Backend changes for conversational qa

StudybotAPI/backend/utils/chain_loader.py

@@ -10,6 +10,7 @@ from langchain.chains import (
 )
 from langchain.llms import Clarifai
 from langchain.prompts import PromptTemplate
+from langchain.memory import ConversationBufferMemory
 
 
 async def llm_chain_loader(DATA_PATH: str):
@@ -19,7 +20,9 @@ async def llm_chain_loader(DATA_PATH: str):
     with open("backend/utils/prompt.txt", "r", encoding="utf8") as f:
         prompt = f.read()
 
-    prompt = PromptTemplate(template=prompt, input_variables=["context", "question"])
+    prompt = PromptTemplate(
+        template=prompt, input_variables=["context", "chat_history", "question"]
+    )
 
     llm = Clarifai(
         pat=config.CLARIFAI_PAT,
@@ -29,12 +32,23 @@ async def llm_chain_loader(DATA_PATH: str):
         model_version_id=config.MODEL_VERSION_ID,
     )
 
-    qa_chain = RetrievalQA.from_chain_type(
+    # qa_chain = RetrievalQA.from_chain_type(
+    #     llm=llm,
+    #     chain_type="stuff",
+    #     retriever=db.as_retriever(search_type="similarity",search_kwargs={"k": 2}),
+    #     return_source_documents=True,
+    #     chain_type_kwargs={"prompt": prompt},
+    # )
+
+    memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
+    qa_chain = ConversationalRetrievalChain.from_llm(
         llm=llm,
         chain_type="stuff",
-        retriever=db.as_retriever(search_type="similarity",search_kwargs={"k": 2}),
-        return_source_documents=True,
-        chain_type_kwargs={"prompt": prompt},
+        retriever=db.as_retriever(search_type="similarity", search_kwargs={"k": 2}),
+        # return_source_documents=True,
+        # chain_type_kwargs={"prompt": prompt},
+        condense_question_prompt=prompt,
+        memory=memory,
     )
 
     app.state.qa_chain = qa_chain

StudybotAPI/backend/utils/prompt.txt

@@ -9,4 +9,10 @@ The "SOURCES" part should be a reference to the source of the document from whic
 Consider a student engaged in the study of any theoretical subject, where the abundance of concepts and events poses a challenge to memorization. The aim is to overcome this hurdle and be capable of providing brief answers to specific queries. For example, if a student forgets a key concept, date, or event, they can ask the bot a question like "What is [specific query]?" for a concise answer.
 Note that students can also ask multiple questions in a single query. For example, "What is [specific query 1]?, What is [specific query 2]?, What is [specific query 3]?".
 
-{question} [/INST]
+Chat History:
+{chat_history}
+
+Follow Up Input: {question}
+Standalone question:
+
+[/INST]