# Urgent help needed for Pinecone metadata filter issue in chatbot using OpenAI Langchain Streamlit; **URL:** \ **Category:** Support\ **Tags:** semantic-search, vector-database, pinecone-assistant\ **Created:** [May 7, 2023, 1:05pm UTC](https://community.pinecone.io/t/urgent-help-needed-for-pinecone-metadata-filter-issue-in-chatbot-using-openai-langchain-streamlit/1731 "2023-05-07T13:05:58Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![CrudenBoy](https://avatars.discourse-cdn.com/v4/letter/c/b77776/32.png) [@CrudenBoy](https://community.pinecone.io/u/CrudenBoy)\ **Post date:** [May 7, 2023, 1:05pm UTC](https://community.pinecone.io/t/urgent-help-needed-for-pinecone-metadata-filter-issue-in-chatbot-using-openai-langchain-streamlit/1731/1 "2023-05-07T13:05:58Z")
Hello Pinecone Community, I hope this message finds you well. I am reaching out for your help in resolving a issue that I have been struggling with for days. I have a chatbot that’s currently functioning and I have created Ui on Streamlit to load the vectors and metadata, but I want to filter the results by a metadata field ‘pdf\_id’ that I added. I am using the - OpenAI - Langchain - Streamlit solution for this project. The issue I’m encountering is that I can’t get the metadata filter to work. My chatbot script calls functions from a **utils scrip** t. When I run this script in Visual Studio Code, I keep getting an error: **“Import ‘pinecone’ could not be resolved PylancereportMissingImports.”** However, this issue doesn’t prevent the script from running when I use the functions in the **1) Functions work but do not filter by metadata** section below. Now, I want to replace these functions so I can filter metadata using the following code in my chatbot script: ```auto filter = {"pdf_id": {"$eq": "fd31d0e8"}} response = get_answer(query, filter=filter) # Replace ``` I have tried to use metadata filtering documents on pinecone, and countless other solutions, but nothing seems to work. You can see one of the many attempts in the **2) Functions do not work with metadata filters** section below. I would really appreciate your assistance. I am at my wit’s end with this problem and would be grateful for any advice or help you can provide. Thank you in advance for your support! Here is the **utlis script** : ```auto import os from dotenv import load_dotenv load_dotenv("API.env") import langchain import openai import pinecone from langchain.embeddings.openai import OpenAIEmbeddings from langchain.vectorstores import Pinecone from langchain.llms import OpenAI as ChatOpenAI from langchain.chains.question_answering import load_qa_chain # Define Pinecone API key and environment PINECONE_API_KEY = os.getenv("PINECONE_API_KEY") PINECONE_API_ENV = os.getenv("PINECONE_API_ENV") # Initialize Pinecone pinecone.init( api_key=PINECONE_API_KEY, environment=PINECONE_API_ENV ) index_name = "langchain2" # Embeddings embeddings = OpenAIEmbeddings(openai_api_key=openai.api_key) # Pinecone index docs=[] index = Pinecone.from_documents(docs, embeddings, index_name=index_name) # Load the LLM and QA Chain llm = ChatOpenAI(temperature=0) chain = load_qa_chain(llm, chain_type="stuff") def get_initial_message(): messages = [ {"role": "system", "content": "You are a helpful assistant that supports users who are international development practitioners. If you do not know the answer to any question, you will seek further clarification from the user"}, {"role": "user", "content": "I need you to help me research some documents in your knowledgebase"}, {"role": "assistant", "content": "Thats awesome, what do you want to know?"} ] return messages ################################################################ # **1) Functions work but do not filter by metadata** def get_similiar_docs(query, k=2, score=False): if score: similar_docs = index.similarity_search_with_score(query, k=k) else: similar_docs = index.similarity_search(query, k=k) return similar_docs def get_answer(prompt): similar_docs = get_similiar_docs(prompt) answer = chain.run(input_documents=similar_docs, question=prompt) return answer # **2) Functions do not work with metadata filters** # **AttributeError:**'Pinecone' object has no attribute 'query' def get_similar_docs(query_vector, index, k=5, filter=None): similar_docs = index.query(queries=[query_vector], top_k=k, filter=filter) return similar_docs def get_answer(prompt, filter=None): similar_docs = get_similar_docs(prompt, index, k=5, filter=filter) answer = chain.run(input_documents=similar_docs, question=prompt) return answer ##################################################################### def get_chatgpt_response(messages, model="gpt-3.5-turbo"): print("model: ", model) response = openai.ChatCompletion.create( model=model, messages=messages ) return response['choices'][0]['message']['content'] def update_chat(messages, role, content): messages.append({"role": role, "content": content}) return messages ``` Here is the chatbot script ```auto # Streamlit Chatbot UI script import pinecone import streamlit as st from streamlit_chat import message from chat_utils1 import get_initial_message, update_chat, get_answer # Import the get_answer function import os from dotenv import load_dotenv load_dotenv('C:/Users/david/My Drive/Colab Notebooks/Building a GPT-4 Chatbot using ChatGPT API and Streamlit Chat/API.env') # Pinecone initialization PINECONE_API_KEY = os.getenv("PINECONE_API_KEY") PINECONE_API_ENV = os.getenv("PINECONE_API_ENV") index_name = "langchain2" index = pinecone.Index(index_name=index_name) # Create the Pinecone instance st.title("Chatbot : ChatGPT for Evaluators") st.subheader("AI Evaluation Companion:") model = st.selectbox( "Select a model", ("gpt-3.5-turbo", "gpt-4") ) #Initialize the session states to store the generated messages, past queries, and the initial set of messages. if 'generated' not in st.session_state: st.session_state['generated'] = [] if 'past' not in st.session_state: st.session_state['past'] = [] query = st.text_input("Query: ", key="input") if 'messages' not in st.session_state: st.session_state['messages'] = get_initial_message() #Process the user's query and generate the AI response. if query: with st.spinner("generating..."): messages = st.session_state['messages'] messages = update_chat(messages, "user", query) filter = {"pdf_id": {"$eq": "fd31d0e8"}} response = get_answer(query, filter=filter) # Replace get_chatgpt_response with get_answer messages = update_chat(messages, "assistant", response) st.session_state.past.append(query) st.session_state.generated.append(response) #Display the chat messages and an expander to show the full message history. if st.session_state['generated']: for i in range(len(st.session_state['generated'])-1, -1, -1): message(st.session_state['past'][i], is_user=True, key=str(i) + '_user') message(st.session_state["generated"][i], key=str(i)) with st.expander("Show Messages"): st.write(messages) ``` ---
**Author:** ![anthonyjohnarena](https://sea2.discourse-cdn.com/flex020/user_avatar/community.pinecone.io/anthonyjohnarena/32/712_2.png) [@anthonyjohnarena](https://community.pinecone.io/u/anthonyjohnarena)\ **Post date:** [June 11, 2023, 1:08am UTC](https://community.pinecone.io/t/urgent-help-needed-for-pinecone-metadata-filter-issue-in-chatbot-using-openai-langchain-streamlit/1731/2 "2023-06-11T01:08:35Z")
Did you ever figure out what the problem/solution was?