app.py

"""
Quantum Scrutiny Platform | Groq-Powered
Single-file Streamlit app (refactored, Groq streaming-compatible)
"""

import os
import io
import re
import json
import base64
import traceback
from typing import Optional, List

from dotenv import load_dotenv
load_dotenv()

import streamlit as st
import pandas as pd

# File parsing
import fitz                   # PyMuPDF
from docx import Document     # python-docx

# Groq client
from groq import Groq

# Validation
from pydantic import BaseModel, Field, ValidationError

# --- Page config ---
st.set_page_config(layout="wide", page_title="Quantum Scrutiny Platform | Groq-Powered")

# --- Config / Secrets ---
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
ADMIN_PASSWORD = os.getenv("ADMIN_PASSWORD", "admin")

# Initialize Groq client (no API key -> UI warning but app still loads)
groq_client = None
if GROQ_API_KEY:
    try:
        groq_client = Groq(api_key=GROQ_API_KEY)
    except Exception as e:
        st.error(f"Failed to initialize Groq client: {e}")
else:
    st.warning("GROQ_API_KEY not found. Set it as an environment variable or in .env for model calls to work.")

# --- Session state defaults ---
if 'is_admin_logged_in' not in st.session_state:
    st.session_state.is_admin_logged_in = False
if 'analyzed_data' not in st.session_state:
    initial_cols = [
        'Name', 'Job Role', 'Resume Score (100)', 'Shortlisted', 'Email', 'Phone',
        # NEW SCORE COLUMNS
        'Experience Score (40)', 'Skills Score (30)', 'Communication Score (20)', 'Certifications Score (10)',
        'Experience Summary', 'Education Summary', 'Communication Rating (1-10)',
        'Skills/Technologies', 'Certifications',
        # THERAPIST FIELDS
        'ABA Skills (1-10)', 'RBT/BCBA Cert', 'Autism-Care Exp (1-10)'
    ]
    st.session_state.analyzed_data = pd.DataFrame(columns=initial_cols)
if 'individual_analysis' not in st.session_state:
    st.session_state.individual_analysis = []
if 'run_analysis' not in st.session_state:
    st.session_state.run_analysis = False

# --- Pydantic schema (No change needed here, as the new scores are derived) ---
class ResumeAnalysis(BaseModel):
    name: str = Field(default="Unknown")
    email: str = Field(default="")
    phone: str = Field(default="")
    certifications: List[str] = Field(default_factory=list)
    experience_summary: str = Field(default="")
    education_summary: str = Field(default="")
    communication_skills: str = Field(default="N/A")
    technical_skills: List[str] = Field(default_factory=list)
    aba_therapy_skills: Optional[str] = Field(default="N/A")
    rbt_bcba_certification: Optional[str] = Field(default="N/A")
    autism_care_experience_score: Optional[str] = Field(default="N/A")

# --- Helpers: file text extraction (No change) ---
def extract_text_from_file(uploaded_file) -> str:
    """Extract text from PDF or DOCX. Returns empty string on failure."""
    try:
        content = uploaded_file.read()
        filename = uploaded_file.name.lower()
        if filename.endswith(".pdf") or content[:5] == b"%PDF-":
            try:
                with fitz.open(stream=content, filetype="pdf") as doc:
                    text = ""
                    for p in doc:
                        text += p.get_text()
                return text.strip()
            except Exception:
                return ""
        elif filename.endswith(".docx"):
            try:
                doc = Document(io.BytesIO(content))
                paragraphs = [p.text for p in doc.paragraphs if p.text and p.text.strip()]
                return "\n".join(paragraphs).strip()
            except Exception:
                return ""
        else:
            # fallback: decode bytes as text
            try:
                return content.decode('utf-8', errors='ignore')
            except Exception:
                return ""
    except Exception:
        return ""

# --- Groq call with streaming (collects chunks) (No change) ---
def call_groq_stream_collect(prompt: str, model_name: str = "llama-3.3-70b-versatile", temperature: float = 0.2, max_completion_tokens: int = 2048, top_p: float = 1.0) -> Optional[str]:
    """
    Calls Groq with streaming enabled and collects the textual output.
    Returns the full model text, or None on failure.
    """
    if not groq_client:
        st.error("Groq client not initialized. Set GROQ_API_KEY in environment/secrets.")
        return None

    try:
        completion = groq_client.chat.completions.create(
            model=model_name,
            messages=[
                {"role": "system", "content": "You are a professional Resume Analyzer. Return JSON only when asked."},
                {"role": "user", "content": prompt}
            ],
            temperature=temperature,
            max_completion_tokens=max_completion_tokens,
            top_p=top_p,
            stream=True
        )

        # completion is an iterator/streamable object; collect chunks
        collected = ""
        # some SDKs yield dict-like chunks, some objects; handle both
        for chunk in completion:
            try:
                # Common pattern: chunk.choices[0].delta.content
                delta = getattr(chunk.choices[0].delta, "content", None) if hasattr(chunk, "choices") else None
                if delta is None:
                    # fallback for dict-like object
                    if isinstance(chunk, dict):
                        delta = chunk.get("choices", [{}])[0].get("delta", {}).get("content")
                if delta:
                    collected += delta
                else:
                    # Some SDKs return final message in chunk.choices[0].message.content
                    try:
                        msg = getattr(chunk.choices[0].message, "content", None)
                        if msg:
                            collected += msg
                    except Exception:
                        pass
            except Exception:
                # last-resort: append str(chunk)
                try:
                    collected += str(chunk)
                except Exception:
                    pass

        return collected.strip()
    except Exception as e:
        st.error(f"Groq API call failed: {e}")
        return None

# --- Parsing model output safely to JSON (No change) ---
def extract_first_json(text: str) -> Optional[dict]:
    """
    Find the first JSON object in text and parse it; return dict or None.
    """
    if not text:
        return None
    # find first balanced braces block
    # quick heuristic regex for {...}
    try:
        match = re.search(r"(\{(?:[^{}]|(?R))*\})", text, re.DOTALL)
    except re.error:
        # Python's re doesn't support (?R); fallback to simpler greedy
        match = re.search(r"(\{.*\})", text, re.DOTALL)
    if match:
        json_text = match.group(1)
    else:
        # maybe the model returned only JSON-like lines -> try to parse full text
        json_text = text

    try:
        parsed = json.loads(json_text)
        return parsed
    except Exception:
        # try to clean common issues: single quotes -> double quotes
        try:
            json_text_fixed = json_text.replace("'", '"')
            parsed = json.loads(json_text_fixed)
            return parsed
        except Exception:
            return None

# --- Analyze with Groq (cached by resume text + role) (No change) ---
@st.cache_data(show_spinner=False)
def analyze_resume_with_groq_cached(resume_text: str, job_role: str) -> ResumeAnalysis:
    """
    Calls Groq (streaming) and returns a ResumeAnalysis instance.
    Uses caching to avoid duplicate calls for same resume_text+role.
    """
    # Build prompt instructing JSON structure
    therapist_instructions = ""
    if job_role.lower() == "therapist":
        therapist_instructions = (
            "Because the role is 'Therapist', carefully search for ABA Therapy Skills, "
            "RBT/BCBA Certification, and Autism-Care Experience. Provide scores 1-10 as STRINGS, or 'N/A'."
        )
    else:
        therapist_instructions = "If therapist-specific fields are not relevant, set them to 'N/A'."

    system_user_prompt = (
        "Return a single JSON object with the following keys exactly: "
        "name (string), email (string), phone (string), certifications (array of strings), "
        "experience_summary (string), education_summary (string), communication_skills (STRING, e.g., '8'), "
        "technical_skills (array of strings), aba_therapy_skills (STRING or 'N/A'), "
        "rbt_bcba_certification (STRING 'Yes'/'No'/'N/A'), autism_care_experience_score (STRING or 'N/A'). "
        f"{therapist_instructions}\n\nResume Text:\n\n{resume_text}\n\nReturn only the JSON object."
    )

    raw = call_groq_stream_collect(system_user_prompt, model_name="llama-3.3-70b-versatile", temperature=0.0, max_completion_tokens=2048)

    if not raw:
        # fallback empty object
        return ResumeAnalysis(
            name="Extraction Failed",
            email="",
            phone="",
            certifications=[],
            experience_summary="",
            education_summary="",
            communication_skills="N/A",
            technical_skills=[],
            aba_therapy_skills="N/A",
            rbt_bcba_certification="N/A",
            autism_care_experience_score="N/A"
        )

    parsed = extract_first_json(raw)
    if not parsed:
        # show raw output for debugging when developer runs app locally (admin panel will show too)
        st.warning("Failed to parse model JSON output. See raw output below for debugging.")
        st.text_area("Raw model output (debug)", raw, height=200)
        return ResumeAnalysis(
            name="Extraction Failed",
            email="",
            phone="",
            certifications=[],
            experience_summary="",
            education_summary="",
            communication_skills="N/A",
            technical_skills=[],
            aba_therapy_skills="N/A",
            rbt_bcba_certification="N/A",
            autism_care_experience_score="N/A"
        )

    # Ensure keys exist and coerce types
    parsed.setdefault("name", "Unknown")
    parsed.setdefault("email", "")
    parsed.setdefault("phone", "")
    parsed.setdefault("certifications", [])
    parsed.setdefault("experience_summary", "")
    parsed.setdefault("education_summary", "")
    parsed.setdefault("communication_skills", "N/A")
    parsed.setdefault("technical_skills", [])
    parsed.setdefault("aba_therapy_skills", "N/A")
    parsed.setdefault("rbt_bcba_certification", "N/A")
    parsed.setdefault("autism_care_experience_score", "N/A")

    # Ensure string coercions for some fields
    try:
        parsed["communication_skills"] = str(parsed.get("communication_skills") or "N/A")
        parsed["aba_therapy_skills"] = str(parsed.get("aba_therapy_skills") or "N/A")
        parsed["rbt_bcba_certification"] = str(parsed.get("rbt_bcba_certification") or "N/A")
        parsed["autism_care_experience_score"] = str(parsed.get("autism_care_experience_score") or "N/A")
    except Exception:
        pass

    # Validate via Pydantic
    try:
        analysis = ResumeAnalysis.parse_obj(parsed)
        return analysis
    except ValidationError as ve:
        st.error("Model output failed schema validation.")
        st.text_area("Raw model output (debug)", raw, height=200)
        st.exception(ve)
        return ResumeAnalysis(
            name="Extraction Failed",
            email="",
            phone="",
            certifications=[],
            experience_summary="",
            education_summary="",
            communication_skills="N/A",
            technical_skills=[],
            aba_therapy_skills="N/A",
            rbt_bcba_certification="N/A",
            autism_care_experience_score="N/A"
        )

# --- Scoring logic (MODIFIED) ---
def calculate_resume_score(analysis: ResumeAnalysis, role: str) -> tuple[float, float, float, float, float]:
    """
    Calculates the overall score and the individual component scores.
    Returns (final_score, exp_score, skills_score, comm_score, certs_score)
    """
    total_score = 0.0

    # 1. Experience score: up to 40 points
    exp_len = len(analysis.experience_summary or "")
    # Cap factor at 1.0 (e.g., 100+ chars = 1.0)
    exp_factor = min(exp_len / 100.0, 1.0)
    exp_score = round(exp_factor * 40.0)
    total_score += exp_score

    # 2. Skills score: up to 30 points
    skills_count = len(analysis.technical_skills or [])
    # Cap factor at 1.0 (e.g., 10+ skills = 1.0)
    skills_factor = min(skills_count / 10.0, 1.0)
    skills_score = round(skills_factor * 30.0)
    total_score += skills_score

    # 3. Communication score: up to 20 points (expects 0-10 in string)
    try:
        m = re.search(r"(\d+(\.\d+)?)", str(analysis.communication_skills))
        # Use regex match if available, otherwise try direct float conversion
        comm_val = float(m.group(1)) if m else float(str(analysis.communication_skills))
        comm_val = max(0.0, min(10.0, comm_val)) # Clamp to 0-10
    except Exception:
        comm_val = 5.0 # Default if model extraction failed
    comm_score = round((comm_val / 10.0) * 20.0)
    total_score += comm_score

    # 4. Certifications score: up to 10 points
    # Max 10 points for 10 or more certifications
    certs_score = min(len(analysis.certifications or []), 10) * 1.0
    total_score += certs_score

    # 5. Therapist bonus: up to 10 points (added to overall score if applicable)
    spec_bonus = 0.0
    if role.lower() == "therapist":
        def safe_score(x):
            try:
                m = re.search(r"(\d+(\.\d+)?)", str(x))
                return float(m.group(1)) if m else 0.0
            except Exception:
                return 0.0
        aba = safe_score(analysis.aba_therapy_skills)
        autism = safe_score(analysis.autism_care_experience_score)
        # Average of the two specialized scores, scaled to a max of 10 points
        spec_bonus = ((aba + autism) / 20.0) * 10.0
        total_score += spec_bonus

    final_score = round(min(total_score, 100))
    
    return (float(final_score), float(exp_score), float(skills_score), float(comm_score), float(certs_score))

# --- Append to DataFrame (MODIFIED) ---
def append_analysis_to_dataframe(job_role: str, analysis: ResumeAnalysis, scores: tuple[float, float, float, float, float]):
    final_score, exp_score, skills_score, comm_score, certs_score = scores
    
    data = analysis.dict()
    tech = ", ".join(data.get("technical_skills") or [])
    certs = ", ".join(data.get("certifications") or [])
    
    row = {
        'Name': data.get("name") or "",
        'Job Role': job_role,
        'Resume Score (100)': final_score,
        'Shortlisted': 'No',
        'Email': data.get("email") or "",
        'Phone': data.get("phone") or "",
        
        # NEW SCORE COLUMNS
        'Experience Score (40)': exp_score,
        'Skills Score (30)': skills_score,
        'Communication Score (20)': comm_score,
        'Certifications Score (10)': certs_score,
        
        'Experience Summary': data.get("experience_summary") or "",
        'Education Summary': data.get("education_summary") or "",
        'Communication Rating (1-10)': str(data.get("communication_skills") or "N/A"),
        'Skills/Technologies': tech,
        'Certifications': certs,
        'ABA Skills (1-10)': str(data.get("aba_therapy_skills") or "N/A"),
        'RBT/BCBA Cert': str(data.get("rbt_bcba_certification") or "N/A"),
        'Autism-Care Exp (1-10)': str(data.get("autism_care_experience_score") or "N/A"),
    }
    new_df = pd.DataFrame([row])
    st.session_state.analyzed_data = pd.concat([st.session_state.analyzed_data, new_df], ignore_index=True)

# --- Excel export helper (No change) ---
def df_to_excel_bytes(df: pd.DataFrame) -> bytes:
    output = io.BytesIO()
    with pd.ExcelWriter(output, engine="openpyxl") as writer:
        df.to_excel(writer, index=False, sheet_name="Resume Analysis Data")
    return output.getvalue()

# --- UI Layout ---
st.title("🌌 Quantum Scrutiny Platform: AI Resume Analysis (Single-file)")

tab_user, tab_admin = st.tabs(["👤 Resume Uploader (User Panel)", "🔒 Admin Dashboard (Password Protected)"])

# --- User Panel (Minor change for scoring) ---
with tab_user:
    st.header("Upload Resumes for Analysis")
    st.info("Upload multiple PDF or DOCX files. The Groq AI engine will extract and score fields.")

    job_role_options = ["Software Engineer", "ML Engineer", "Therapist", "Data Analyst", "Project Manager"]
    selected_role = st.selectbox("1. Select the Target Job Role", options=job_role_options, key="selected_role")

    uploaded_files = st.file_uploader("2. Upload Resumes (PDF or DOCX)", type=["pdf", "docx"], accept_multiple_files=True)

    if st.button("🚀 Analyze All Uploaded Resumes"):
        if not uploaded_files:
            st.warning("Please upload one or more resume files to begin analysis.")
        else:
            st.session_state.run_analysis = True
            st.rerun()

    if st.session_state.get("run_analysis", False):
        if not uploaded_files:
            st.warning("No files found. Upload files and try again.")
            st.session_state.run_analysis = False
        else:
            total = len(uploaded_files)
            progress = st.progress(0)
            st.session_state.individual_analysis = []
            idx = 0
            with st.spinner("Processing resumes..."):
                for f in uploaded_files:
                    idx += 1
                    try:
                        st.write(f"Analyzing **{f.name}**...")
                        resume_text = extract_text_from_file(f)
                        if not resume_text:
                            st.error(f"Could not extract text from {f.name}. Skipping.")
                            progress.progress(idx / total)
                            continue

                        analysis = analyze_resume_with_groq_cached(resume_text, selected_role)

                        if analysis.name == "Extraction Failed":
                            st.error(f"Extraction failed for {f.name}. See debug output.")
                            progress.progress(idx / total)
                            continue

                        scores = calculate_resume_score(analysis, selected_role)
                        final_score = scores[0]
                        
                        append_analysis_to_dataframe(selected_role, analysis, scores)

                        st.session_state.individual_analysis.append({
                            'name': analysis.name,
                            'score': final_score,
                            'role': selected_role,
                            'file_name': f.name
                        })
                    except Exception as e:
                        st.error(f"Error analyzing {f.name}: {e}")
                        st.exception(traceback.format_exc())
                    finally:
                        progress.progress(idx / total)

            st.success(f"✅ Successfully processed {len(st.session_state.individual_analysis)} of {total} resumes.")
            st.session_state.run_analysis = False

    # Display last results summary
    if st.session_state.individual_analysis:
        st.subheader("Last Analysis Summary")
        for item in st.session_state.individual_analysis:
            st.markdown(f"**{item['name']}** (for **{item['role']}**) - **Score: {item['score']}/100**")
        st.markdown("---")
        st.caption("All analyzed data is stored in the Admin Dashboard.")

# --- Admin Panel (MODIFIED for new columns) ---
with tab_admin:
    if not st.session_state.is_admin_logged_in:
        st.header("Admin Login")
        password = st.text_input("Enter Admin Password", type="password")
        if st.button("🔑 Login"):
            if password == ADMIN_PASSWORD:
                st.session_state.is_admin_logged_in = True
                st.rerun()
            else:
                st.error("Incorrect password.")
        st.stop()

    st.header("🎯 Recruitment Dashboard")
    if st.button("🚪 Logout"):
        st.session_state.is_admin_logged_in = False
        st.rerun()

    if st.session_state.analyzed_data.empty:
        st.warning("No resume data has been analyzed yet. Please upload files in the User Panel.")
    else:
        df = st.session_state.analyzed_data.copy()
        st.subheader("Candidate Data Table")
        st.success(f"**Total Candidates Analyzed: {len(df)}**")

        # Updated columns for display in the data editor
        display_cols = [
            'Name', 
            'Job Role', 
            'Resume Score (100)', 
            'Experience Score (40)', 
            'Skills Score (30)', 
            'Communication Score (20)', 
            'Certifications Score (10)',
            'Shortlisted', 
            'Email', 
            'Skills/Technologies'
        ]
        # Filter columns to only those present in the current dataframe (safety check)
        current_display_cols = [col for col in display_cols if col in df.columns]

        edited_df = st.data_editor(
            df[current_display_cols],
            column_config={
                "Shortlisted": st.column_config.SelectboxColumn(
                    "Shortlisted",
                    help="Mark the candidate as Shortlisted or Rejected.",
                    options=["No", "Yes"],
                    required=True
                ),
                "Resume Score (100)": st.column_config.ProgressColumn(
                    "Total Score",
                    format="%f",
                    min_value=0, max_value=100,
                ),
                "Experience Score (40)": st.column_config.ProgressColumn(
                    "Experience (40)",
                    format="%f",
                    min_value=0, max_value=40,
                ),
                "Skills Score (30)": st.column_config.ProgressColumn(
                    "Skills (30)",
                    format="%f",
                    min_value=0, max_value=30,
                ),
                "Communication Score (20)": st.column_config.ProgressColumn(
                    "Comms (20)",
                    format="%f",
                    min_value=0, max_value=20,
                ),
                "Certifications Score (10)": st.column_config.ProgressColumn(
                    "Certs (10)",
                    format="%f",
                    min_value=0, max_value=10,
                ),
            },
            key="dashboard_editor",
            hide_index=True
        )

        # The logic to update the session state with the edited 'Shortlisted' column remains the same
        try:
            # Update the master dataframe with the edited 'Shortlisted' column
            for col in edited_df.columns:
                if col in st.session_state.analyzed_data.columns and not edited_df[col].equals(st.session_state.analyzed_data[col]):
                    # Only update 'Shortlisted' which is the only editable field
                    if col == 'Shortlisted':
                        st.session_state.analyzed_data.loc[:, 'Shortlisted'] = edited_df['Shortlisted'].values
        except Exception:
            # Fallback for index issues on data_editor changes
            for i, val in enumerate(edited_df.get('Shortlisted', []).tolist()):
                if i < len(st.session_state.analyzed_data):
                    st.session_state.analyzed_data.at[i, 'Shortlisted'] = val


        st.markdown("---")
        st.subheader("📥 Download Data")
        df_export = st.session_state.analyzed_data.copy()
        excel_bytes = df_to_excel_bytes(df_export)

        st.download_button(
            label="💾 Download All Data as Excel (.xlsx)",
            data=excel_bytes,
            file_name="quantum_scrutiny_report.xlsx",
            mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
            help="Downloads the full table including all extracted fields and shortlist status."
        )

# --- End of file ---