qa chain modifications

StudybotAPI/backend/utils/chain_loader.py

@@ -21,7 +21,7 @@ async def llm_chain_loader(DATA_PATH: str):
         prompt = f.read()
 
     prompt = PromptTemplate(
-        template=prompt, input_variables=["context", "chat_history", "question"]
+        template=prompt, input_variables=["context", "question"]
     )
 
     llm = Clarifai(
@@ -32,24 +32,26 @@ async def llm_chain_loader(DATA_PATH: str):
         model_version_id=config.MODEL_VERSION_ID,
     )
 
-    # 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(
+    qa_chain = RetrievalQA.from_chain_type(
         llm=llm,
         chain_type="stuff",
         retriever=db.as_retriever(
-            search_type="mmr", search_kwargs={"k": 2, "fetch_k": 4}),
-        # return_source_documents=True,
-        # chain_type_kwargs={"prompt": prompt},
-        condense_question_prompt=prompt,
-        memory=memory,
+            search_type="mmr", search_kwargs={"k": 2, "fetch_k": 4}
+        ),
+        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="mmr", search_kwargs={"k": 2, "fetch_k": 4}),
+    #     # 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

@@ -9,10 +9,7 @@ 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]?".
 
-Chat History:
-{chat_history}
 
-Follow Up Input: {question}
-Standalone question:
+{question}
 
 [/INST]