app.py

import gradio as gr
import requests
import json
import os
import hashlib
import uuid
from datetime import datetime
import io

# ── Gemini client (REST API — no SDK dependency) ──────────────────────────────
GEMINI_KEY   = os.environ.get("GEMINI_API_KEY", "")
GEMINI_MODEL = "gemini-2.0-flash"

def ask_gemini(prompt: str, short: bool = False) -> str:
    if not GEMINI_KEY:
        return "⚠️ GEMINI_API_KEY not set. Add it in Space → Settings → Secrets."
    
    if short:
        prompt = "Answer concisely in 3-5 sentences or bullet points. No long paragraphs.\n\n" + prompt
    
    url = f"https://generativelanguage.googleapis.com/v1beta/models/{GEMINI_MODEL}:generateContent?key={GEMINI_KEY}"
    
    payload = {
        "contents": [{
            "parts": [{"text": prompt}]
        }]
    }
    
    try:
        resp = requests.post(url, json=payload, timeout=60)
        data = resp.json()
        
        if resp.status_code != 200:
            error_msg = data.get("error", {}).get("message", str(data))
            return f"⚠️ API Error: {error_msg}"
        
        candidates = data.get("candidates", [])
        if not candidates:
            return "⚠️ No response from Gemini."
        
        text = candidates[0].get("content", {}).get("parts", [{}])[0].get("text", "")
        return text if text else "⚠️ Empty response from Gemini."
        
    except Exception as e:
        return f"⚠️ API Error: {str(e)}"

# ── Persistent storage paths ───────────────────────────────────────────────────
DATA_DIR = "/data"
USERS_FILE = os.path.join(DATA_DIR, "users.json")
SESSIONS_DIR = os.path.join(DATA_DIR, "sessions")

def _ensure_dirs():
    os.makedirs(DATA_DIR, exist_ok=True)
    os.makedirs(SESSIONS_DIR, exist_ok=True)

def _load_users() -> dict:
    _ensure_dirs()
    if not os.path.exists(USERS_FILE):
        return {}
    try:
        with open(USERS_FILE) as f:
            return json.load(f)
    except Exception:
        return {}

def _save_users(users: dict):
    _ensure_dirs()
    with open(USERS_FILE, "w") as f:
        json.dump(users, f, indent=2)

def _hash_pw(pw: str) -> str:
    return hashlib.sha256(pw.encode()).hexdigest()

def _user_session_file(username: str) -> str:
    return os.path.join(SESSIONS_DIR, f"{username}.json")

def load_user_session(username: str) -> dict:
    path = _user_session_file(username)
    if os.path.exists(path):
        try:
            with open(path) as f:
                return json.load(f)
        except Exception:
            pass
    return {"username": username, "phases": {}, "last_saved": None, "search_count": 0}

def save_user_session(username: str, session: dict):
    _ensure_dirs()
    session["last_saved"] = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
    with open(_user_session_file(username), "w") as f:
        json.dump(session, f, indent=2)

# ── Auth functions ─────────────────────────────────────────────────────────────
FREE_SEARCH_LIMIT = 5

def register_user(username: str, password: str, confirm: str) -> tuple[str, str]:
    username = username.strip().lower()
    if not username or len(username) < 3:
        return "❌ Username must be at least 3 characters.", ""
    if not password or len(password) < 6:
        return "❌ Password must be at least 6 characters.", ""
    if password != confirm:
        return "❌ Passwords do not match.", ""
    users = _load_users()
    if username in users:
        return "❌ Username already taken. Choose another.", ""
    users[username] = {
        "pw_hash": _hash_pw(password),
        "created": datetime.now().strftime("%Y-%m-%d"),
        "plan": "free",
    }
    _save_users(users)
    return f"✅ Account created! Welcome, {username}. You can now log in.", ""

def login_user(username: str, password: str) -> tuple[str, str, dict]:
    username = username.strip().lower()
    users = _load_users()
    if username not in users:
        return "❌ Username not found.", "", {}
    if users[username]["pw_hash"] != _hash_pw(password):
        return "❌ Incorrect password.", "", {}
    session = load_user_session(username)
    plan = users[username].get("plan", "free")
    count = session.get("search_count", 0)
    msg = f"✅ Welcome back, {username}! Plan: {plan.upper()} | Searches used: {count}"
    return msg, username, session

def reset_password(username: str, new_pw: str, confirm: str) -> str:
    username = username.strip().lower()
    if new_pw != confirm:
        return "❌ Passwords do not match."
    if len(new_pw) < 6:
        return "❌ Password too short (min 6 characters)."
    users = _load_users()
    if username not in users:
        return "❌ Username not found."
    users[username]["pw_hash"] = _hash_pw(new_pw)
    _save_users(users)
    return f"✅ Password reset for {username}. You can now log in."

def check_search_limit(username: str, session: dict) -> tuple[bool, str]:
    if not username:
        return False, "❌ Please log in first."
    users = _load_users()
    plan = users.get(username, {}).get("plan", "free")
    if plan != "free":
        return True, ""
    count = session.get("search_count", 0)
    if count >= FREE_SEARCH_LIMIT:
        return False, f"⚠️ Free plan limit reached ({FREE_SEARCH_LIMIT} searches). Contact admin to upgrade."
    return True, f"Free searches remaining: {FREE_SEARCH_LIMIT - count - 1}"

def increment_search(username: str, session: dict) -> dict:
    session["search_count"] = session.get("search_count", 0) + 1
    if username:
        save_user_session(username, session)
    return session

# ── Domains & Study Types ──────────────────────────────────────────────────────
DOMAINS = {
    "🫁 Respiratory Therapy": "respiratory therapy pulmonology ventilation",
    "🏥 Emergency Medicine": "emergency medicine acute care trauma",
    "💉 Anesthesia": "anesthesia anesthesiology perioperative",
    "🔬 Clinical Laboratory": "clinical laboratory diagnostics biomarkers hematology",
    "❤️ Cardiac Surgery (ICVT/ECVT)": "cardiac surgery cardiopulmonary bypass ECMO perfusion",
    "🫀 Cardiology Programs": "cardiology cardiac rehabilitation interventional cardiology",
    "📡 Radiology": "radiology diagnostic imaging MRI CT ultrasound",
    "🦻 Audiology": "audiology hearing loss vestibular rehabilitation",
    "🤲 Occupational Therapy": "occupational therapy rehabilitation functional outcomes",
    "🦯 Physical Therapy": "physical therapy physiotherapy musculoskeletal",
    "🧠 Neurology": "neurology neuroscience stroke rehabilitation",
    "🍼 Pediatrics": "pediatrics child health neonatal care",
    "🌡️ Critical Care / ICU": "critical care intensive care sepsis mechanical ventilation",
    "🧬 Oncology": "oncology cancer treatment chemotherapy immunotherapy",
    "💊 Pharmacy / Pharmacology": "clinical pharmacy pharmacology drug interactions",
    "🧪 Medical Biotechnology": "medical biotechnology molecular diagnostics genomics",
    "🏋️ Sports Medicine": "sports medicine athletic training exercise physiology",
    "👁️ Ophthalmology": "ophthalmology eye disease retina cornea",
    "🦷 Dental / Oral Health": "dentistry oral health periodontics",
    "🧘 Mental Health / Psychiatry": "psychiatry mental health cognitive behavioral therapy",
    "🤱 Midwifery / Obstetrics": "midwifery obstetrics maternal health",
    "🌍 Public Health / Epidemiology": "public health epidemiology health policy",
    "💻 Health Informatics": "health informatics digital health EHR AI in medicine",
    "🔩 Biomedical Engineering": "biomedical engineering medical devices prosthetics",
    "🧑‍⚕️ Nursing": "nursing patient care clinical outcomes",
}

STUDY_TYPES = [
    "Systematic Review & Meta-Analysis", "Randomized Controlled Trial (RCT)",
    "Cohort Study", "Case-Control Study", "Cross-Sectional Study",
    "Case Series / Case Report", "Scoping Review", "Narrative Review",
    "Qualitative Study", "Mixed Methods", "Pilot / Feasibility Study",
    "Diagnostic Accuracy Study", "Economic Evaluation",
]

JOURNALS = {
    "High Impact General": ["NEJM", "The Lancet", "JAMA", "BMJ", "Nature Medicine"],
    "Systematic Reviews": ["Systematic Reviews (BMC)", "Cochrane Database", "JBI Evidence Synthesis"],
    "Open Access Fast": ["PLOS Medicine", "PLOS ONE", "BMC Medicine", "Frontiers in Medicine"],
    "Digital Health": ["JMIR", "npj Digital Medicine", "Health Informatics Journal"],
    "Allied Health": ["Journal of Allied Health", "Physical Therapy", "AJOT", "Annals of Emergency Medicine"],
    "Gulf / Regional": ["Saudi Medical Journal", "Eastern Mediterranean Health Journal", "AAMJ"],
    "Student / First Pub": ["Cureus", "Medical Education Online", "BMC Medical Education"],
}

# ── Semantic Scholar ───────────────────────────────────────────────────────────
def search_papers(query: str, year_from: int = 2020, limit: int = 20):
    try:
        params = {
            "query": query, "limit": limit,
            "fields": "title,authors,year,abstract,citationCount,externalIds,publicationVenue",
            "year": f"{year_from}-2026",
        }
        r = requests.get("https://api.semanticscholar.org/graph/v1/paper/search",
                         params=params, timeout=15)
        if r.status_code == 200:
            return r.json().get("data", [])
    except Exception:
        pass
    return []

def format_papers_table(papers):
    if not papers:
        return "*No papers found. Try broadening your search terms.*"
    rows = []
    for i, p in enumerate(papers[:15], 1):
        authors = ", ".join(a.get("name", "") for a in p.get("authors", [])[:2])
        if len(p.get("authors", [])) > 2:
            authors += " et al."
        year = p.get("year", "N/A")
        title = (p.get("title") or "N/A")[:80]
        cites = p.get("citationCount", 0)
        venue = ((p.get("publicationVenue") or {}).get("name") or "Unknown")[:35]
        doi = (p.get("externalIds") or {}).get("DOI", "")
        doi_str = f"[DOI]({doi})" if doi else "—"
        rows.append(f"| {i} | {title} | {authors} | {year} | {venue} | {cites} | {doi_str} |")
    header = "| # | Title | Authors | Year | Journal | Cites | DOI |\n|---|-------|---------|------|---------|-------|-----|\n"
    return header + "\n".join(rows)

# ── Phase functions ────────────────────────────────────────────────────────────
def run_gap_analysis(domain, topic, year_from, study_type, username, session):
    ok, msg = check_search_limit(username, session)
    if not ok:
        return msg, "", "", session
    if not topic.strip():
        return "❌ Please enter a research topic.", "", "", session
    domain_kw = DOMAINS.get(domain, domain)
    papers = search_papers(f"{topic} {domain_kw}", int(year_from))
    table_md = format_papers_table(papers)
    abstracts = "\n\n".join(
        f"[{p.get('year')}] {p.get('title','')}: {(p.get('abstract') or '')[:280]}"
        for p in papers[:10] if p.get("abstract")
    )
    prompt = f"""Expert research analyst in {domain}. Researcher topic: "{topic}". Study type: {study_type}.

Recent papers 2020-2026:
{abstracts or "Use expert knowledge of the field."}

Provide:
## Executive Summary
## What Is Well-Researched (4-5 areas)
## 5 Research Gaps (each: Gap Title, Why it matters, Recommended study type, Target population, Expected contribution)
## Top Recommended Gap
## Methodological Gaps
## Gulf & Asian Context Gaps

Flag unverified claims with [CITE NEEDED]."""
    gap = ask_gemini(prompt)
    session = increment_search(username, session)
    session.setdefault("phases", {})
    session["phases"]["p1_topic"] = topic
    session["phases"]["p1_domain"] = domain
    session["phases"]["p1_gap"] = gap
    session["phases"]["p1_table"] = table_md
    if username:
        save_user_session(username, session)
    return table_md, gap, f"✅ Auto-saved | Searches used: {session['search_count']}", session

def generate_design(domain, topic, gap_focus, study_type, username, session):
    if not topic.strip() or not gap_focus.strip():
        return "❌ Enter topic and gap focus.", session
    prompt = f"""Senior research methodologist in {domain}.
Topic: {topic} | Gap: {gap_focus} | Study: {study_type}

Generate complete research design:
## PICO/SPIDER Framework
## Research Objectives (primary + 2 secondary)
## Research Questions
## Methodology (design, setting, sample, inclusion/exclusion, sample size, tools, stats)
## Structured Abstract Draft (Background/Objective/Methods/Expected Results/Conclusion/Keywords)
## Ethics & Registration (IRB, PROSPERO yes/no, checklist)
## 12-Month Timeline

Flag claims with [CITE NEEDED]."""
    result = ask_gemini(prompt)
    session.setdefault("phases", {})
    session["phases"]["p2_design"] = result
    if username:
        save_user_session(username, session)
    return result, session

def writing_assistant(domain, topic, section, rough_notes, username, session):
    if not rough_notes.strip():
        return "❌ Add your rough notes first.", session
    prompt = f"""Expert medical writer in {domain}. Paper: "{topic}". Section: {section}.

Rough notes:
{rough_notes}

Transform into polished publication-ready academic prose for {section}.
- Formal medical English
- Tag unverified claims: [CITE NEEDED — search: "keywords"]
- Tag missing data: [DATA NEEDED]
- Vary sentence structure (avoid AI-detectable patterns)

After text add:
## Writing Coach Notes (3 tips)
## Self-Check Checklist"""
    result = ask_gemini(prompt)
    session.setdefault("phases", {})
    prev = session["phases"].get("p3_writing", "")
    session["phases"]["p3_writing"] = prev + f"\n\n=== {section} ===\n" + result
    if username:
        save_user_session(username, session)
    return result, session

def match_journals(domain, topic, study_type, open_access, target_if, username, session):
    prompt = f"""Journal selection expert. Profile: {domain} | {topic} | {study_type} | {open_access} | IF target: {target_if}

Recommend 8 journals. For each:
### [Rank]. [Journal] — IF: X.X
- Publisher, Indexing, Scope fit, Acceptance rate, Review time, APC fee
- Gulf/Asian author note, Strategic tip
- ⚠️ Predatory check: SAFE/CAUTION

Include 2 Gulf/regional journals, 1 fast-track. Flag non-Scopus/WoS journals."""
    result = ask_gemini(prompt)
    session.setdefault("phases", {})
    session["phases"]["p4_journals"] = result
    if username:
        save_user_session(username, session)
    return result, session

def reviewer_response(domain, topic, reviewer_comments, username, session):
    if not reviewer_comments.strip():
        return "❌ Paste reviewer comments.", session
    prompt = f"""Expert in peer review responses for {domain}. Paper: "{topic}"

Comments:
{reviewer_comments}

Write professional response letter with:
- Opening paragraph
- Response to each comment (quote → author response → manuscript change)
- Cover letter for revised submission
- Re-submission checklist

Professional tone, never defensive. [CITE NEEDED] where needed."""
    result = ask_gemini(prompt)
    session.setdefault("phases", {})
    session["phases"]["p5_reviewer"] = result
    if username:
        save_user_session(username, session)
    return result, session

def export_document(username, session):
    if not username:
        return None
    phases = session.get("phases", {})
    
    md_content = f"""# Research Portfolio: {phases.get("p1_topic", "Untitled")}

**ScholarPath** | User: {username} | Saved: {session.get("last_saved", "—")}

---
"""
    content_map = [
        ("Phase 1 — Gap Analysis", "p1_gap"),
        ("Phase 2 — Research Design", "p2_design"),
        ("Phase 3 — Writing", "p3_writing"),
        ("Phase 4 — Journals", "p4_journals"),
        ("Phase 5 — Reviewer Response", "p5_reviewer"),
    ]
    for heading, key in content_map:
        if phases.get(key):
            md_content += f"\n\n## {heading}\n\n{phases[key]}\n\n---\n"
    
    topic_slug = (phases.get("p1_topic", "portfolio") or "portfolio")[:25].replace(" ", "_")
    path = f"/tmp/{username}_{topic_slug}.md"
    with open(path, "w", encoding="utf-8") as f:
        f.write(md_content)
    return path

def load_saved_work(username, session):
    phases = session.get("phases", {})
    saved = session.get("last_saved")
    if not phases:
        return "No saved work yet.", "", "", "", ""
    return (
        phases.get("p1_gap", ""),
        phases.get("p2_design", ""),
        phases.get("p3_writing", ""),
        phases.get("p4_journals", ""),
        f"Last saved: {saved}" if saved else "Not saved yet.",
    )

# ── CSS ────────────────────────────────────────────────────────────────────────
CSS = """
@import url('https://fonts.googleapis.com/css2?family=Cormorant+Garamond:wght@400;600;700&family=Plus+Jakarta+Sans:wght@300;400;500;600&display=swap');

:root {
    --navy: #0a1628; --gold: #c9a84c; --gold-light: #e8c87a;
    --cream: #faf7f0; --sage: #1e3a5f; --text: #1a1a2e;
}
body, .gradio-container { font-family: 'Plus Jakarta Sans', sans-serif !important; background: var(--cream) !important; }
h1,h2,h3 { font-family: 'Cormorant Garamond', serif !important; }
.gradio-container { max-width: 1200px !important; margin: 0 auto !important; }

/* Buttons */
.btn-primary button {
    background: linear-gradient(135deg, #0a1628, #1e3a5f) !important;
    color: #faf7f0 !important; font-weight: 600 !important;
    border: none !important; border-radius: 8px !important;
    box-shadow: 0 4px 12px rgba(10,22,40,.3) !important;
    transition: all .25s !important;
}
.btn-primary button:hover { background: linear-gradient(135deg,#1e3a5f,#c9a84c) !important; transform: translateY(-1px) !important; }
.btn-gold button {
    background: linear-gradient(135deg, #c9a84c, #e8c87a) !important;
    color: #0a1628 !important; font-weight: 700 !important;
    border: none !important; border-radius: 8px !important;
}

/* Inputs */
textarea, input[type=text], input[type=password] {
    border: 1.5px solid #e2d9c8 !important; border-radius: 8px !important;
    background: white !important; color: var(--text) !important;
    font-family: 'Plus Jakarta Sans', sans-serif !important;
}
textarea:focus, input:focus { border-color: #c9a84c !important; box-shadow: 0 0 0 3px rgba(201,168,76,.15) !important; }

/* Tabs */
.tab-nav button { font-family: 'Plus Jakarta Sans', sans-serif !important; font-weight: 500 !important; color: #1e3a5f !important; }
.tab-nav button.selected { color: #c9a84c !important; border-bottom: 2px solid #c9a84c !important; }

/* Notice */
.notice { background: linear-gradient(135deg,#fef3c7,#fde68a) !important; border: 1px solid #d97706 !important;
    border-radius: 8px !important; padding: 10px 14px !important; color: #92400e !important; font-size: 13px !important; }
.notice-green { background: #d1fae5 !important; border: 1px solid #065f46 !important;
    border-radius: 8px !important; padding: 10px 14px !important; color: #065f46 !important; font-size: 13px !important; }

/* Floating chat button */
#float-chat-btn {
    position: fixed !important; bottom: 24px !important; right: 24px !important;
    width: 60px !important; height: 60px !important;
    background: linear-gradient(135deg,#0a1628,#1e3a5f) !important;
    border-radius: 50% !important; border: 2px solid #c9a84c !important;
    display: flex !important; align-items: center !important; justify-content: center !important;
    cursor: pointer !important; z-index: 9999 !important;
    box-shadow: 0 4px 20px rgba(10,22,40,.5) !important;
    font-size: 26px !important; transition: transform .2s !important;
}
#float-chat-btn:hover { transform: scale(1.12) !important; }
#chat-panel {
    position: fixed !important; bottom: 96px !important; right: 24px !important;
    width: 360px !important; height: 480px !important;
    background: white !important; border-radius: 16px !important;
    border: 1px solid #e2d9c8 !important; z-index: 9998 !important;
    box-shadow: 0 8px 40px rgba(10,22,40,.25) !important;
    display: none; flex-direction: column !important; overflow: hidden !important;
}
#chat-panel.open { display: flex !important; }
#chat-panel-header {
    background: linear-gradient(135deg,#0a1628,#1e3a5f) !important;
    padding: 12px 16px !important; color: #e8c87a !important;
    font-family: 'Cormorant Garamond',serif !important; font-size: 16px !important;
    font-weight: 600 !important; display: flex !important; justify-content: space-between !important;
    align-items: center !important;
}
#chat-messages {
    flex: 1 !important; overflow-y: auto !important; padding: 12px !important;
    display: flex !important; flex-direction: column !important; gap: 8px !important;
    background: #faf7f0 !important;
}
.msg-user { background: #0a1628 !important; color: white !important; padding: 8px 12px !important;
    border-radius: 12px 12px 2px 12px !important; font-size: 13px !important;
    align-self: flex-end !important; max-width: 80% !important; }
.msg-ai { background: white !important; color: #1a1a2e !important; padding: 8px 12px !important;
    border-radius: 12px 12px 12px 2px !important; font-size: 13px !important;
    border: 1px solid #e2d9c8 !important; align-self: flex-start !important; max-width: 85% !important; }
#chat-input-row { display: flex !important; padding: 8px !important; gap: 6px !important;
    border-top: 1px solid #e2d9c8 !important; background: white !important; }
#chat-input-row input {
    flex: 1 !important; border: 1px solid #e2d9c8 !important; border-radius: 20px !important;
    padding: 8px 14px !important; font-size: 13px !important; outline: none !important;
}
#chat-send-btn {
    background: linear-gradient(135deg,#c9a84c,#e8c87a) !important;
    border: none !important; border-radius: 50% !important;
    width: 36px !important; height: 36px !important; cursor: pointer !important;
    font-size: 16px !important; color: #0a1628 !important; font-weight: 700 !important;
}
"""

HEADER = """
<div style="background:linear-gradient(135deg,#0a1628 0%,#1e3a5f 50%,#0a1628 100%);
    padding:36px 48px;border-radius:16px;margin-bottom:8px;position:relative;overflow:hidden;">
    <div style="position:absolute;top:-20px;right:-20px;width:200px;height:200px;
        border:1px solid rgba(201,168,76,.15);border-radius:50%;"></div>
    <div style="position:absolute;top:10px;right:40px;width:120px;height:120px;
        border:1px solid rgba(201,168,76,.1);border-radius:50%;"></div>
    <div style="position:relative;z-index:1;">
        <div style="display:flex;align-items:center;gap:14px;margin-bottom:10px;">
            <div style="width:48px;height:48px;background:linear-gradient(135deg,#c9a84c,#e8c87a);
                border-radius:12px;display:flex;align-items:center;justify-content:center;font-size:22px;">🔬</div>
            <div>
                <h1 style="font-family:'Cormorant Garamond',serif;color:#e8c87a;font-size:30px;
                    font-weight:700;margin:0;letter-spacing:1px;">ScholarPath</h1>
                <p style="color:rgba(255,255,255,.5);font-size:12px;margin:0;">
                    Research Intelligence Portal · Allied Health & Medical Sciences</p>
            </div>
        </div>
        <p style="color:rgba(255,255,255,.85);font-size:14px;margin:0;max-width:660px;line-height:1.6;">
            End-to-end AI research support — from gap to publication. Built for Gulf & Asian researchers.
        </p>
        <div style="display:flex;gap:10px;margin-top:14px;flex-wrap:wrap;">
            <span style="background:rgba(201,168,76,.2);color:#e8c87a;padding:3px 10px;
                border-radius:20px;font-size:11px;border:1px solid rgba(201,168,76,.3);">✦ 200M+ Papers</span>
            <span style="background:rgba(201,168,76,.2);color:#e8c87a;padding:3px 10px;
                border-radius:20px;font-size:11px;border:1px solid rgba(201,168,76,.3);">✦ Gemini 2.5 Flash</span>
            <span style="background:rgba(201,168,76,.2);color:#e8c87a;padding:3px 10px;
                border-radius:20px;font-size:11px;border:1px solid rgba(201,168,76,.3);">✦ Auto-Save</span>
            <span style="background:rgba(201,168,76,.2);color:#e8c87a;padding:3px 10px;
                border-radius:20px;font-size:11px;border:1px solid rgba(201,168,76,.3);">✦ 25 Disciplines</span>
            <span style="background:rgba(201,168,76,.2);color:#e8c87a;padding:3px 10px;
                border-radius:20px;font-size:11px;border:1px solid rgba(201,168,76,.3);">✦ Account & History</span>
        </div>
    </div>
</div>
"""

FLOATING_CHAT_HTML = """
<div id="float-chat-btn" onclick="toggleChat()" title="Ask AI Assistant">🤖</div>
<div id="chat-panel">
    <div id="chat-panel-header">
        <span>🔬 ScholarPath Assistant</span>
        <span onclick="toggleChat()" style="cursor:pointer;opacity:.7;font-size:18px;">✕</span>
    </div>
    <div id="chat-messages">
        <div class="msg-ai">👋 Hi! I'm your research assistant. Ask me anything about research methods, journals, citations, or how to use this portal.</div>
    </div>
    <div id="chat-input-row">
        <input id="chat-float-input" type="text" placeholder="Ask a question..." onkeydown="if(event.key==='Enter')sendFloatChat()"/>
        <button id="chat-send-btn" onclick="sendFloatChat()">➤</button>
    </div>
</div>
<script>
function toggleChat(){
    const p=document.getElementById('chat-panel');
    p.classList.toggle('open');
}
async function sendFloatChat(){
    const inp=document.getElementById('chat-float-input');
    const msgs=document.getElementById('chat-messages');
    const text=inp.value.trim();
    if(!text)return;
    inp.value='';
    msgs.innerHTML+=`<div class="msg-user">${text}</div>`;
    msgs.innerHTML+=`<div class="msg-ai" id="typing">⏳ Thinking...</div>`;
    msgs.scrollTop=msgs.scrollHeight;
    try{
        const r=await fetch('/float_chat',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({q:text})});
        const d=await r.json();
        document.getElementById('typing').outerHTML=`<div class="msg-ai">${d.answer.replace(/\\n/g,'<br>')}</div>`;
    }catch(e){
        document.getElementById('typing').outerHTML=`<div class="msg-ai">⚠️ Error. Please try again.</div>`;
    }
    msgs.scrollTop=msgs.scrollHeight;
}
</script>
"""

# ── Gradio App ─────────────────────────────────────────────────────────────────
with gr.Blocks(title="ScholarPath") as demo:

    gr.HTML(HEADER)
    gr.HTML('<div class="notice">🔒 Privacy: Only topics & public abstracts are processed. Never paste patient data or unpublished raw data. Nothing is shared externally.</div>')

    # Shared state
    state_user = gr.State("")
    state_session = gr.State({})

    with gr.Tabs():

        # ── Account Tab ──────────────────────────────────────────────────────
        with gr.Tab("👤 My Account"):
            with gr.Row():
                # Login
                with gr.Column():
                    gr.Markdown("### 🔑 Login")
                    li_user = gr.Textbox(label="Username")
                    li_pw = gr.Textbox(label="Password", type="password")
                    li_btn = gr.Button("Login", elem_classes=["btn-primary"], variant="primary")
                    li_msg = gr.Markdown("")

                # Register
                with gr.Column():
                    gr.Markdown("### ✏️ Create Account")
                    reg_user = gr.Textbox(label="Choose Username")
                    reg_pw = gr.Textbox(label="Password (min 6 chars)", type="password")
                    reg_pw2 = gr.Textbox(label="Confirm Password", type="password")
                    reg_btn = gr.Button("Create Account", elem_classes=["btn-primary"], variant="primary")
                    reg_msg = gr.Markdown("")

                # Reset
                with gr.Column():
                    gr.Markdown("### 🔄 Reset Password")
                    rs_user = gr.Textbox(label="Username")
                    rs_pw = gr.Textbox(label="New Password", type="password")
                    rs_pw2 = gr.Textbox(label="Confirm New Password", type="password")
                    rs_btn = gr.Button("Reset Password", elem_classes=["btn-gold"], variant="primary")
                    rs_msg = gr.Markdown("")

            gr.Markdown("---")
            gr.Markdown("### 📂 Your Saved Work")
            load_btn = gr.Button("📥 Load My Saved Work", elem_classes=["btn-primary"], variant="primary")
            save_status = gr.Markdown("")
            with gr.Row():
                sv_gap = gr.Textbox(label="Saved Gap Analysis", lines=4, interactive=False)
                sv_design = gr.Textbox(label="Saved Research Design", lines=4, interactive=False)
            with gr.Row():
                sv_writing = gr.Textbox(label="Saved Writing", lines=4, interactive=False)
                sv_journals = gr.Textbox(label="Saved Journals", lines=4, interactive=False)

            gr.HTML('<div class="notice-green">✅ Your work auto-saves after every action. If internet drops, your last save is restored when you log in again. Use the Export tab to download a .md backup anytime.</div>')

            def do_login(u, p):
                msg, username, session = login_user(u, p)
                return msg, username, session
            def do_register(u, p, p2):
                msg, _ = register_user(u, p, p2)
                return msg
            def do_reset(u, p, p2):
                return reset_password(u, p, p2)
            def do_load(username, session):
                if not username:
                    return "❌ Log in first.", "", "", "", ""
                g, d, w, j, s = load_saved_work(username, session)
                return s, g, d, w, j

            li_btn.click(do_login, [li_user, li_pw], [li_msg, state_user, state_session])
            reg_btn.click(do_register, [reg_user, reg_pw, reg_pw2], reg_msg)
            rs_btn.click(do_reset, [rs_user, rs_pw, rs_pw2], rs_msg)
            load_btn.click(do_load, [state_user, state_session], [save_status, sv_gap, sv_design, sv_writing, sv_journals])

        # ── Phase 1 ──────────────────────────────────────────────────────────
        with gr.Tab("🔍 Phase 1 · Gap Analysis"):
            gr.Markdown("### Discover Research Gaps — *Login required*")
            with gr.Row():
                with gr.Column(scale=1):
                    p1_domain = gr.Dropdown(choices=list(DOMAINS.keys()), label="Domain", value="🫁 Respiratory Therapy")
                    p1_topic = gr.Textbox(label="Research Topic", placeholder="e.g., NIV weaning in ICU patients", lines=2)
                    p1_year = gr.Slider(2015, 2025, value=2020, step=1, label="Search From Year")
                    p1_study = gr.Dropdown(choices=STUDY_TYPES, label="Study Type", value=STUDY_TYPES[0])
                    p1_btn = gr.Button("🚀 Run Gap Analysis", elem_classes=["btn-primary"], variant="primary")
                    p1_status = gr.Markdown("")
                with gr.Column(scale=2):
                    p1_table = gr.Markdown("*Papers appear here after search.*")
            p1_gap_out = gr.Markdown("*Gap report appears here.*")

            def do_gap(domain, topic, year, study, username, session):
                tbl, gap, status, session = run_gap_analysis(domain, topic, year, study, username, session)
                return tbl, gap, status, session
            p1_btn.click(do_gap, [p1_domain, p1_topic, p1_year, p1_study, state_user, state_session],
                         [p1_table, p1_gap_out, p1_status, state_session])

        # ── Phase 2 ──────────────────────────────────────────────────────────
        with gr.Tab("📐 Phase 2 · Research Design"):
            gr.Markdown("### Build Your Research Protocol")
            with gr.Row():
                with gr.Column(scale=1):
                    p2_domain = gr.Textbox(label="Domain", placeholder="e.g., Respiratory Therapy")
                    p2_topic = gr.Textbox(label="Topic", placeholder="From Phase 1")
                    p2_gap = gr.Textbox(label="Specific Gap to Address", lines=4, placeholder="Paste the gap from Phase 1...")
                    p2_study = gr.Dropdown(choices=STUDY_TYPES, label="Study Type", value=STUDY_TYPES[0])
                    p2_btn = gr.Button("🏗️ Generate Design", elem_classes=["btn-primary"], variant="primary")
                with gr.Column(scale=2):
                    p2_out = gr.Markdown("*Research design appears here.*")
            def do_design(domain, topic, gap, study, username, session):
                result, session = generate_design(domain, topic, gap, study, username, session)
                return result, session
            p2_btn.click(do_design, [p2_domain, p2_topic, p2_gap, p2_study, state_user, state_session], [p2_out, state_session])

        # ── Phase 3 ──────────────────────────────────────────────────────────
        with gr.Tab("✍️ Phase 3 · Writing"):
            gr.HTML('<div class="notice">⚠️ Do not paste patient data or unpublished raw datasets here.</div>')
            with gr.Row():
                with gr.Column(scale=1):
                    p3_domain = gr.Textbox(label="Domain")
                    p3_topic = gr.Textbox(label="Paper Topic")
                    p3_section = gr.Dropdown(
                        choices=["Abstract","Introduction","Literature Review","Methodology",
                                 "Results","Discussion","Conclusion","References (APA 7)"],
                        label="Section", value="Introduction")
                    p3_notes = gr.Textbox(label="Your Rough Notes", lines=10,
                                          placeholder="Bullet points, rough draft, key ideas...")
                    p3_btn = gr.Button("✨ Polish Section", elem_classes=["btn-primary"], variant="primary")
                with gr.Column(scale=2):
                    p3_out = gr.Markdown("*Output appears here.*")
            def do_writing(domain, topic, section, notes, username, session):
                result, session = writing_assistant(domain, topic, section, notes, username, session)
                return result, session
            p3_btn.click(do_writing, [p3_domain, p3_topic, p3_section, p3_notes, state_user, state_session], [p3_out, state_session])

        # ── Phase 4 ──────────────────────────────────────────────────────────
        with gr.Tab("📰 Phase 4 · Journals"):
            with gr.Row():
                with gr.Column(scale=1):
                    p4_domain = gr.Textbox(label="Domain")
                    p4_topic = gr.Textbox(label="Topic")
                    p4_study = gr.Dropdown(choices=STUDY_TYPES, label="Study Type", value=STUDY_TYPES[0])
                    p4_oa = gr.Radio(["Open Access Required","Open Access Preferred","Any"], label="Open Access", value="Open Access Preferred")
                    p4_if = gr.Dropdown(
                        choices=["Any (student/first pub)","IF 1–3","IF 3–6","IF 6+","Top 10 only"],
                        label="Target Impact Factor", value="IF 1–3")
                    p4_btn = gr.Button("🎯 Match Journals", elem_classes=["btn-primary"], variant="primary")
                with gr.Column(scale=2):
                    p4_out = gr.Markdown("*Journal recommendations appear here.*")
            def do_journals(domain, topic, study, oa, if_t, username, session):
                result, session = match_journals(domain, topic, study, oa, if_t, username, session)
                return result, session
            p4_btn.click(do_journals, [p4_domain, p4_topic, p4_study, p4_oa, p4_if, state_user, state_session], [p4_out, state_session])

        # ── Phase 5 ──────────────────────────────────────────────────────────
        with gr.Tab("📨 Phase 5 · Reviewer"):
            with gr.Row():
                with gr.Column(scale=1):
                    p5_domain = gr.Textbox(label="Domain")
                    p5_topic = gr.Textbox(label="Paper Title")
                    p5_comments = gr.Textbox(label="Reviewer Comments", lines=12, placeholder="Paste full reviewer comments...")
                    p5_btn = gr.Button("📝 Generate Response", elem_classes=["btn-primary"], variant="primary")
                with gr.Column(scale=2):
                    p5_out = gr.Markdown("*Response letter appears here.*")
            def do_reviewer(domain, topic, comments, username, session):
                result, session = reviewer_response(domain, topic, comments, username, session)
                return result, session
            p5_btn.click(do_reviewer, [p5_domain, p5_topic, p5_comments, state_user, state_session], [p5_out, state_session])

        # ── Export ───────────────────────────────────────────────────────────
        with gr.Tab("💾 Export"):
            gr.Markdown("""### Download Your Complete Research Portfolio
Work auto-saves after every action. Use this to download a .md backup anytime.""")
            exp_btn = gr.Button("📥 Download My Portfolio (.md)", elem_classes=["btn-gold"], variant="primary")
            exp_file = gr.File(label="Download")
            exp_btn.click(export_document, [state_user, state_session], exp_file)

        # ── Inline AI Chat (full tab) ─────────────────────────────────────
        with gr.Tab("🤖 AI Assistant"):
            gr.Markdown("### Ask Research Questions — Concise answers")
            chat_hist = gr.State([])
            chatbox = gr.Chatbot(height=380, show_label=False)
            with gr.Row():
                chat_in = gr.Textbox(placeholder="Ask anything about research...", label="", scale=5, lines=1)
                chat_btn = gr.Button("Send", elem_classes=["btn-primary"], scale=1)

            ASSISTANT_SYS = """You are ScholarPath, a concise research assistant for health sciences researchers (Gulf/Asian).
Rules: SHORT answers (3-6 sentences, bullet points for lists). Direct and practical.
Portal phases: 1=Gap Analysis, 2=Research Design, 3=Writing, 4=Journals, 5=Reviewer Response.
"""
            def chat_respond(message, history):
                if not message.strip():
                    return history, ""
                resp = ask_gemini(ASSISTANT_SYS + "\nQuestion: " + message, short=True)
                history = history + [[message, resp]]
                return history, ""
            chat_btn.click(chat_respond, [chat_in, chat_hist], [chatbox, chat_in])
            chat_in.submit(chat_respond, [chat_in, chat_hist], [chatbox, chat_in])

        # ── Journal Index ─────────────────────────────────────────────────
        with gr.Tab("📋 Journals"):
            jmd = ""
            for cat, jlist in JOURNALS.items():
                jmd += f"\n#### {cat}\n" + "".join(f"- {j}\n" for j in jlist)
            gr.Markdown(jmd)

    # Floating chatbot (calls Gradio API endpoint)
    gr.HTML(FLOATING_CHAT_HTML)

    gr.HTML("""
<div style="margin-top:20px;padding:18px 24px;background:linear-gradient(135deg,#0a1628,#1e3a5f);
    border-radius:12px;display:flex;justify-content:space-between;align-items:center;flex-wrap:wrap;gap:10px;">
    <div>
        <p style="color:#e8c87a;font-family:'Cormorant Garamond',serif;font-size:17px;margin:0;font-weight:600;">ScholarPath Research Portal</p>
        <p style="color:rgba(255,255,255,.45);font-size:11px;margin:3px 0 0;">Gulf & Asian Health Sciences Researchers</p>
    </div>
    <div style="font-size:11px;color:rgba(255,255,255,.4);">Gemini 2.5 Flash · Semantic Scholar · Free · No data stored externally</div>
</div>
""")

# ── Float chat API endpoint ────────────────────────────────────────────────────
from fastapi import Request as FastRequest
from fastapi.responses import JSONResponse

@demo.app.post("/float_chat")
async def float_chat_endpoint(request: FastRequest):
    body = await request.json()
    q = body.get("q", "")
    if not q:
        return JSONResponse({"answer": "Please ask a question."})
    sys = """You are ScholarPath AI assistant for health sciences researchers.
Give SHORT answers (3-5 sentences max). Bullet points for lists. Direct and practical.
Portal: Phase 1=Gap Analysis, 2=Research Design, 3=Writing, 4=Journal Match, 5=Reviewer Response.
"""
    answer = ask_gemini(sys + "\nQuestion: " + q, short=True)
    return JSONResponse({"answer": answer})

demo.launch(css=CSS)