core: update load chat prompt template multi mesages

[jhpiedrahitao]

• Description: update load chat prompt template method to allow chat prompt templates
  with multiple messages (loles) when using file loading (load from config)
• Issue: Mechansim to load prompt templates from a file (e.g. yaml) #21672

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Skipped Deployment

Name	Status	Preview	Comments	Updated (UTC)
langchain	⬜️ Ignored (Inspect)	Visit Preview		Nov 14, 2024 5:21pm

[baskaryan]

could we add a unit test for the change

[jhpiedrahitao]

@baskaryan tests and examples for loading added

[jhpiedrahitao]

@baskaryan please take a look when you have the time

[ccurme]

Closing as I believe this functionality exists:

from langchain_core.prompts import ChatPromptTemplate
from langchain_core.load import dumpd
from langchain_core.load.load import load
import json

prompt = ChatPromptTemplate(
    [
        ("system", "system message is {system_message}"),
        ("user", "user message is {user_message}"),
    ]
)

with open("foo.json", "w") as f:
  json.dump(dumpd(prompt), f)

with open("foo.json", "r") as f:
  reloaded = load(json.load(f))

See guide here: https://python.langchain.com/docs/how_to/serialization/

Let me know if this does not cover your use case. Thank you!

[jhpiedrahitao]

@ccurme I think this not covers loading a full conversation (multiple turns with roles in a template), The current implementation loads only one message into the chat template