app.py

import streamlit as st
from langchain_community.document_loaders import PyPDFLoader, PyPDFDirectoryLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import FAISS
from langchain_groq import ChatGroq



loader = PyPDFDirectoryLoader("Dataset")
docx = loader.load()
#st.write("Number of documents loaded:", len(docx))


text_sp = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
chunks = text_sp.split_documents(docx)
#st.write("Number of chunks created:", len(chunks))


if not chunks:
    st.error("No chunks were created. Please check the documents or text splitter settings.")
else:
    # Step 3: Create metadata
    metadatas = [{"source": f"{i}-pl"} for i in range(len(chunks))]

    
    embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2",
                                       model_kwargs={'device': "cpu"})

   
    try:
        sample_embedding = embeddings.embed_documents([chunks[0].page_content])
        #st.write("Sample embedding length:", len(sample_embedding[0]))
    except Exception as e:
        st.error(f"Error in generating embeddings: {str(e)}")

    
    try:
        vector_store = FAISS.from_documents(chunks, embeddings)
        #st.write("Vector store created successfully.")
    except IndexError as e:
        st.error("IndexError in creating vector store: " + str(e))
        st.write("Check if chunks and embeddings are non-empty and correctly formatted.")
        st.write("Chunks:", chunks)
    except Exception as e:
        st.error(f"An error occurred: {str(e)}")

llm = ChatGroq(model_name="mixtral-8x7b-32768", groq_api_key="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")

def retrieve(query, vector_store, embeddings):
    query_embedding = embeddings.embed_query(query)
    results = vector_store.similarity_search_by_vector(query_embedding, k=5)
    return results

from langchain.schema import HumanMessage, SystemMessage

def generate_response(query, retrieved_docs, llm):
    context = " ".join([doc.page_content for doc in retrieved_docs])
    messages = [
        SystemMessage(content="You are an expert in prompt engineering."),
        HumanMessage(content=f"Context: {context}\n\nQuestion: {query}\n\nAnswer:")
    ]
    response = llm(messages)
    return response.content.strip()



st.title("PromptGuru 🚀📖")

st.sidebar.markdown("PromptGuru By OpenRAG 🗣️")
st.sidebar.markdown(
    """
PromptGuru is a tool you can use for asking any queries related Prompt Engineering and Get it solved within a couple of minutes.
    """
    )

st.sidebar.markdown(
    """
Note -- This tool is in a beta stage. Kindly have some patience while generating the response and give the model time to think. 
    """
    )

st.sidebar.markdown(
    """
📧 **Get in Touch**

Federation of Entrepreneurship Development (FED)

The student body of KIIT TBI, FED fosters entrepreneurship by uniting all startup ideas on one platform. We provide access to top resources, contacts, and mentors, and nurture entrepreneurial skills through creative, authentic, and efficient techniques.

Skill Hunt 2.0

Our ongoing event features four cohorts: AI/ML, Digital Marketing, Prompt Engineering, and UI/UX. This chatbot is part of the Prompt Engineering cohort, aiming to enhance skills in creating effective AI prompts.

Follow us on:

Instagram: [Click Here](https://www.instagram.com/fedkiit/)

LinkedIn: [Click Here](https://www.linkedin.com/company/fedkiit/) 
    """
    
)

query = st.text_input("Ask your question about prompt engineering:")
if query:
    with st.spinner("Retrieving documents..."):
        retrieved_docs = retrieve(query, vector_store, embeddings)
       # st.write("Retrieved Documents:", retrieved_docs)
    
    with st.spinner("Generating response..."):
        response = generate_response(query, retrieved_docs, llm)
        st.write("Response:", response)