Upload 5 files

make_answer.py

@@ -0,0 +1,3 @@
+def result(rag_chain,question):
+    response= rag_chain.invoke(question)
+    print(f"[풋풋이의 답변]\n{response}")

make_chain_gguh.py

@@ -0,0 +1,32 @@
+from langchain_core.runnables import RunnablePassthrough
+from langchain_core.output_parsers import StrOutputParser
+from langchain_community.chat_models import ChatOllama
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_pinecone import PineconeVectorStore
+from langchain_community.embeddings import SentenceTransformerEmbeddings
+
+def make_chain_ollama(retriever):
+    def format_docs(docs):
+        # 검색한 문서 결과를 하나의 문단으로 합쳐줍니다.
+        return "\n\n".join(doc.page_content for doc in docs)
+
+    # LangChain이 지원하는 다른 채팅 모델을 사용합니다. 여기서는 Ollama를 사용합니다.
+    llm = ChatOllama(model="zephyr:latest")
+
+    template = "\"```\" Below is an instruction that describes a task. Write a response that appropriately completes the request."\
+    "제시하는 context에서만 대답하고 context에 없는 내용은 모르겠다고 대답해"\
+    "make answer in korean. 한국어로 대답하세요"\
+    "\n\nContext:\n{context}\n;"\
+    "Question: {question}"\
+    "\n\nAnswer:"
+
+    prompt = ChatPromptTemplate.from_template(template)
+
+    rag_chain = (
+    {"context": retriever| format_docs, "question": RunnablePassthrough()}
+    | prompt
+    | llm
+    | StrOutputParser()
+    )
+
+    return rag_chain

make_chain_model.py

@@ -0,0 +1,32 @@
+from langchain_core.runnables import RunnablePassthrough
+from langchain_core.output_parsers import StrOutputParser
+from langchain_community.chat_models import ChatOllama
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_pinecone import PineconeVectorStore
+from langchain_community.embeddings import SentenceTransformerEmbeddings
+
+def make_chain_llm(retriever,llm):
+    def format_docs(docs):
+        # 검색한 문서 결과를 하나의 문단으로 합쳐줍니다.
+        return "\n\n".join(doc.page_content for doc in docs)
+
+    # LangChain이 지원하는 다른 채팅 모델을 사용합니다. 여기서는 Ollama를 사용합니다.
+    # llm = ChatOllama(model="zephyr:latest")
+
+    template = "\"```\" Below is an instruction that describes a task. Write a response that appropriately completes the request."\
+    "제시하는 context에서만 대답하고 context에 없는 내용은 생성하지마"\
+    "make answer in korean. 한국어로 대답하세요"\
+    "\n\nContext:\n{context}\n;"\
+    "Question: {question}"\
+    "\n\nAnswer:"
+
+    prompt = ChatPromptTemplate.from_template(template)
+
+    rag_chain = (
+    {"context": retriever| format_docs, "question": RunnablePassthrough()}
+    | prompt
+    | llm
+    | StrOutputParser()
+    )
+
+    return rag_chain

retrieve_docs.py

@@ -0,0 +1,5 @@
+def print_search_results(retrievers, query):
+    print(f"Query: {query}")
+    for i in range(len(retrievers.invoke(query))):
+        print(f'찾은 문장{i+1}:',
+            retrievers.invoke(query)[i].page_content)

retriever.py

@@ -0,0 +1,38 @@
+from langchain_core.runnables import RunnablePassthrough
+from langchain_core.output_parsers import StrOutputParser
+from langchain_community.chat_models import ChatOllama
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_pinecone import PineconeVectorStore
+from langchain_community.embeddings import SentenceTransformerEmbeddings
+
+import os
+from dotenv import load_dotenv
+from langchain.retrievers import BM25Retriever, EnsembleRetriever
+from kiwipiepy import Kiwi
+load_dotenv()
+
+kiwi = Kiwi()
+
+def kiwi_tokenize(text):
+    return [token.form for token in kiwi.tokenize(text)]
+# embedding_model = SentenceTransformerEmbeddings(model_name='BM-K/KoSimCSE-roberta-multitask', model_kwargs={"trust_remote_code":True}) 
+
+def retriever(pc, bm25):
+    pcretriever = pc.as_retriever(search_kwargs={'k':4})
+    kiwi_bm25 = BM25Retriever.from_documents(bm25,preprocess_func=kiwi_tokenize)
+    kiwi_bm25.k=4
+    
+    kiwibm25_pc_37 = EnsembleRetriever(
+        retrievers=[kiwi_bm25, pcretriever],  # 사용할 검색 모델의 리스트
+        weights=[0.3, 0.7],  # 각 검색 모델의 결과에 적용할 가중치
+        search_type="mmr",  # 검색 결과의 다양성을 증진시키는 MMR 방식을 사용
+    ) 
+        # Pinecone vector store 초기화  
+    # vectorstore = PineconeVectorStore(  
+    # index_name=os.getenv("INDEX_NAME"), embedding=embedding_model
+    # )  
+
+    # retriever = vectorstore.as_retriever(search_kwargs={'k': 2})
+
+    return kiwibm25_pc_37
+