Update app.py

app.py

@@ -15,6 +15,7 @@ from langchain.chat_models import ChatOpenAI
 #from langchain.chains import ChatVectorDBChain
 from langchain.chains import RetrievalQA
 from langchain.document_loaders import PyPDFLoader
+from langchain.chains.question_answering import load_qa_chain
 
 # Use Chroma in Colab to create vector embeddings, I then saved them to HuggingFace so now I have to set it use them here.
 #from chromadb.config import Settings
@@ -29,7 +30,7 @@ def get_empty_state():
 
     
 #Initial prompt template, others added below from TXT file
-prompt_templates = {"All Needs Gurus": "I want you to act as a needs assessment expert."}
+prompt_templates = {"All Needs Experts": "I want you to act as a needs assessment expert."}
 
 def download_prompt_templates():
     url = "https://huggingface.co/spaces/ryanrwatkins/needs/raw/main/gurus.txt"
@@ -94,7 +95,7 @@ def submit_message(prompt, prompt_template, temperature, max_tokens, context_len
     if prompt_template:
         system_prompt = [{ "role": "system", "content": prompt_template }]
 
-    prompt_msg = { "role": "user", "content": prompt }
+    prompt_msg = [{ "role": "user", "content": prompt }]
 
 
     try:
@@ -102,14 +103,15 @@ def submit_message(prompt, prompt_template, temperature, max_tokens, context_len
 
 # completion = openai.ChatCompletion.create(model="gpt-3.5-turbo", messages=system_prompt + history[-context_length*2:] + [prompt_msg], temperature=temperature, max_tokens=max_tokens)
 
-        completion = RetrievalQA.from_chain_type(llm=ChatOpenAI(temperature=temperature, max_tokens=max_tokens, model_name="gpt-3.5-turbo"), chain_type="stuff", retriever=vectordb.as_retriever() , return_source_documents=False)
-        completion = completion({"query": system_prompt + history[-context_length*2:] + [prompt_msg] })
+        completion_chain = load_qa_chain(ChatOpenAI(temperature=temperature, max_tokens=max_tokens, model_name="gpt-3.5-turbo"), chain_type="stuff" )
+        completion = RetrievalQA(combine_documents_chain=completion_chain, retriever=vectordb.as_retriever())
+        completion = completion.run({"query":  prompt_msg })
         # from https://blog.devgenius.io/chat-with-document-s-using-openai-chatgpt-api-and-text-embedding-6a0ce3dc8bc8
         
         history.append(prompt_msg)
         history.append(completion.choices[0].message.to_dict())
 
-        #state['total_tokens'] += completion['usage']['total_tokens']
+        state['total_tokens'] += completion['usage']['total_tokens']
     
     except Exception as e:
         history.append(prompt_msg)