Spec out API for all required vs. all possible input variables - Githubissues

langchain-ai / langchain

🦜🔗 Build context-aware reasoning applications

https://python.langchain.com

MIT License

88.4k stars 13.88k forks source link

Spec out API for all required vs. all possible input variables #22832

Open eyurtsev opened 2 weeks ago

eyurtsev commented 2 weeks ago

Privileged issue

[X] I am a LangChain maintainer, or was asked directly by a LangChain maintainer to create an issue here.

Issue Content

Context:

MessagePlaceholder can be optional or non optional
Current LangChain API for input variables doesn't distinguish between possible input variables vs. required input variables

See: https://github.com/langchain-ai/langchain/pull/21640

Requirements

get_input_schema() should reflect optional and required inputs
Expose another property to either fetch all required or all possible input variables (with explanation about why this is the correct approach) alternatively delegate to get_input_schema(), and make semantics of input_variables clear (e.g., all possible values)

from langchain import LLMChain
prompt = ChatPromptTemplate.from_messages([MessagesPlaceholder("history", optional=True), ('user', '${input}')])
model = ChatOpenAI()
chain = LLMChain(llm=model, prompt=prompt)
chain({'input': 'what is your name'})
prompt.get_input_schema()

keenborder786 commented 2 weeks ago

@eyurtsev since I was already working on this issue in #21640, I can tackle this issue.