updated to langchain latest

Pipfile

@@ -19,6 +19,10 @@ streamlit = "*"
 streamlit-chat = "*"
 tqdm = "*"
 langchain-pinecone = "*"
+langchain-openai = "*"
+langchain-community = "*"
+isort = "*"
+langchainhub = "*"
 
 [dev-packages]
 

backend/core.py

@@ -1,27 +1,71 @@
 from dotenv import load_dotenv
 
 load_dotenv()
+from langchain_core.output_parsers import StrOutputParser
+from langchain_core.runnables import RunnablePassthrough
 
-from langchain_openai import ChatOpenAI, OpenAIEmbeddings
 
 from typing import Any, Dict, List
-from langchain.chains import ConversationalRetrievalChain
-from langchain_pinecone import PineconeVectorStore
 
+from langchain import hub
+from langchain.chains.combine_documents import create_stuff_documents_chain
+from langchain.chains.history_aware_retriever import create_history_aware_retriever
+from langchain.chains.retrieval import create_retrieval_chain
+from langchain_openai import ChatOpenAI, OpenAIEmbeddings
+from langchain_pinecone import PineconeVectorStore
 
-INDEX_NAME = "langchain-doc-index"
+from consts import INDEX_NAME
 
 
 def run_llm(query: str, chat_history: List[Dict[str, Any]] = []):
     embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
-    docsearch = PineconeVectorStore(embedding=embeddings, index_name=INDEX_NAME)
+    docsearch = PineconeVectorStore(index_name=INDEX_NAME, embedding=embeddings)
+    chat = ChatOpenAI(verbose=True, temperature=0)
+
+    rephrase_prompt = hub.pull("langchain-ai/chat-langchain-rephrase")
+
+    retrieval_qa_chat_prompt = hub.pull("langchain-ai/retrieval-qa-chat")
+    stuff_documents_chain = create_stuff_documents_chain(chat, retrieval_qa_chat_prompt)
+
+    history_aware_retriever = create_history_aware_retriever(
+        llm=chat, retriever=docsearch.as_retriever(), prompt=rephrase_prompt
+    )
+    qa = create_retrieval_chain(
+        retriever=history_aware_retriever, combine_docs_chain=stuff_documents_chain
+    )
 
-    chat = ChatOpenAI(
-        verbose=True,
-        temperature=0,
+    result = qa.invoke(input={"input": query, "chat_history": chat_history})
+    return result
+
+
+def format_docs(docs):
+    return "\n\n".join(doc.page_content for doc in docs)
+
+
+def run_llm2(query: str, chat_history: List[Dict[str, Any]] = []):
+    embeddings = OpenAIEmbeddings()
+    docsearch = PineconeVectorStore(index_name=INDEX_NAME, embedding=embeddings)
+    chat = ChatOpenAI(model_name="gpt-4o", verbose=True, temperature=0)
+
+    rephrase_prompt = hub.pull("langchain-ai/chat-langchain-rephrase")
+
+    retrieval_qa_chat_prompt = hub.pull("langchain-ai/retrieval-qa-chat")
+
+    rag_chain = (
+        {
+            "context": docsearch.as_retriever() | format_docs,
+            "input": RunnablePassthrough(),
+        }
+        | retrieval_qa_chat_prompt
+        | chat
+        | StrOutputParser()
     )
 
-    qa = ConversationalRetrievalChain.from_llm(
-        llm=chat, retriever=docsearch.as_retriever(), return_source_documents=True
+    retrieve_docs_chain = (lambda x: x["input"]) | docsearch.as_retriever()
+
+    chain = RunnablePassthrough.assign(context=retrieve_docs_chain).assign(
+        answer=rag_chain
     )
-    return qa.invoke({"question": query, "chat_history": chat_history})
+
+    result = chain.invoke({"input": query, "chat_history": chat_history})
+    return result

consts.py

@@ -0,0 +1 @@
+INDEX_NAME = "langchain-doc-index"

ingestion.py

@@ -2,23 +2,18 @@
 
 load_dotenv()
 
-import os
-
-from langchain_community.document_loaders import ReadTheDocsLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
-from langchain_pinecone import PineconeVectorStore
+from langchain_community.document_loaders import ReadTheDocsLoader
 from langchain_openai import OpenAIEmbeddings
+from langchain_pinecone import PineconeVectorStore
 
-
-INDEX_NAME = "langchain-doc-index"
+from consts import INDEX_NAME
 
 embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
 
 
 def ingest_docs():
-    loader = ReadTheDocsLoader(
-        "langchain-docs/api.python.langchain.com/en/latest/chains"
-    )
+    loader = ReadTheDocsLoader("langchain-docs/api.python.langchain.com/en/latest")
 
     raw_documents = loader.load()
     print(f"loaded {len(raw_documents)} documents")

main.py

@@ -1,9 +1,13 @@
+from dotenv import load_dotenv
+
+load_dotenv()
 from typing import Set
 
-from backend.core import run_llm
 import streamlit as st
 from streamlit_chat import message
 
+from backend.core import run_llm
+
 
 def create_sources_string(source_urls: Set[str]) -> str:
     if not source_urls:
@@ -37,16 +41,17 @@ def create_sources_string(source_urls: Set[str]) -> str:
             query=prompt, chat_history=st.session_state["chat_history"]
         )
 
-        sources = set(
-            [doc.metadata["source"] for doc in generated_response["source_documents"]]
-        )
+        sources = set(doc.metadata["source"] for doc in generated_response["context"])
+
         formatted_response = (
             f"{generated_response['answer']} \n\n {create_sources_string(sources)}"
         )
 
-        st.session_state.chat_history.append((prompt, generated_response["answer"]))
-        st.session_state.user_prompt_history.append(prompt)
-        st.session_state.chat_answers_history.append(formatted_response)
+        st.session_state["user_prompt_history"].append(prompt)
+        st.session_state["chat_answers_history"].append(formatted_response)
+        st.session_state["chat_history"].append(("human", prompt))
+        st.session_state["chat_history"].append(("ai", generated_response["answer"]))
+
 
 if st.session_state["chat_answers_history"]:
     for generated_response, user_query in zip(