Encountered " <RANGE_GOOP> "25 "" at line 1, column 18.\r\nWas expecting:\r\n "TO" ...\r\n "}}}}],"grouped":true,"type":"search_phase_execution_exception","reason":"all shards failed","root_cause":[{"index_uuid":"70QktWW0T7OhYORdpL1Lsg","index":"my-index","type":"query_shard_exception","reason":"Failed to parse query [(metadata.id:[15, 25])]"}]}"

[markpollack]

Please try to translate to english in future.

"When you use Elasticsearch as a vector database, an error is reported when you use the in condition query"

There are tests for this here -

spring-ai/vector-stores/spring-ai-elasticsearch-store/src/test/java/org/springframework/ai/vectorstore/elasticsearch/ElasticsearchVectorStoreIT.java

Line 198 in 4f4da30

public void searchWithFilters(String vectorStoreBeanName) {

perhaps that helps.

Please provide a fully working repo example and we can investigate further.

[HellerZhang]

I suspect there is an issue with the string format in the expression. When I use .filterExpression("id in ['25','26']"), there is no problem. However, I noticed that in the result generated by b.in("id", documentIds).build();, the metadata.id:[15, 25] does not have quotation marks. The data stored in my vector database is in string format. Could this be related to the problem?