CompressingQueryTransformer NullPointException

[zxyyang]

Visit CompressingQueryTransformer transform method to query. The metadata () when no memory context model of null pointer error

[zxyyang]

If you create the Assistant without specifying ChatMemoryProvider in the Build method but specifying retrievalAugmentor and then access it after the Assistant is built, you'll get a null pointer problem
（如果创建Assistant时候Build方法没有指定 ChatMemoryProvider 但是指定了retrievalAugmentor则在传入构建好后访问，会产出空指针问题）

[zxyyang]

@langchain4j

[langchain4j]

Hi @zxyyang, thank you, but what is the point of using CompressingQueryTransformer without ChatMemory?

[zxyyang]

Hi @zxyyang, thank you, but what is the point of using CompressingQueryTransformer without ChatMemory?

build AiServices use .retrievalAugmentor(retrievalAugmentor) but null .chatMemoryProvider(chatMemoryProvider) ，Use a embedding database but no memory context

[langchain4j]

@zxyyang I mean using CompressingQueryTransformer makes sense only when ChatMemory/chatMemoryProvider is present. Without the memory, there is nothing to compress...

[zxyyang]

CompressingQueryTransformer

My scenario is to use retrievalAugmentor but not set chatMemoryProvider automatically to call CompressingQueryTransformer and error ，

[langchain4j]

@zxyyang using CompressingQueryTransformer makes no sense without ChatMemory/chatMemoryProvider. What are you trying to achieve? Do you understand how CompressingQueryTransformer works? It takes all messages from chat memory (for this you need ChatMemory/chatMemoryProvider) and compresses them into a standalone query. If there is no memory, CompressingQueryTransformer will just waste time and tokens. It makes no sense.

[zxyyang]

@zxyyang using CompressingQueryTransformer makes no sense without ChatMemory/chatMemoryProvider. What are you trying to achieve? Do you understand how CompressingQueryTransformer works? It takes all messages from chat memory (for this you need ChatMemory/chatMemoryProvider) and compresses them into a standalone query. If there is no memory, CompressingQueryTransformer will just waste time and tokens. It makes no sense.

You mean that if I use embedding database I must have to configure the memory context store

[langchain4j]

You mean that if I use embedding database I must have to configure the memory context store

No, embedding store can be used without memory.
CompressingQueryTransformer is not needed to use embeding store/retreiver.
Hovewer, ChatMemory/chatMemoryProvider are needed to use CompressingQueryTransformer.

[zxyyang]

You mean that if I use embedding database I must have to configure the memory context store

No, embedding store can be used without memory. CompressingQueryTransformer is not needed to use embeding store/retreiver. Hovewer, ChatMemory/chatMemoryProvider are needed to use CompressingQueryTransformer.

    Assistant userMessageAssistant = AiServices.builder(Assistant.class)
      //       .chatMemoryProvider(chatMemoryProvider)
            .chatLanguageModel(OpenAiChatModel
                    .builder()
                    .logRequests(true)
                    .logResponses(true)
                    .modelName(apiConfig.getModel()) // 指定使用的模型名称
                    .apiKey(apiConfig.getKey()) // 设置API密钥
                    .maxTokens(apiConfig.getMaxToken()) // 最大令牌数
                    .maxRetries(apiConfig.getRetryTimes())
                    .baseUrl(StringUtils.isBlank(apiConfig.getBaseUrl()) ? null : apiConfig.getBaseUrl()) // 设置基础URL
                    .build())
            .tools(indicatorQueryTool)
            .retrievalAugmentor(retrievalAugmentor)
            .build();     This is my code If I use retrievalAugmentor will use CompressingQueryTransformer error he would himself, and because there is no context and error null pointer, If you had.chatMemoryProvider(chatMemoryProvider), it would work

[zxyyang]

You see is I use a problem or this code is not rigorous@ @langchain4j