llm.py

# llm.py
from langchain_openai import ChatOpenAI
from langchain_anthropic import ChatAnthropic
from langchain_google_genai import GoogleGenerativeAI
from langchain_groq import ChatGroq
from langchain_community.chat_models import ChatOllama
from langchain_core.runnables import ConfigurableField
from langchain.callbacks.base import BaseCallbackHandler


class StreamCallback(BaseCallbackHandler):
    def on_llm_new_token(self, token: str, **kwargs):
        print(token, end="", flush=True)


def get_llm(streaming=True):
    return ChatOpenAI(
        model="gpt-4",
        temperature=0,
        streaming=streaming,
        callbacks=[StreamCallback()],
    ).configurable_alternatives(
        ConfigurableField(id="llm"),
        default_key="gpt4",
        claude=ChatAnthropic(
            model="claude-3-opus-20240229",
            temperature=0,
            streaming=streaming,
            callbacks=[StreamCallback()],
        ),
        gpt3=ChatOpenAI(
            model="gpt-3.5-turbo",
            temperature=0,
            streaming=streaming,
            callbacks=[StreamCallback()],
        ),
        gemini=GoogleGenerativeAI(
            model="gemini-1.5-flash",
            temperature=0,
            streaming=streaming,
            callbacks=[StreamCallback()],
        ),
        llama3=ChatGroq(
            model_name="llama3-70b-8192",
            temperature=0,
            streaming=streaming,
            callbacks=[StreamCallback()],
        ),
        ollama=ChatOllama(
            model="EEVE-Korean-10.8B:long",
            streaming=streaming,
            callbacks=[StreamCallback()],
        ),
    )