How can I retrieve the context documents from RetrievalAugmentationAdvisor in a RAG pipeline?

[rkudryashov]

I'm using Spring AI with Qdrant to implement a RAG (Retrieval-Augmented Generation) pipeline, and I've integrated RetrievalAugmentationAdvisor (or QuestionAnswerAdvisor) to automate the retrieval and generation flow.

While the advisor successfully generates accurate answers, I need to return the retrieved context documents (e.g., code snippets, text chunks) alongside the answer to show evidence or source references in the UI.

However, it appears that RetrievalAugmentationAdvisor only returns the final answer string and does not expose the list of retrieved Document objects used in the context.

My Questions:

1. Is there a built-in way to access the retrieved documents from RetrievalAugmentationAdvisor after retrieval?
2. If not, what is the recommended approach to capture or intercept the retrieved chunks while still benefiting from Spring AI’s RAG abstractions?
3. Should I instead bypass the advisor and manually orchestrate retrieval (VectorStore.similaritySearch) + generation (ChatClient) for more control?

Environment:

• Spring AI Version: 1.0.1
• Vector Store: Qdrant (QdrantVectorStore)

Any guidance or examples (e.g., custom advisor, decorated vector store, etc.) would be greatly appreciated!

Thanks!

[rkudryashov]

done:

        return chatClient.prompt()
            .advisors(retrievalAugmentationAdvisor)
            .user(question)
            .stream()
            .chatResponse()
            .flatMap { chatResponse ->
                ...
                val sources = chatResponse.metadata.get<List<Document>>("rag_document_context")
                ...
            }