app.py

import streamlit as st
import logging
import os
from langchain_community.vectorstores import FAISS
from langchain_community.document_loaders import CSVLoader
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain.prompts import PromptTemplate
from langchain.llms import HuggingFaceHub
import dotenv
import yaml

# Load environment variables from .env file
dotenv.load_dotenv()

# Load configuration from YAML
def load_config():
    with open("yaml-editor-online.yaml", "r") as f:
        config = yaml.safe_load(f)
    return config

# Securely fetch Hugging Face API key
hf_token =  os.getenv('HUGGING') # Set this in your environment

# Load configuration
config = load_config()
logging.basicConfig(level=logging.INFO)

# Initialize Embeddings Model
instructor_embeddings = HuggingFaceEmbeddings(model_name=config["embedding_model"])

# Initialize FAISS Vector Database
def create_vector_db():
    try:
        loader = CSVLoader(file_path="plant_biodata.csv", source_column="Information about the disease")
        data = loader.load()
        
        # Create FAISS vector store
        vectordb = FAISS.from_documents(documents=data, embedding=instructor_embeddings)
        vectordb.save_local(config["vector_db_path"])
        logging.info("Vector database successfully created and saved.")
    except Exception as e:
        logging.error("Error creating vector database:", exc_info=e)

# Load FAISS and retrieve relevant documents
import textwrap

def get_qa_chain(query):
    try:
        if not os.path.exists(config["vector_db_path"]):
            logging.error(f"FAISS index path does not exist: {config['vector_db_path']}")
            return "Error: No data found."

        vectordb = FAISS.load_local(
            config["vector_db_path"], instructor_embeddings, allow_dangerous_deserialization=True
        )
        retriever = vectordb.as_retriever(score_threshold=config["score_threshold"])

        # Retrieve top-k relevant documents
        relevant_docs = retriever.get_relevant_documents(query)[:3]

        if not relevant_docs:
            return "No relevant information found."

        # Summarizing relevant documents
        summarized_context = " ".join(doc.page_content for doc in relevant_docs)
        summarized_context = textwrap.shorten(summarized_context, width=600, placeholder="...")

        # Refined prompt enforcing bullet-point output
        prompt_template = """Given the following context and a question, generate a structured answer in bullet points.
        QUESTION: {query}
        
        Ensure the response is simple, clear, and formatted as bullet points avoiding complex terms.
        """
        prompt = PromptTemplate(input_variables=["query"], template=prompt_template).format(query=query)

        # Call LLM
        llm = HuggingFaceHub(
            repo_id=config["model_name"],
            model_kwargs={
                "temperature": config["temperature"],
                "max_length": 150,
                "top_p": config["top_p"],
                "top_k": config["top_k"]
            },
            huggingfacehub_api_token=hf_token
        )

        response = llm(prompt)
        return response.strip()

    except Exception as e:
        logging.error("Error getting response:", exc_info=e)
        return "Sorry, there was an error processing your request."

# Streamlit UI with Dark Mode
def main():
    st.set_page_config(page_title="Crop Disease Assistant", page_icon="🌱", layout="centered")
    
    # Custom CSS for Dark Mode
    dark_theme_css = """
    <style>
    body {
        background-color: #121212;
        color: #ffffff;
    }
    .stTextInput, .stButton>button {
        background-color: #1e1e1e;
        color: white;
        border-radius: 5px;
        border: 1px solid #444;
    }
    .stMarkdown {
        background-color: #1e1e1e;
        padding: 10px;
        border-radius: 5px;
    }
    </style>
    """
    st.markdown(dark_theme_css, unsafe_allow_html=True)
    
    st.title("🌾 Crop Disease Assistant")
    st.write("Enter a crop disease-related question, and get simplified information.")

    # Text input for the user query
    query = st.text_input("Enter your query:")

    # Display response on clicking the button
    if st.button("Get Information"): 
        if query:
            response = get_qa_chain(query)
            print("HELLOOOO")
            print(response[300:])
            st.markdown(f"**📌 Response:**\n\n{response[250:]}")
        else:
            st.write("Please enter a query to get a response.")

if __name__ == "__main__":
    if not os.path.exists(config["vector_db_path"]):
        logging.info(f"Vector database not found at {config['vector_db_path']}, creating it now.")
        create_vector_db()
    main()