m

main.py

@@ -7,7 +7,7 @@ from question import chat_with_doc
 from langchain.embeddings import HuggingFaceInferenceAPIEmbeddings
 from langchain.vectorstores import SupabaseVectorStore
 from supabase import Client, create_client
-from stats import add_usage
+from stats import get_usage
 
 supabase_url = st.secrets.SUPABASE_URL
 supabase_key = st.secrets.SUPABASE_KEY
@@ -71,6 +71,8 @@ if 'max_tokens' not in st.session_state:
 if 'username' not in st.session_state:
     st.session_state['username'] = username
 
-chat_with_doc(st.session_state['model'], vector_store, stats_db=supabase)
+stats = str(get_usage(supabase)) 
+
+chat_with_doc(st.session_state['model'], vector_store, stats_db=supabase, stats=stats)
 
 st.markdown("---\n\n")

question.py

@@ -7,7 +7,7 @@ from langchain.llms import OpenAI
 from langchain.llms import HuggingFaceEndpoint
 from langchain.chat_models import ChatAnthropic
 from langchain.vectorstores import SupabaseVectorStore
-from stats import add_usage, get_usage
+from stats import add_usage
 
 memory = ConversationBufferMemory(memory_key="chat_history", input_key='question', output_key='answer', return_messages=True)
 openai_api_key = st.secrets.openai_api_key
@@ -15,13 +15,12 @@ anthropic_api_key = st.secrets.anthropic_api_key
 hf_api_key = st.secrets.hf_api_key
 logger = get_logger(__name__)
 
-def chat_with_doc(model, vector_store: SupabaseVectorStore, stats_db):
+def chat_with_doc(model, vector_store: SupabaseVectorStore, stats_db, stats):
     
     if 'chat_history' not in st.session_state:
         st.session_state['chat_history'] = []
       
-    stats = str(get_usage(stats_db)) 
-    question = st.text_area("## Ask a question (" + stats + " queries answered so far)", max_chars=500)
+    query = st.text_area("## Ask a question (" + stats + " queries answered so far)", max_chars=500)
     columns = st.columns(2)
     with columns[0]:
         button = st.button("Ask")
@@ -38,7 +37,7 @@ def chat_with_doc(model, vector_store: SupabaseVectorStore, stats_db):
 
     if button:
         qa = None
-        add_usage(stats_db, "chat", "prompt" + question, {"model": model, "temperature": st.session_state['temperature']})
+        add_usage(stats_db, "chat", "prompt" + query, {"model": model, "temperature": st.session_state['temperature']})
         if model.startswith("gpt"):
             logger.info('Using OpenAI model %s', model)
             qa = ConversationalRetrievalChain.from_llm(
@@ -62,12 +61,14 @@ def chat_with_doc(model, vector_store: SupabaseVectorStore, stats_db):
                 huggingfacehub_api_token=hf_api_key,
                 model_kwargs=model_kwargs
             )
-            qa = ConversationalRetrievalChain.from_llm(hf, retriever=vector_store.as_retriever(search_kwargs={"score_threshold": 0.8, "k": 4,"filter": {"user": st.session_state["username"]}}), memory=memory, verbose=True, return_source_documents=True)
+            qa = ConversationalRetrievalChain.from_llm(hf, retriever=vector_store.as_retriever(search_kwargs={"score_threshold": 0.6, "k": 4,"filter": {"user": st.session_state["username"]}}), memory=memory, verbose=True, return_source_documents=True)
         
-        st.session_state['chat_history'].append(("You", question))
+        print("Question>")
+        print(query)
+        st.session_state['chat_history'].append(("You", query))
 
         # Generate model's response and add it to chat history
-        model_response = qa({"question": question})
+        model_response = qa({"question": query})
         logger.info('Result: %s', model_response["answer"])
         sources = model_response["source_documents"]
         logger.info('Sources: %s', model_response["source_documents"])
@@ -77,9 +78,8 @@ def chat_with_doc(model, vector_store: SupabaseVectorStore, stats_db):
         else:
             st.session_state['chat_history'].append(("Safety Copilot", "I am sorry, I do not have enough information to provide an answer. If there is a public source of data that you would like to add, please email copilot@securade.ai."))
         
-        
-    # Display chat history
-    st.empty()
-    chat_history = st.session_state['chat_history']
-    for speaker, text in chat_history:
-        st.markdown(f"**{speaker}:** {text}")
+        # Display chat history
+        st.empty()
+        chat_history = st.session_state['chat_history']
+        for speaker, text in chat_history:
+            st.markdown(f"**{speaker}:** {text}")

requirements.txt

@@ -1,9 +1,9 @@
-langchain
+langchain==0.1.0
 Markdown==3.4.3
 openai==0.27.6
 pdf2image==1.16.3
 pypdf==3.8.1
-streamlit==1.30.0
+streamlit==1.22.0
 StrEnum==0.4.10
 supabase==1.0.3
 tiktoken==0.4.0
@@ -12,5 +12,5 @@ anthropic==0.2.8
 fastapi==0.95.2
 python-multipart==0.0.6
 uvicorn==0.22.0
-docx2txt
-st-login-form
+docx2txt==0.8
+st-login-form==0.2.1