Search code examples
langchain

PromptTemplate Validation in ConversationChain

Been stuck on trying to add variables to the prompt. template using this example from Langchain Docs as a starting point.

ValidationError: 1 validation error for ConversationChain __root__ Got unexpected prompt input variables. The prompt expects ['history', 'input', 'daily_context'], but got ['history'] as inputs from memory, and input as the normal input key. (type=value_error

Any suggestions here to get past the validation?

daily_context = "It is a Saturday, the office is open."

template = """You are Jane's Boss.  

Here is some context about today: {daily_context}

Current conversation:
{history}
Jane: {input}
AI:"""
PROMPT = PromptTemplate(input_variables=[ "history", "input", "daily_context"], template=template)

llm = OpenAI(temperature=0.4, model="gpt-3.5-turbo-instruct")
conversation = ConversationChain(
 prompt=PROMPT,
 llm=llm,
 verbose=True,
 memory=ConversationBufferMemory(),
)
Solution

  • When the context is available at this point you can prepopulate the prompt like so:

    PROMPT = PromptTemplate.from_template(template).partial(daily_context=daily_context)

    from_template is the recommended way to instantiate a prompt, which saves you from specifying input variables.