import streamlit as st import chromadb from chromadb.utils import embedding_functions from sentence_transformers import SentenceTransformer from openai import OpenAI client = chromadb.PersistentClient(path="./chromadb/") MODEL_NAME: str = "mixedbread-ai/mxbai-embed-large-v1" # ~ 0.5 gb COLLECTION_NAME: str = "scheme" EMBEDDING_FUNC = embedding_functions.SentenceTransformerEmbeddingFunction(model_name=MODEL_NAME) schemer = client.get_collection( name=COLLECTION_NAME, embedding_function=EMBEDDING_FUNC, ) DATA_AVAL: bool = schemer.count() > 0 APP_NAME: str = "Groove-GPT" st.title(APP_NAME) st.header("What is Groovy-GPT?") st.write("Groovy-GPT is a RAG (Retrieval-Augmented Generation) model that uses ChromaDB to retrieve relevant documents and then uses OpenAI's models to generate a response.") st.write("The model is trained on the MIT Scheme textbook and a handful of Discrete Math and Paradigms related content that Professor Troeger posted") st.write("Data Avaliable: ", DATA_AVAL) user_question: str = st.text_area("Enter your groovy questions here") access_key: str = st.text_input("Enter your gpt key here", type="password") st.markdown("*For more information about how to get an access key, read [this article](https://platform.openai.com/api-keys).*", unsafe_allow_html=True) gpt_type: str = st.selectbox(label="Choose GPT Type", options=["gpt-3.5-turbo", "gpt-3.5-turbo-1106", "gpt-3.5-turbo-0125", "gpt-4-32k-0613", "gpt-4-0613", "gpt-4-0125-preview"], index=0) st.markdown("*For more information about GPT types, read [this article](https://platform.openai.com/docs/models). Make sure it has money in it ☠️*", unsafe_allow_html=True) if st.button('Query Database') & (access_key != "") & (user_question != ""): openai_client = OpenAI(api_key=access_key) with st.spinner('Loading...'): st.header("Results") # Perform the Chromadb query. results = schemer.query( query_texts=[user_question], n_results=10, include = ['documents'] ) documents = results["documents"] response = openai_client.chat.completions.create( model="gpt-3.5-turbo", messages=[ {"role": "system", "content": "You are an expert in functional programming in Scheme, with great knowledge on programming paradigms."}, {"role": "user", "content": user_question}, {"role": "assistant", "content": str(documents)}, ] ) st.write(response.choices[0].message.content)