app.py

import streamlit as st
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
import PyPDF2
from dotenv import load_dotenv
import os

# Load environment variables from .env
load_dotenv()

# API Key
access_token = os.getenv("API_KEY")

# Streamlit App Title
st.title("Job Description and CV-Based Email Generator")
st.write("""
This app uses Hugging Face's Gemma model to generate a professional email based on a pre-parsed CV and a job description.
Upload your CV once in the sidebar, and the system will reuse the parsed details for generating emails.
""")

# Sidebar for Settings and CV Upload
st.sidebar.title("Settings and CV Upload")

# File Upload for CV in Sidebar
uploaded_file = st.sidebar.file_uploader("Upload your CV (PDF format):", type=["pdf"])

if "parsed_cv" not in st.session_state:
    st.session_state.parsed_cv = None

if "email_history" not in st.session_state:
    st.session_state.email_history = []

if uploaded_file is not None:
    try:
        # Extract text from PDF
        pdf_reader = PyPDF2.PdfReader(uploaded_file)
        cv_text = "".join([page.extract_text() for page in pdf_reader.pages])
        st.sidebar.success("CV uploaded and text extracted successfully!")

        # Parse CV details and save to session state
        def parse_cv(cv_text):
            return f"""
            Name: [Extracted Name]
            Contact Information: [Extracted Contact Info]
            Skills: [Extracted Skills]
            Experience: [Extracted Experience]
            Education: [Extracted Education]
            Summary: {cv_text[:500]}...  # Truncated summary of the CV
            """

        st.session_state.parsed_cv = parse_cv(cv_text)
        st.sidebar.success("CV parsed successfully!")
    except Exception as e:
        st.sidebar.error(f"Failed to extract text from CV: {e}")

if st.session_state.parsed_cv:
    st.sidebar.write("### Parsed CV Details:")
    st.sidebar.text(st.session_state.parsed_cv)

# Ensure Access Token is Provided
if access_token:
    @st.cache_resource
    def initialize_pipeline(access_token):
        try:
            tokenizer = AutoTokenizer.from_pretrained("google/gemma-2b-it", token=access_token)
            model = AutoModelForCausalLM.from_pretrained(
                "google/gemma-2b-it",
                token=access_token,
            )
            return pipeline(
                "text-generation",
                model=model,
                tokenizer=tokenizer,
                max_new_tokens=2048,
                temperature=0.7,
                top_p=0.95
            )
        except Exception as e:
            st.error(f"Failed to initialize the model: {str(e)}")
            return None

    text_gen_pipeline = initialize_pipeline(access_token)

    # Input job description
    job_description = st.text_area("Enter the job description:", "")

    # Display generated email
    if st.button("Generate Email"):
        if st.session_state.parsed_cv and job_description.strip():
            try:
                # Improved prompt template
                prompt = f"""Task: Write a professional job application email.

CV Summary:
{st.session_state.parsed_cv}

Job Description:
{job_description}

Instructions: Write a concise and professional email expressing interest in the position. 
Highlight relevant experience and skills from the CV that match the job requirements.
Keep the tone professional and enthusiastic.

Email:
"""
                # Generate email using the pipeline
                if text_gen_pipeline:
                    response = text_gen_pipeline(
                        prompt,
                        clean_up_tokenization_spaces=True,
                        return_full_text=False
                    )[0]['generated_text']

                    # Save response in history
                    st.session_state.email_history.append({
                        "job_description": job_description,
                        "email": response
                    })

                    # Display response
                    st.subheader("Generated Email:")
                    st.write(response)

                    # Display conversation history
                    if st.session_state.email_history:
                        st.subheader("Previous Generations:")
                        for idx, entry in enumerate(st.session_state.email_history, 1):
                            st.write(f"### Email {idx}")
                            st.write(f"**Job Description:** {entry['job_description']}")
                            st.write(f"**Generated Email:** {entry['email']}")
                else:
                    st.error("Text generation pipeline not properly initialized.")
            except Exception as e:
                st.error(f"Error generating email: {str(e)}")
        else:
            st.warning("Please upload your CV in the sidebar and enter a job description.")
else:
    st.warning("Please enter your Hugging Face access token in the sidebar to use the app.")