I was following an old tutorial about chaining in Langchain. With it, I was writing some demo chains of my own, such as:

prompt_candidates = ChatPromptTemplate.from_template(
    """A research has shown that the following SUBJECTS are somehow related with the CAREER '{career}':

{research_list}

From the SUBJECTS list, get me a list of ACADEMIC THEMES. The ACADEMIC THEMES list must accomplish the following requirements:

[Long list of requirements, shortened for both readability and propietary contents]
"""
)

prompt_finalists = ChatPromptTemplate.from_template(
    """I have a list of ACADEMIC THEMES, all of them for the CAREER '{career}':

{academic_themes}

Another Curricular Design Expert has determined however, that these ACADEMIC THEMES need the following CORRECTIONS:

[Another long list of requirements, shortened for both readability and propietary contents]

Rewrite the ACADEMIC THEMES, so they become compliant with the CORRECTIONS raised.
"""
)

# Chains definition
candidates_chain = LLMChain(llm=llm, prompt=prompt_candidates, output_key="academic_themes")
finalists_chain = LLMChain(llm=llm, prompt=prompt_finalists, output_key="finalists")

# Chaining
final_chain = SequentialChain(
    chains=[candidates_chain, finalists_chain],
    input_variables=["career", "research_list"],
    output_variables=["finalists"],
    verbose=False
)

However, I got the following warning:

LangChainDeprecationWarning: The class `LLMChain` was deprecated in LangChain 0.1.17 and will be removed in 1.0. Use RunnableSequence, e.g., `prompt | llm` instead.
  candidates_chain = LLMChain(llm=llm, prompt=prompt_candidates, output_key="academic_themes")

Indeed, I was reading the docs, which ask you to use the pipe "|" operator; however the examples provided there are very simple, and usually involve a prompt and a llm, which are very straightforward (and are even provided in the same warning message); however I could not figure out how to adapt the pipe operator in my own chain.

I was thinking of something like:

from langchain_core.output_parsers import StrOutputParser

chain_a = prompt_candidates | llm | StrOutputParser()
chain_b = prompt_finalists | llm | StrOutputParser()
composed_chain = chain_a | chain_b
output_chain=composed_chain.invoke(
    {
        "career": "Artificial Intelligence",
        "research_list": "\n".join(research_col)
    }
)

But this gets me:

TypeError: Expected mapping type as input to ChatPromptTemplate. Received <class 'str'>.

I have tried several stuff, but nothing functional. What am I doing wrong?