app.py

import streamlit as st
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate
from langchain.llms import HuggingFaceHub
import fitz
from PIL import Image
import os
import pytesseract
import re

# Set Hugging Face API Key
os.environ["HUGGINGFACEHUB_API_TOKEN"] = st.secrets["HF_TOKEN"]

# Initialize LLM
llm = HuggingFaceHub(repo_id="mistralai/Mistral-7B-Instruct-v0.3", model_kwargs={"temperature": 0.5})

# App Configuration
st.set_page_config(page_title="DocuMentorAI", layout="wide", page_icon="📄")
st.title("📄 DocuMentorAI")

# Improved CSS
st.markdown("""
<style>
    .output-container {
        background-color: #f0f2f6;
        padding: 20px;
        border-radius: 10px;
        margin-top: 20px;
        white-space: pre-wrap;
    }
    .stTextArea textarea {
        font-size: 16px !important;
    }
    .stButton button {
        width: 100%;
    }
</style>
""", unsafe_allow_html=True)

# Helper Functions
def extract_text_from_pdf(pdf_file):
    try:
        pdf_bytes = pdf_file.read()
        with fitz.open(stream=pdf_bytes, filetype="pdf") as doc:
            return " ".join([page.get_text() for page in doc])
    except Exception as e:
        st.error(f"Error extracting text from PDF: {e}")
        return ""

def extract_text_from_image(image_file):
    try:
        image = Image.open(image_file)
        return pytesseract.image_to_string(image)
    except Exception as e:
        st.error(f"Error extracting text from image: {e}")
        return ""

def extract_text(uploaded_file):
    if not uploaded_file:
        return ""
    return extract_text_from_pdf(uploaded_file) if uploaded_file.type == "application/pdf" else extract_text_from_image(uploaded_file)
def parse_resume(resume_text):
    """Extract key information from resume text"""
    parsed_info = {
        'education': [],
        'skills': [],
        'experience': [],
        'projects': [],
        'publications': []
    }
    
    # Find education details
    edu_markers = ['Education:', 'EDUCATION', 'Academic Background']
    exp_markers = ['Experience:', 'EXPERIENCE', 'Work History', 'Employment']
    skill_markers = ['Skills:', 'SKILLS', 'Technical Skills', 'Technologies']
    proj_markers = ['Projects:', 'PROJECTS', 'Key Projects']
    pub_markers = ['Publications:', 'PUBLICATIONS', 'Research Papers']
    
    # Helper function to extract section content
    def extract_section(text, start_markers, end_markers):
        content = []
        for start in start_markers:
            start_idx = text.find(start)
            if start_idx != -1:
                section_start = start_idx + len(start)
                section_end = len(text)
                
                # Find the next section marker
                for end in end_markers:
                    next_section = text.find(end, section_start)
                    if next_section != -1:
                        section_end = min(section_end, next_section)
                
                section_content = text[section_start:section_end].strip()
                content.append(section_content)
        
        return '\n'.join(content)
    
    # Extract sections
    all_markers = edu_markers + exp_markers + skill_markers + proj_markers + pub_markers
    
    parsed_info['education'] = extract_section(resume_text, edu_markers, all_markers)
    parsed_info['experience'] = extract_section(resume_text, exp_markers, all_markers)
    parsed_info['skills'] = extract_section(resume_text, skill_markers, all_markers)
    parsed_info['projects'] = extract_section(resume_text, proj_markers, all_markers)
    parsed_info['publications'] = extract_section(resume_text, pub_markers, all_markers)
    
    return parsed_info

def extract_professor_details(text):
    professor_pattern = r"(Dr\.|Professor|Prof\.?)\s+([A-Z][a-z]+\s[A-Z][a-z]+)"
    university_pattern = r"(University|Institute|College|School of [A-Z][A-Za-z\s]+)"
    
    professor_match = re.search(professor_pattern, text)
    university_match = re.search(university_pattern, text)
    
    return (professor_match.group(0) if professor_match else "Not Found",
            university_match.group(0) if university_match else "Not Found")

def clean_email_output(email_text):
    """Clean and format email content"""
    start_idx = email_text.find("Dear")
    if start_idx == -1:
        start_idx = 0
        
    end_markers = ["Best regards,", "Sincerely,", "Yours sincerely,", "Kind regards,"]
    end_idx = len(email_text)
    for marker in end_markers:
        idx = email_text.find(marker)
        if idx != -1:
            end_idx = email_text.find("\n\n", idx) if email_text.find("\n\n", idx) != -1 else len(email_text)
            break
    
    email_content = email_text[start_idx:end_idx].strip()
    
    if "Phone:" in email_text or "Email:" in email_text:
        contact_info = "\n\n" + "\n".join([
            line for line in email_text[end_idx:].split("\n")
            if any(info in line for info in ["Phone:", "Email:"])
        ]).strip()
        email_content += contact_info
    
    return email_content

def clean_cover_letter_output(letter_text):
    """Clean and format cover letter content"""
    start_markers = ["Dear", "To Whom", "Hiring"]
    start_idx = len(letter_text)
    for marker in start_markers:
        idx = letter_text.find(marker)
        if idx != -1:
            start_idx = min(start_idx, idx)
    
    end_markers = ["Sincerely,", "Best regards,", "Yours truly,", "Regards,"]
    end_idx = len(letter_text)
    for marker in end_markers:
        idx = letter_text.find(marker)
        if idx != -1:
            end_idx = letter_text.find("\n\n", idx) if letter_text.find("\n\n", idx) != -1 else len(letter_text)
            break
    
    return letter_text[start_idx:end_idx].strip()

def clean_research_statement_output(statement_text):
    """Clean and format research statement content"""
    # Remove common headers
    headers = ["Research Statement", "Statement of Research", "Research Interests"]
    cleaned_text = statement_text
    for header in headers:
        if cleaned_text.startswith(header):
            cleaned_text = cleaned_text[len(header):].lstrip(":\n")
    
    # Remove any trailing references or bibliography sections
    end_markers = ["References", "Bibliography", "Citations"]
    for marker in end_markers:
        idx = cleaned_text.find(marker)
        if idx != -1:
            cleaned_text = cleaned_text[:idx].strip()
    
    return cleaned_text.strip()

def clean_sop_output(sop_text):
    """Clean and format Statement of Purpose content"""
    # Remove common headers
    headers = ["Statement of Purpose", "Personal Statement", "Academic Statement"]
    cleaned_text = sop_text
    for header in headers:
        if cleaned_text.startswith(header):
            cleaned_text = cleaned_text[len(header):].lstrip(":\n")
    
    # Remove any trailing sections
    end_markers = ["Thank you", "References", "Additional Information"]
    for marker in end_markers:
        idx = cleaned_text.find(marker)
        if idx != -1:
            cleaned_text = cleaned_text[:idx].strip()
    
    return cleaned_text.strip()

# Initialize session state
if 'generated_content' not in st.session_state:
    st.session_state.generated_content = {
        'email': None,
        'cover_letter': None,
        'research_statement': None,
        'sop': None
    }

# Template Definitions
templates = {
    'email': PromptTemplate.from_template("""
Write ONLY a formal cold email for a research position.
Start with 'Dear Professor' and end with a signature.

Use these specific details from the CV:
Education: {education}
Relevant Experience: {experience}
Key Skills: {skills}
Notable Projects: {projects}
Publications: {publications}

Additional Context:
Professor: {professor_name}
University: {university_name}
Research Interests: {research_interests}
Why This Lab: {reason}

Guidelines:
1. Keep the email concise (max 400 words)
2. Focus on the most relevant experience and skills that match the lab's research
3. Mention 1-2 specific projects or publications that align with the lab's work
4. Include a clear statement of interest and why you're a good fit
5. End with your contact information
"""),
    
    'cover_letter': PromptTemplate.from_template("""
Write ONLY a professional cover letter.
Use these specific details from the CV:
Education: {education}
Relevant Experience: {experience}
Technical Skills: {skills}
Notable Projects: {projects}
Publications: {publications}

Position Details:
Job Title: {job_title}
Company: {company}
Required Skills: {key_skills}

Guidelines:
1. Start with a formal greeting
2. Focus on experiences and skills that directly match the job requirements
3. Provide specific examples from your projects and work history
4. Demonstrate how your background makes you an ideal candidate
5. End with a professional closing
"""),
    
    'research_statement': PromptTemplate.from_template("""
Write ONLY a research statement focused on your academic journey and research goals.
Use these specific details from your background:
Education: {education}
Research Experience: {experience}
Technical Skills: {skills}
Research Projects: {projects}
Publications: {publications}

Additional Context:
Research Background: {research_background}
Key Projects: {key_projects}
Future Goals: {future_goals}

Guidelines:
1. Describe your research journey and motivation
2. Highlight key research achievements and findings
3. Connect past work to future research goals
4. Demonstrate technical expertise and methodological knowledge
5. End with your vision for future contributions to the field
"""),
    
    'sop': PromptTemplate.from_template("""
Write ONLY a Statement of Purpose (SOP).
Use these specific details from your background:
Education: {education}
Research Experience: {experience}
Technical Skills: {skills}
Notable Projects: {projects}
Publications: {publications}

Additional Context:
Motivation: {motivation}
Academic Goals: {academic_background}
Research Interests: {research_experiences}
Career Objectives: {career_goals}
Program Interest: {why_this_program}

Guidelines:
1. Tell a coherent story about your academic journey
2. Connect your background to your future goals
3. Demonstrate why you're prepared for graduate study
4. Show alignment between your interests and the program
5. Make a compelling case for why you should be admitted
""")
}


# Create LangChain instances
chains = {key: LLMChain(llm=llm, prompt=template) for key, template in templates.items()}

# Sidebar for Input Collection
with st.sidebar:
    st.subheader("📝 Input Details")
    job_opening_text = st.text_area("Job/Research Opening Details", height=150)
    cv_resume_file = st.file_uploader("Upload CV/Resume", type=["pdf", "png", "jpg", "jpeg"])
    cv_resume_text = extract_text(cv_resume_file)

# Tab Layout
tab1, tab2, tab3, tab4 = st.tabs(["Cold Email", "Cover Letter", "Research Statement", "SOP"])

# Cold Email Tab
with tab1:
    professor_name, university_name = extract_professor_details(job_opening_text)
    research_interests = st.text_input("Research Interests")
    reason = st.text_input("Why this professor/lab?")
    
    if st.button("Generate Email", key="email_btn"):
        if job_opening_text and cv_resume_text:
            with st.spinner("Generating..."):
                try:
                    # Parse resume information
                    resume_info = parse_resume(cv_resume_text)
                    
                    # Generate email with parsed information
                    generated_email = chains['email'].run({
                        "professor_name": professor_name,
                        "university_name": university_name,
                        "research_interests": research_interests,
                        "reason": reason,
                        "education": resume_info['education'],
                        "experience": resume_info['experience'],
                        "skills": resume_info['skills'],
                        "projects": resume_info['projects'],
                        "publications": resume_info['publications']
                    })
                    st.session_state.generated_content['email'] = clean_email_output(generated_email)
                except Exception as e:
                    st.error(f"Generation error: {e}")
        else:
            st.error("Please provide all required inputs")

# Cover Letter Tab
with tab2:
    job_title = st.text_input("Job Title")
    company_name = university_name if university_name != "Not Found" else st.text_input("Company/University")
    key_skills = st.text_input("Key Skills")
    
    if st.button("Generate Cover Letter", key="cover_letter_btn"):
        if job_opening_text and cv_resume_text:
            with st.spinner("Generating..."):
                try:
                    resume_info = parse_resume(cv_resume_text)
                    generated_letter = chains['cover_letter'].run({
                        "job_title": job_title,
                        "company": company_name,
                        "key_skills": key_skills,
                        "reason": reason,
                        "skills": resume_info['skills'],
                        "education": resume_info['education'],
                        "experience": resume_info['experience']
                    })
                    st.session_state.generated_content['cover_letter'] = clean_cover_letter_output(generated_letter)
                except Exception as e:
                    st.error(f"Generation error: {e}")
        else:
            st.error("Please provide all required inputs")

    if st.session_state.generated_content['cover_letter']:
        st.markdown('<div class="output-container">', unsafe_allow_html=True)
        st.markdown(st.session_state.generated_content['cover_letter'])
        st.download_button("Download Cover Letter", st.session_state.generated_content['cover_letter'],
                         file_name="cover_letter.txt", key="cover_letter_download")
        st.markdown('</div>', unsafe_allow_html=True)

# Research Statement Tab
with tab3:
    
    key_projects = st.text_input("Key Research Projects")
    future_goals = st.text_input("Future Research Goals")
    
    if st.button("Generate Research Statement", key="research_stmt_btn"):
        with st.spinner("Generating..."):
            try:
                resume_info = parse_resume(cv_resume_text)
                generated_statement = chains['research_statement'].run({
                    "reason": reason,
                    "education": resume_info['education'],
                    "experience": resume_info['experience'],
                    "skills": resume_info['skills'],
                    "projects": resume_info['projects'],
                    "publications": resume_info['publications'],
                    "research_background": resume_info['publications'],
                    "key_projects": key_projects,
                    "future_goals": future_goals
                })
                st.session_state.generated_content['research_statement'] = clean_research_statement_output(generated_statement)
            except Exception as e:
                st.error(f"Generation error: {e}")

    if st.session_state.generated_content['research_statement']:
        st.markdown('<div class="output-container">', unsafe_allow_html=True)
        st.markdown(st.session_state.generated_content['research_statement'])
        st.download_button("Download Research Statement", st.session_state.generated_content['research_statement'],
                         file_name="research_statement.txt", key="research_stmt_download")
        st.markdown('</div>', unsafe_allow_html=True)

# SOP Tab
with tab4:
    motivation = st.text_input("Motivation for Graduate Studies")
    
    
    career_goals = st.text_input("Career Goals")
    why_this_program = st.text_input("Why This Program")
    
    if st.button("Generate SOP", key="sop_btn"):
        with st.spinner("Generating..."):
            try:
                resume_info = parse_resume(cv_resume_text)
                generated_sop = chains['sop'].run({
                    "motivation": motivation,
                    "academic_background": resume_info['education'],
                    "research_experiences": resume_info['publications'],
                    "career_goals": career_goals,
                    "why_this_program": why_this_program,
                    "experience": resume_info['experience'],
                    "skills": resume_info['skills'],
                    "projects": resume_info['projects']

                })
                st.session_state.generated_content['sop'] = clean_sop_output(generated_sop)
            except Exception as e:
                st.error(f"Generation error: {e}")

    if st.session_state.generated_content['sop']:
        st.markdown('<div class="output-container">', unsafe_allow_html=True)
        st.markdown(st.session_state.generated_content['sop'])
        st.download_button("Download SOP", st.session_state.generated_content['sop'],
                         file_name="sop.txt", key="sop_download")
        st.markdown('</div>', unsafe_allow_html=True)

# Reset Button
if st.sidebar.button("🔄 Reset All"):
    st.session_state.generated_content = {key: None for key in st.session_state.generated_content}
    st.experimental_rerun()