Update app.py

app.py

@@ -10,6 +10,7 @@ from langchain_community.llms import HuggingFacePipeline
 from langchain.chains import ConversationChain
 from langchain.memory import ConversationBufferMemory
 from langchain_community.llms import HuggingFaceEndpoint
+from langchain import PromptTemplate
 
 from pathlib import Path
 import chromadb
@@ -166,12 +167,22 @@ def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db, pr
     # retriever=vector_db.as_retriever(search_type="similarity", search_kwargs={'k': 3})
     retriever=vector_db.as_retriever()
     progress(0.8, desc="Defining retrieval chain...")
+
+    prompt_template_string = f'''Your task is as follows: 1. Determine if the input {query} is compliant with the provided {context}. 2. If the requirement is compliant, report "This requirement is compliant." 3. If the requirement is not compliant report "This requirement is not compliant." 4. If the requirement is not compliant, give the reason for non compliance and return the specific rule or guideline the requirement violates. 5. If the requirement is not compliant, report a refined version of the requirement delimited in quotes that is compliant with the provided {context}.'''
+
+    # Create a PromptTemplate object
+    prompt_template = PromptTemplate(
+        template=prompt_template_string,
+        input_variables=["user_question"]
+    )
+
+    
     qa_chain = ConversationalRetrievalChain.from_llm(
         llm,
         retriever=retriever,
         chain_type="stuff", 
         memory=memory,
-        combine_docs_chain_kwargs={"prompt": """Your task is as follows: 1. Determine if the input {query} is compliant with the provided {context}. 2. If the requirement is compliant, report "This requirement is compliant." 3. If the requirement is not compliant report "This requirement is not compliant." 4. If the requirement is not compliant, give the reason for non compliance and return the specific rule or guideline the requirement violates. 5. If the requirement is not compliant, report a refined version of the requirement delimited in quotes that is compliant with the provided {context}."""},
+        combine_docs_chain_kwargs={"prompt": prompt_template},
         return_source_documents=True,
         #return_generated_question=False,
         verbose=False,
@@ -226,11 +237,11 @@ def initialize_database(list_file_obj, chunk_size, chunk_overlap, progress=gr.Pr
     return vector_db, collection_name, "Complete!"
 
 
-def initialize_LLM(llm_option, llm_temperature, max_tokens, top_k, vector_db, progress=gr.Progress()):
+def initialize_LLM(llm_option, prompt_template, llm_temperature, max_tokens, top_k, vector_db, progress=gr.Progress()):
     # print("llm_option",llm_option)
     llm_name = list_llm[llm_option]
     print("llm_name: ",llm_name)
-    qa_chain = initialize_llmchain(llm_name, llm_temperature, max_tokens, top_k, vector_db, progress)
+    qa_chain = initialize_llmchain(llm_name, prompt = prompt_template, llm_temperature, max_tokens, top_k, vector_db, progress)
     return qa_chain, "Complete!"