Issue with Trulens Evaluation and TruChain(using LangChain)

I'm currently working on LLM evaluation using Trulens and Azure openai and encountering some issues with the code. Here's the relevant portion of my code:

retriever = vectorstore .as_retriever()

prompt = hub.pull("rlm/rag-prompt")

def format_docs(docs):
    return "\n\n".join(doc.page_content for doc in docs)

rag_chain = (
    {"context": retriever | format_docs, "question": RunnablePassthrough()}
    | prompt
    | llm
    | StrOutputParser()
)

When attempting to apply the Trulens recorder with the following code:


# TruLens Eval chain recorder
chain_recorder = TruChain(
    rag_chain, app_id="Chain1_ChatApplication", feedbacks=[f_qa_relevance, f_context_relevance, f_groundedness])
I encounter the error:

ValidationError: 1 validation error for TruChain
app
  Can't instantiate abstract class Chain with abstract methods _call, input_keys, output_keys (type=type_error)

I also tried applying TruChain with an llm Langchain chain:

chain = LLMChain(llm=llm, prompt=prompt)

qa = RetrievalQA.from_chain_type(
    llm=llm,
    retriever=vectorstore.as_retriever()
)
chain_recorder = TruChain(
    qa, app_id="contextual-chatbot", feedbacks=[f_qa_relevance, f_context_relevance, f_groundedness]
)

However, this results in the error:(in the chain_recorder part)

TypeError: Object of type 'OpenAI' is not JSON serializable I'm using Python 3.10.11 and have also tested with Python 3.9. The relevant package versions are as follows:

langchain==0.1.0 langchain-community==0.0.9 langchain-core==0.1.7 langchainhub==0.1.14 trulens-eval==0.18.1

truera / trulens

Issue with Trulens Evaluation and TruChain(using LangChain) #776