app.py

import streamlit as st
import streamlit.components.v1 as components
import os
import tempfile
import json
import textwrap
import re
import ast
from typing import Optional
from pathlib import Path
import asyncio
import requests

# API and instructor imports
import instructor
from google import genai
import anthropic
from openai import AsyncOpenAI

# Project imports
from resumer import ResumeTailorPipeline
from resumer.utils.latex_ops import json_to_latex_pdf
from streamlit_pdf_viewer import pdf_viewer

# ============================================
# PAGE CONFIGURATION
# ============================================

st.set_page_config(
    page_title="Resume Tailor AI",
    page_icon="📄",
    layout="wide",
    initial_sidebar_state="expanded"
)

st.markdown("""
    <style>
    .main { padding-top: 1rem; }
    .stTabs [data-baseweb="tab-list"] button { font-size: 1.1em; }
    </style>
    """, unsafe_allow_html=True)

# ============================================
# MODEL CONFIGURATIONS
# ============================================

MODELS = {
    "Gemini": [
        "gemini-3-flash-preview",
        "gemini-3-pro-image-preview",
        "gemini-2.5-pro",
        "gemini-2.5-flash",
        "gemini-2.5-flash-lite"
    ],
    "Claude": [
        "claude-sonnet-4-5",
        "claude-haiku-4-5",
        "claude-opus-4-5",
    ],
    "OpenAI": [
        "gpt-5-mini",
        "gpt-5-nano",
        "gpt-4o-mini",
        "gpt-4o",
    ]
}

# ============================================
# SESSION STATE INITIALIZATION
# ============================================

def init_session_state():
    defaults = {
        "authenticated": False,
        "api_provider": None,
        "selected_model": None,
        "api_key": None,
        "resume_file": None,
        "resume_path": None,
        "resume_bytes": None,
        "job_url": None,
        "job_text": None,
        "pipeline": None,
        "tailored_resume_path": None,
        "tailored_resume_pdf": None,
        "tailored_resume_tex": None,
        "tailored_resume_json": None,
        "processing_log": [],
    }
    for key, value in defaults.items():
        if key not in st.session_state:
            st.session_state[key] = value

init_session_state()

# ============================================
# API CLIENT INITIALIZATION
# ============================================

def get_gemini_instructor_client(api_key: str):
    """Initialize Instructor-patched Gemini client"""
    native_client = genai.Client(api_key=api_key)
    aclient = instructor.from_genai(
        native_client,
        mode=instructor.Mode.GENAI_TOOLS,
        use_async=True
    )
    return aclient

def get_claude_instructor_client(api_key: str):
    """Initialize Instructor-patched Claude client"""
    native_client = anthropic.Anthropic(api_key=api_key)
    aclient = instructor.from_anthropic(
        native_client,
        mode=instructor.Mode.TOOLS,
    )
    return aclient

def get_openai_instructor_client(api_key: str):
    """Initialize Instructor-patched OpenAI client"""
    native_client = AsyncOpenAI(api_key=api_key)
    aclient = instructor.from_openai(
        native_client,
        mode=instructor.Mode.TOOLS,
    )
    return aclient

# ============================================
# UTILITY FUNCTIONS
# ============================================
import base64

import base64

def mermaid_chart(code: str, height: int = 600):
    """
    Renders Mermaid.js diagrams in Streamlit by fetching SVG from mermaid.ink.
    Saves the SVG locally and displays it.
    """
    # Clean up code
    code = textwrap.dedent(code).strip()
    
    # Encode to base64
    graphbytes = code.encode("utf8")
    base64_bytes = base64.urlsafe_b64encode(graphbytes)
    base64_string = base64_bytes.decode("ascii")
    
    # Construct URL
    url = f"https://mermaid.ink/svg/{base64_string}"
    
    try:
        # Fetch the SVG
        response = requests.get(url)
        if response.status_code == 200:
            # Display as image
            st.image(response.text, width="stretch")
        else:
            # Fallback: Try without the init block
            import re
            code_no_init = re.sub(r'%%\{init:.*?\}%%', '', code, flags=re.DOTALL).strip()
            graphbytes_fallback = code_no_init.encode("utf8")
            base64_bytes_fallback = base64.urlsafe_b64encode(graphbytes_fallback)
            base64_string_fallback = base64_bytes_fallback.decode("ascii")
            url_fallback = f"https://mermaid.ink/svg/{base64_string_fallback}"
            
            response_fallback = requests.get(url_fallback)
            if response_fallback.status_code == 200:
                 st.image(response_fallback.text, width="stretch")
            else:
                st.error(f"Failed to render diagram (Status: {response.status_code})")
                st.code(code, language="mermaid")
    except Exception as e:
        st.error(f"Error rendering diagram: {str(e)}")
        st.code(code, language="mermaid")

def log_message(message: str):
    """Add message to processing log"""
    st.session_state.processing_log.append(message)

def save_uploaded_file(uploaded_file) -> str:
    """Save uploaded file to temporary location and store bytes"""
    # Read the file bytes first
    file_bytes = uploaded_file.getvalue()
    st.session_state.resume_bytes = file_bytes
    
    # Save to temporary location
    with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp:
        tmp.write(file_bytes)
        return tmp.name

async def run_pipeline(
    aclient,
    model_name: str,
    resume_path: str,
    job_url: Optional[str] = None,
    job_text: Optional[str] = None,
    progress_callback=None
) -> Optional[tuple]:
    """Run the ResumeTailorPipeline asynchronously"""
    try:
        if progress_callback:
            progress_callback("📖 Initializing pipeline...")
        
        with tempfile.TemporaryDirectory() as tmpdir:
            pipeline = ResumeTailorPipeline(
                aclient=aclient,
                model_name=model_name,
                resume_path=resume_path,
                output_dir=tmpdir,
                log_callback=progress_callback
            )
            
            # Store pipeline in session state
            st.session_state.pipeline = pipeline
            
            # Generate tailored resume asynchronously
            result = await pipeline.generate_tailored_resume(
                job_url=job_url,
                job_site_content=job_text
            )
            
            # Result is now a tuple: (pdf_path, tex_path)
            if isinstance(result, tuple):
                tailored_pdf_path, tailored_tex_path = result
            else:
                tailored_pdf_path = result
                tailored_tex_path = None
            
            if progress_callback:
                progress_callback("💾 Reading generated files...")
            
            # Read the PDF and store in session state
            if tailored_pdf_path and os.path.exists(tailored_pdf_path):
                with open(tailored_pdf_path, "rb") as f:
                    st.session_state.tailored_resume_pdf = f.read()
            
            # Read the TEX file and store in session state
            if tailored_tex_path and os.path.exists(tailored_tex_path):
                with open(tailored_tex_path, "r", encoding="utf-8") as f:
                    st.session_state.tailored_resume_tex = f.read()
            
            # Also store the JSON details
            st.session_state.tailored_resume_json = pipeline.resume_details
            
            if progress_callback:
                progress_callback("✅ Cleanup and finalization...")
            
            pipeline.close_cache()
            return (tailored_pdf_path, tailored_tex_path)
            
    except Exception as e:
        if progress_callback:
            progress_callback(f"❌ Error: {str(e)}")
        st.error(f"Pipeline Error: {str(e)}")
        import traceback
        st.error(traceback.format_exc())
        return None

# ============================================
# MAIN APP UI
# ============================================

def main():
    # Header
    col1, col2 = st.columns([0.7, 0.3])
    with col1:
        st.title("📄 ResFit: Resume Tailor AI")
        st.markdown("*Tailor your resume for any job using AI - **Preserving your Links!***")
        st.info("💡 **Why ResFit?** Unlike other tools, this app preserves all hyperlinks in your resume while tailoring the content.")
        
        with st.expander("🔄 How ResFit Works"):
            # Read flowchart from file
            flowchart_path = Path(__file__).parent / "docs" / "flowchart.mmd"
            if flowchart_path.exists():
                with open(flowchart_path, "r") as f:
                    flowchart_code = f.read()
                mermaid_chart(flowchart_code, height=800)
            else:
                st.error(f"Flowchart definition not found at {flowchart_path}")
    
    # ========== SIDEBAR: AUTHENTICATION ==========
    with st.sidebar:
        st.header("🔐 Authentication")
        
        # Step 1: Select Provider
        api_provider = st.radio(
            "Step 1: Select API Provider",
            ["Gemini", "Claude", "OpenAI"],
            key="provider_select"
        )
        st.session_state.api_provider = api_provider
        
        # Step 2: Select Model based on provider
        available_models = MODELS[api_provider]
        selected_model = st.selectbox(
            "Step 2: Select Model",
            available_models,
            key=f"model_select_{api_provider}",
            index=0
        )
        st.session_state.selected_model = selected_model
        
        # Display model info
        model_info = {
            "Gemini": {
                "gemini-3-flash-preview": "⚡ Fastest, latest (recommended)",
                "gemini-3-pro-image-preview": "🖼️ Vision capabilities, advanced",
                "gemini-2.5-pro": "💪 Most capable but slower",
                "gemini-2.5-flash": "⚡ Fast & capable",
                "gemini-2.5-flash-lite": "💨 Fastest, most affordable",
            },
            "Claude": {
                "claude-sonnet-4-5": "⚡ Latest Sonnet (recommended)",
                "claude-haiku-4-5": "💨 Fastest, most affordable",
                "claude-opus-4-5": "💪 Most capable but slower",
            },
            "OpenAI": {
                "gpt-5-mini": "⚡ Latest & fastest (recommended)",
                "gpt-5-nano": "💨 Most affordable",
                "gpt-4o-mini": "💪 Good balance",
                "gpt-4o": "🦾 Most capable",
            }
        }
        
        if selected_model in model_info.get(api_provider, {}):
            st.caption(f"ℹ️ {model_info[api_provider][selected_model]}")
        
        st.divider()
        
        # Step 3: Enter API Key
        api_key = st.text_input(
            "Step 3: Enter API Key",
            type="password",
            key="api_key_input",
            help=f"Your {api_provider} API key will not be stored"
        )
        
        st.divider()
        
        # Authenticate button
        if st.button("🔓 Authenticate", width="stretch", type="primary"):
            if api_key:
                try:
                    if api_provider == "Gemini":
                        aclient = get_gemini_instructor_client(api_key)
                    elif api_provider == "Claude":
                        aclient = get_claude_instructor_client(api_key)
                    else:  # OpenAI
                        aclient = get_openai_instructor_client(api_key)
                    
                    st.session_state.authenticated = True
                    st.session_state.api_key = api_key
                    st.session_state.aclient = aclient
                    st.success(f"✅ Authenticated!\n\n**Provider:** {api_provider}\n**Model:** {selected_model}")
                except Exception as e:
                    st.error(f"❌ Authentication failed: {str(e)}")
            else:
                st.error("Please enter an API key")
        
        st.divider()
        
        # Display current auth status
        if st.session_state.authenticated:
            st.info(f"""
            ✅ **Authenticated**
            
            **Provider:** {st.session_state.api_provider}
            **Model:** {st.session_state.selected_model}
            """)
            
            if st.button("🚪 Logout", width="stretch"):
                st.session_state.authenticated = False
                st.session_state.api_key = None
                st.session_state.api_provider = None
                st.session_state.selected_model = None
                st.session_state.aclient = None
                st.rerun()


        st.markdown("[![GitHub](https://img.shields.io/badge/GitHub-ResFit-181717?logo=github)](https://github.com/AwaleSajil/resfit)")
    
    # ========== MAIN CONTENT ==========
    if not st.session_state.authenticated:
        st.warning("⚠️ Please authenticate with an API provider in the sidebar to continue")
        st.info("""
        **How to get an API key:**
        
        🔵 **Gemini**: Free API key at [https://aistudio.google.com/app/apikey](https://aistudio.google.com/app/apikey)
        
        🔴 **Claude**: API key at [https://console.anthropic.com/](https://console.anthropic.com/)
        
        🟢 **OpenAI**: API key at [https://platform.openai.com/api-keys](https://platform.openai.com/api-keys)
        """)
        return
    
    # Main tabs
    tab1, tab2, tab3 = st.tabs(["📤 Upload", "⚙️ Process", "📊 Results"])
    
    # ========== TAB 1: UPLOAD ==========
    with tab1:
        st.header("Upload Your Materials")
        
        col1, col2 = st.columns(2)
        
        with col1:
            st.subheader("📄 Resume PDF")
            resume_file = st.file_uploader(
                "Select your resume (PDF only)",
                type=["pdf"],
                key="resume_uploader"
            )
            
            if resume_file:
                # Save to temporary location
                resume_path = save_uploaded_file(resume_file)
                st.session_state.resume_file = resume_file
                st.session_state.resume_path = resume_path
                st.success(f"✅ Uploaded: {resume_file.name}")
                st.info(f"📊 Size: {resume_file.size / 1024:.1f} KB")
        
        with col2:
            st.subheader("🎯 Job Description")
            
            job_source = st.radio(
                "Provide job description via:",
                ["📎 URL", "📝 Text"],
                horizontal=False,
                key="job_source_select"
            )
            
            if job_source == "📎 URL":
                job_url = st.text_input(
                    "Paste job posting URL:",
                    placeholder="https://careers.example.com/job/123",
                    key="job_url_input"
                )
                if job_url:
                    st.session_state.job_url = job_url
                    st.session_state.job_text = None
                    st.success("✅ URL saved")
                
            else:  # Text
                job_text = st.text_area(
                    "Paste job description text:",
                    placeholder="Paste the complete job description here...",
                    height=200,
                    key="job_text_input"
                )
                if job_text:
                    st.session_state.job_text = job_text
                    st.session_state.job_url = None
                    st.success("✅ Job description saved")
        
        st.divider()
        
        # Summary
        st.subheader("📋 Upload Summary")
        summary_col1, summary_col2 = st.columns(2)
        
        with summary_col1:
            if st.session_state.resume_path:
                st.metric("Resume", "✅ Ready")
            else:
                st.metric("Resume", "⏳ Waiting")
        
        with summary_col2:
            if st.session_state.job_url or st.session_state.job_text:
                st.metric("Job Description", "✅ Ready")
            else:
                st.metric("Job Description", "⏳ Waiting")
    
    # ========== TAB 2: PROCESS ==========
    with tab2:
        st.header("Process Your Resume")
        
        # Validation
        if not st.session_state.resume_path:
            st.error("❌ Please upload a resume in the Upload tab")
            return
        
        if not st.session_state.job_url and not st.session_state.job_text:
            st.error("❌ Please provide a job description in the Upload tab")
            return
        
        st.info(f"""
        **Processing Configuration:**
        - **Provider:** {st.session_state.api_provider}
        - **Model:** {st.session_state.selected_model}
        
        **This process will:**
        1. Extract your resume structure asynchronously
        2. Extract job requirements asynchronously
        3. Tailor your resume to match the job
        4. Generate a PDF with the tailored version
        """)
        
        st.divider()
        
        # Start processing button
        if st.button("🚀 Generate Tailored Resume", width="stretch", type="primary", key="btn_start"):
            # Clear processing log
            st.session_state.processing_log = []
            
            # Create a single placeholder for live log display
            log_placeholder = st.empty()
            
            def update_progress(message: str):
                """Callback to update progress"""
                # Add message to log
                st.session_state.processing_log.append(message)
                
                # Keep only the latest x logs
                max_logs = 5
                if len(st.session_state.processing_log) > max_logs:
                    latest_logs = st.session_state.processing_log[-max_logs:]
                else:
                    latest_logs = st.session_state.processing_log
                
                # Update the placeholder with latest logs (no duplicates)
                with log_placeholder.container():
                    st.subheader(f"📝 Live Processing Log (Latest {max_logs})")
                    for log in latest_logs:
                        st.write(log)
            
            try:
                update_progress("🔐 Initializing async event loop...")
                
                # Create and run async pipeline
                loop = asyncio.new_event_loop()
                asyncio.set_event_loop(loop)
                
                update_progress("⏳ Starting resume processing...")
                
                result = loop.run_until_complete(
                    run_pipeline(
                        aclient=st.session_state.aclient,
                        model_name=st.session_state.selected_model,
                        resume_path=st.session_state.resume_path,
                        job_url=st.session_state.job_url,
                        job_text=st.session_state.job_text,
                        progress_callback=update_progress
                    )
                )
                
                loop.close()
                
                if result:
                    st.session_state.tailored_resume_path = result
                    st.divider()
                    st.success("✅ Resume tailored successfully!")
                    st.balloons()
                else:
                    st.divider()
                    st.error("❌ Failed to generate tailored resume")
                    
            except Exception as e:
                st.divider()
                st.error(f"❌ Error: {str(e)}")
        
        # Display full processing log history (after processing)
        if st.session_state.processing_log:
            st.divider()
            st.subheader("📋 Full Processing Log")
            with st.expander("View all logs", expanded=False):
                for log in st.session_state.processing_log:
                    st.write(log)
    
    # ========== TAB 3: RESULTS ==========
    with tab3:
        st.header("Results")
        
        if not st.session_state.tailored_resume_path:
            st.info("👈 Complete the processing in the Process tab to see results here")
            return
        
        st.success("✅ Your tailored resume is ready!")
        
        # Download options
        st.subheader("📥 Download Your Resumes")
        
        col1, col2, col3 = st.columns(3)
        
        with col1:
            st.markdown("#### Original Resume")
            if st.session_state.resume_bytes:
                st.download_button(
                    label="📥 Download Original PDF",
                    data=st.session_state.resume_bytes,
                    file_name="original_resume.pdf",
                    mime="application/pdf",
                    width="stretch"
                )
        
        with col2:
            st.markdown("#### Tailored Resume (PDF)")
            if "tailored_resume_pdf" in st.session_state:
                st.download_button(
                    label="📥 Download Tailored PDF",
                    data=st.session_state.tailored_resume_pdf,
                    file_name="tailored_resume.pdf",
                    mime="application/pdf",
                    width="stretch",
                    type="primary"
                )
        
        with col3:
            st.markdown("#### Tailored Resume (LaTeX)")
            if "tailored_resume_tex" in st.session_state and st.session_state.tailored_resume_tex:
                st.download_button(
                    label="📥 Download LaTeX (.tex)",
                    data=st.session_state.tailored_resume_tex.encode('utf-8'),
                    file_name="tailored_resume.tex",
                    mime="text/plain",
                    width="stretch"
                )
            else:
                st.info("LaTeX file not available")
        
        st.divider()
        
        # PDF Preview Section using streamlit-pdf-viewer
        st.subheader("📄 PDF Preview")
        
        preview_col1, preview_col2 = st.columns(2)
        
        with preview_col1:
            with st.expander("👁️ View Original Resume PDF", expanded=True):
                if st.session_state.resume_bytes:
                    pdf_viewer(input=st.session_state.resume_bytes, width=700, height=800)
                else:
                    st.info("No original resume available")
        
        with preview_col2:
            with st.expander("✨ View Tailored Resume PDF", expanded=True):
                if "tailored_resume_pdf" in st.session_state:
                    pdf_viewer(input=st.session_state.tailored_resume_pdf, width=700, height=800)
                else:
                    st.info("No tailored resume available")
        
        st.divider()
        
        # LaTeX Source Code Viewer
        st.subheader("📝 LaTeX Source Code")
        if "tailored_resume_tex" in st.session_state and st.session_state.tailored_resume_tex:
            with st.expander("👁️ View LaTeX Source Code", expanded=False):
                st.code(st.session_state.tailored_resume_tex, language="latex")
        else:
            st.info("No LaTeX source available")
        
        st.divider()
        
        # Data comparison
        st.subheader("📊 Resume Data Comparison")
        
        if st.session_state.pipeline:
            result_col1, result_col2 = st.columns(2)
            
            with result_col1:
                with st.expander("📖 Original Resume Data", expanded=False):
                    if st.session_state.pipeline.resume_info:
                        st.json(st.session_state.pipeline.resume_info.model_dump())
                    else:
                        st.info("No data available")
            
            with result_col2:
                with st.expander("✨ Tailored Resume Data", expanded=False):
                    if "tailored_resume_json" in st.session_state:
                        st.json(st.session_state.tailored_resume_json)
                    else:
                        st.info("No data available")
        
        st.divider()
        
        # Job info display
        st.subheader("🎯 Job Requirements (Extracted)")
        if st.session_state.pipeline and st.session_state.pipeline.job_info:
            with st.expander("View job info", expanded=False):
                if hasattr(st.session_state.pipeline.job_info, 'model_dump'):
                    st.json(st.session_state.pipeline.job_info.model_dump())
                else:
                    st.json(st.session_state.pipeline.job_info)

if __name__ == "__main__":
    main()