CondensePlusContextChatEngine return retriever nodes

[baxteran]

I want to measure the result relevance retrieved from a vector store by a CondensePlusContextChatEngine - similar to the following:

memory = ChatMemoryBuffer.from_defaults(token_limit=3900)

chat_engine = CondensePlusContextChatEngine.from_defaults(
    index.as_retriever(similarity_top_k=5),
    memory=memory,
    llm=llm,
    context_prompt=context_prompt,
    #condense_prompt=condense_prompt,
    verbose=True,
)

Is there a simple way to get the chat engine to return the retriever results when invoking chat_engine.chat()?

Thanks!

[dosubot[bot]]

Hello @baxteran! I'm here to help you with any bugs, questions, or contributions. Let's tackle this together!

To get the CondensePlusContextChatEngine to return retriever results when invoking chat_engine.chat(), use the chat() method. This method retrieves context using the retriever and includes it in the response.

Here's how you can access the retriever results:

response = chat_engine.chat("Your message here")
retriever_results = response.sources  # This contains the retriever results

The chat() method returns an AgentChatResponse that includes both the response and the sources (retriever results) ^[1].

To continue talking to Dosu, mention @dosu.

[baxteran]

This works, though upon further investigation I found response.source_nodes to be more useful for my purpose