app.py

import streamlit as st
import PyPDF2
from docx import Document
from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM
import spacy
import pytextrank
from langchain_google_genai import ChatGoogleGenerativeAI
from langchain_core.messages import SystemMessage
from langchain_core.prompts import ChatPromptTemplate, HumanMessagePromptTemplate, MessagesPlaceholder
from langchain.memory import ConversationBufferMemory
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough, RunnableLambda
import spacy
import subprocess

# Function to check and download spaCy model
def ensure_spacy_model(model_name="en_core_web_sm"):
    try:
        spacy.load(model_name)
    except OSError:
        subprocess.run(["python", "-m", "spacy", "download", model_name])
        spacy.load(model_name)

# Function to extract text from PDF
def extract_text_from_pdf(uploaded_file):
    text = ""
    reader = PyPDF2.PdfReader(uploaded_file)
    for page in reader.pages:
        text += page.extract_text()
    return text

# Function to extract text from Word document
def extract_text_from_word(uploaded_file):
    text = ""
    doc = Document(uploaded_file)
    for paragraph in doc.paragraphs:
        text += paragraph.text + "\n"
    return text

# Function to summarize text
def summarize_text(text, max_length=1000, min_length=30):
    max_length = min(max_length, 1000)  # Ensure max_length doesn't exceed 1000

    try:
        # Initialize the summarizer pipeline
        summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
        summary = summarizer(text, max_length=max_length, min_length=min_length, do_sample=False)
        if isinstance(summary, list) and len(summary) > 0:
            return summary[0]['summary_text']
        else:
            raise ValueError("Unexpected summarizer output format")
    except Exception as e:
        return f"Error in summarization: {e}"

# Function to extract keywords using spaCy and PyTextRank
def extract_keywords(text, top_n=15):
    ensure_spacy_model("en_core_web_sm")
    nlp = spacy.load("en_core_web_sm")
    nlp.add_pipe("textrank", last=True)
    doc = nlp(text)
    keywords = [phrase.text for phrase in doc._.phrases[:top_n]]
    return keywords


# Initialize Google Generative AI chat model
def initialize_chat_model():
    with open("key.txt", "r") as f:
        GOOGLE_API_KEY = f.read().strip()

    chat_model = ChatGoogleGenerativeAI(
        google_api_key=GOOGLE_API_KEY,
        model="gemini-1.5-pro-latest",
        temperature=0.4,
        max_tokens=2000,
        timeout=120,
        max_retries=5,
        top_p=0.9,
        top_k=40,
        presence_penalty=0.6,
        frequency_penalty=0.3
    )
    return chat_model

chat_model = initialize_chat_model()

# Create Chat Template
chat_prompt_template = ChatPromptTemplate.from_messages(
    [
        SystemMessage(
            content="""  You are a language model designed to follow user instructions exactly as given.
            Do not take any actions or provide any information unless specifically directed by the user.
            Your role is to fulfill the user's requests precisely without deviating from the instructions provided."""
        ),
        MessagesPlaceholder(variable_name="chat_history"),
        HumanMessagePromptTemplate.from_template("{human_input}")
    ]
)

# Initialize the Memory
memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)

# Create an Output Parser
output_parser = StrOutputParser()

# Define a chain
chain = RunnablePassthrough.assign(
            chat_history=RunnableLambda(lambda human_input: memory.load_memory_variables(human_input)['chat_history'])
        ) | chat_prompt_template | chat_model | output_parser

# Streamlit App
st.title("Interview Preparation with AI")
st.markdown("## Part-1: Upload Files, Summarize, and Extract Keywords")

# File upload section
file1 = st.file_uploader("Upload your resume (PDF or DOCX):", type=["pdf", "docx"])
file2 = st.file_uploader("Upload the job description (PDF or DOCX):", type=["pdf", "docx"])

if file1 and file2:
    try:
        # Detect file type and extract text for file 1
        if file1.name.endswith('.pdf'):
            text1 = extract_text_from_pdf(file1)
        elif file1.name.endswith('.docx'):
            text1 = extract_text_from_word(file1)
        else:
            st.error("Unsupported file type for file 1")

        # Detect file type and extract text for file 2
        if file2.name.endswith('.pdf'):
            text2 = extract_text_from_pdf(file2)
        elif file2.name.endswith('.docx'):
            text2 = extract_text_from_word(file2)
        else:
            st.error("Unsupported file type for file 2")

        # Summarize texts
        #st.markdown("### Summarizing the uploaded documents...")
        #summary1 = summarize_text(text1)
        #summary2 = summarize_text(text2)

        #st.markdown("### Results for File 1 (Resume)")
        #st.subheader("Summary:")
        #st.write(summary1)

        #st.markdown("### Results for File 2 (Job Description)")
        #st.subheader("Summary:")
        #st.write(summary2)

        # Ensure session state variables are initialized
        if "keywords_extracted" not in st.session_state:
            st.session_state.keywords_extracted = False
        if "ats_score_calculated" not in st.session_state:
            st.session_state.ats_score_calculated = False

        # Button to Extract Keywords
        if st.button("Extract Keywords") or st.session_state.keywords_extracted:
            st.session_state.keywords_extracted = True

            # Extract keywords
            st.markdown("### Extracting keywords...")
            keywords1 = extract_keywords(text1)
            keywords2 = extract_keywords(text2)

            # Display Keywords
            st.markdown("### Results for File 1 (Resume)")
            st.subheader("Keywords:")
            st.write(", ".join(keywords1))

            st.markdown("### Results for File 2 (Job Description)")
            st.subheader("Keywords:")
            st.write(", ".join(keywords2))

            # Button to Calculate ATS Score
            if st.button("ATS Score") or st.session_state.ats_score_calculated:
                
                st.session_state.ats_score_calculated = True
                resume_keywords = set(keywords1)
                job_description_keywords = set(keywords2)
                st.markdown("### ATS Score Calculation")
                query = {"human_input": f"""

"Your task is to act as a highly advanced Applicant Tracking System (ATS) that evaluates the compatibility of a candidate's resume with a given job description. You will meticulously extract and analyze all relevant keywords and information from both the resume and the job description, including but not limited to Role-Specific Keywords, Technical Skills, Certifications, Experience, Soft Skills, Job Responsibilities, Industry Keywords, Methodologies and Practices, Keywords Indicating Preferences, and Core Values.

You will then calculate an ATS score on a scale of 0-100, reflecting how well the resume matches the job description. The score should be based on the following criteria:

Keywords Matching (20%): The extent to which the resume contains the exact keywords and phrases mentioned in the job description.

Skills and Competencies (20%): The presence and relevance of skills and competencies that align with the job requirements.

Formatting (10%): The clarity and simplicity of the resume format, ensuring that the ATS can easily parse the information.

Job Title Match (10%): The similarity between the candidate's previous job titles and the job title in the description.

Experience and Education (20%): Whether the candidate's experience level and education meet the job requirements.

Customization (20%): How well the resume is tailored to the specific job description, including the use of industry-specific language and terminology.

For each criterion, provide a detailed breakdown of the match percentage, highlighting where the candidate meets the requirements and where there are gaps. Finally, provide an overall ATS score and a summary of the candidate's strengths and areas for improvement.

Ensure that the evaluation is done in real-time and with 100% accuracy, taking into account all possible factors that a traditional ATS would consider."


Job Description Keywords:
{list(job_description_keywords)}

Resume Keywords:
{list(resume_keywords)}

"""}

                response = chain.invoke(query)
                memory.save_context(query, {"output": response})

                st.write(response)

            
    except Exception as e:
        st.error(f"An error occurred: {e}")

else:
    st.info("Please upload both files to proceed.")