Spaces:
Sleeping
Sleeping
File size: 1,447 Bytes
5b97f9e 3c5f44b 5b97f9e 3c5f44b 5b97f9e 3c5f44b 5b97f9e 3c5f44b 5b97f9e 3c5f44b 5b97f9e 3c5f44b 5b97f9e 3c5f44b 5b97f9e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 |
from langchain.retrievers import WikipediaRetriever
import wikipedia
from chatbot.llm import gemini_llm
from chatbot.memory import memory
from chatbot.prompts import chat_prompt
from langchain.chains import ConversationalRetrievalChain
def search_wikipedia(query: str, language: str = "vi"):
"""Search Wikipedia in the specified language (vi or en)."""
wikipedia.set_lang(language) # Set Wikipedia language dynamically
retriever = WikipediaRetriever() # Create a new retriever each time to apply language setting
return retriever.get_relevant_documents(query)
def get_retriever(user_input: str):
"""Decide which language retriever to use based on user input."""
# Example logic: If input contains English words, use "en"; otherwise, use "vi".
if any(char.isascii() for char in user_input):
return search_wikipedia(user_input, language="en")
return search_wikipedia(user_input, language="vi")
qa_chain = ConversationalRetrievalChain.from_llm(
llm=gemini_llm,
retriever=get_retriever, # Dynamic Wikipedia search
memory=memory,
return_source_documents=False,
combine_docs_chain_kwargs={"prompt": chat_prompt},
output_key="result"
)
def get_chat_response(user_input: str) -> str:
response = qa_chain(user_input)
# Lưu vào bộ nhớ hội thoại
memory.save_context({"input": user_input}, {"output": response["result"]})
return response["result"]
|