app.py

import streamlit as st
import os
import tempfile
import gc
import base64
import time
import requests
from datetime import datetime
from groq import Groq
from src.agentic_rag.tools.custom_tool import DocumentSearchTool

st.set_page_config(page_title="NeuralDocs — Agentic RAG", page_icon="🧠", layout="wide", initial_sidebar_state="expanded")

st.markdown("""
<style>
@import url('https://fonts.googleapis.com/css2?family=Syne:wght@400;600;700;800&family=JetBrains+Mono:wght@400;500&family=Figtree:wght@300;400;500;600&display=swap');
:root {
    --bg-base:#080c12;--bg-surface:#0d1520;--bg-elevated:#121d2e;
    --border:#1e3048;--border-bright:#2a4460;
    --accent:#00e5ff;--accent-dim:#0097a7;--accent-glow:rgba(0,229,255,0.12);
    --accent2:#7c5cfc;--text-primary:#e8f4f8;--text-secondary:#7a9bb5;--text-muted:#3d5a75;
    --success:#00c896;--warning:#ffb700;
    --font-display:'Syne',sans-serif;--font-body:'Figtree',sans-serif;--font-mono:'JetBrains Mono',monospace;
    --radius-sm:6px;--radius-md:12px;--radius-lg:18px;
}
html,body,[class*="css"]{font-family:var(--font-body);background-color:var(--bg-base)!important;color:var(--text-primary)!important;}
#MainMenu,footer,header{visibility:hidden;}.stDeployButton{display:none;}
.main .block-container{background:var(--bg-base);padding:1.5rem 2rem 3rem;max-width:1100px;}
.app-header{display:flex;align-items:center;gap:14px;padding:1.2rem 0 0.6rem;border-bottom:1px solid var(--border);margin-bottom:1.5rem;}
.app-logo{width:38px;height:38px;background:linear-gradient(135deg,var(--accent),var(--accent2));border-radius:10px;display:flex;align-items:center;justify-content:center;font-size:18px;font-weight:800;color:#000;flex-shrink:0;}
.app-title{font-family:var(--font-display);font-size:1.5rem;font-weight:800;background:linear-gradient(135deg,var(--accent),var(--accent2));-webkit-background-clip:text;-webkit-text-fill-color:transparent;background-clip:text;line-height:1;}
.app-subtitle{font-size:0.7rem;color:var(--text-muted);letter-spacing:2px;text-transform:uppercase;font-family:var(--font-mono);margin-top:3px;}
.header-badge{margin-left:auto;background:var(--accent-glow);border:1px solid var(--accent-dim);color:var(--accent);font-family:var(--font-mono);font-size:0.65rem;padding:4px 10px;border-radius:20px;letter-spacing:1px;}
[data-testid="stSidebar"]{background:var(--bg-surface)!important;border-right:1px solid var(--border)!important;}
[data-testid="stSidebar"] .block-container{background:var(--bg-surface)!important;padding:1.5rem 1.2rem;}
.sidebar-section-label{font-family:var(--font-mono);font-size:0.6rem;letter-spacing:3px;text-transform:uppercase;color:var(--text-muted);margin:1.2rem 0 0.6rem;padding-bottom:0.4rem;border-bottom:1px solid var(--border);}
.status-pill{display:inline-flex;align-items:center;gap:6px;padding:5px 12px;border-radius:20px;font-family:var(--font-mono);font-size:0.7rem;margin:6px 0;}
.status-ready{background:rgba(0,200,150,0.1);border:1px solid var(--success);color:var(--success);}
.status-waiting{background:rgba(255,183,0,0.1);border:1px solid var(--warning);color:var(--warning);}
.status-dot{width:6px;height:6px;border-radius:50%;background:currentColor;}
.status-dot.pulse{animation:pulse 1.5s infinite;}
@keyframes pulse{0%,100%{opacity:1;}50%{opacity:0.3;}}
.metrics-row{display:grid;grid-template-columns:1fr 1fr;gap:8px;margin:10px 0;}
.metric-card{background:var(--bg-elevated);border:1px solid var(--border);border-radius:var(--radius-md);padding:10px 14px;}
.metric-value{font-family:var(--font-display);font-size:1.3rem;font-weight:700;color:var(--accent);line-height:1;}
.metric-label{font-family:var(--font-mono);font-size:0.6rem;color:var(--text-muted);letter-spacing:1px;text-transform:uppercase;margin-top:3px;}
.stChatMessage{background:var(--bg-surface)!important;border:1px solid var(--border)!important;border-radius:var(--radius-lg)!important;padding:1rem 1.2rem!important;margin-bottom:10px!important;}
[data-testid="stChatMessageContent"] p{font-family:var(--font-body);font-size:0.9rem;line-height:1.7;color:var(--text-primary);}
.answer-meta{display:flex;align-items:center;gap:12px;padding:8px 0 4px;border-top:1px solid var(--border);margin-top:10px;flex-wrap:wrap;}
.confidence-bar-wrap{display:flex;align-items:center;gap:6px;}
.confidence-bar{width:60px;height:4px;background:var(--border);border-radius:2px;overflow:hidden;}
.confidence-fill{height:100%;border-radius:2px;background:linear-gradient(90deg,var(--accent2),var(--accent));}
[data-testid="stChatInput"]{background:var(--bg-elevated)!important;border:1px solid var(--border-bright)!important;border-radius:var(--radius-lg)!important;}
[data-testid="stChatInput"] textarea{color:var(--text-primary)!important;font-family:var(--font-body)!important;background:transparent!important;}
.stButton>button{background:var(--bg-elevated)!important;border:1px solid var(--border-bright)!important;color:var(--text-secondary)!important;border-radius:var(--radius-sm)!important;font-family:var(--font-mono)!important;font-size:0.72rem!important;}
.stButton>button:hover{border-color:var(--accent)!important;color:var(--accent)!important;background:var(--accent-glow)!important;}
[data-testid="stFileUploader"]{background:var(--bg-elevated)!important;border:1px dashed var(--border-bright)!important;border-radius:var(--radius-md)!important;}
.history-item{background:var(--bg-elevated);border:1px solid var(--border);border-left:3px solid var(--accent-dim);border-radius:var(--radius-sm);padding:8px 10px;margin:4px 0;}
.history-q{font-size:0.78rem;color:var(--text-secondary);white-space:nowrap;overflow:hidden;text-overflow:ellipsis;}
.history-time{font-family:var(--font-mono);font-size:0.6rem;color:var(--text-muted);margin-top:2px;}
.empty-state{display:flex;flex-direction:column;align-items:center;justify-content:center;padding:3rem 1rem;text-align:center;gap:12px;}
.empty-icon{font-size:2.5rem;opacity:0.3;}
.empty-text{font-family:var(--font-body);color:var(--text-muted);font-size:0.9rem;max-width:320px;line-height:1.6;}
</style>
""", unsafe_allow_html=True)

# ── Session State ─────────────────────────────────────────────────────────────
defaults = {
    "messages": [], "pdf_tool": None, "pdf_name": None,
    "total_queries": 0, "pdf_hits": 0, "web_hits": 0,
    "query_log": [], "pending_followups": [], "pending_prompt": None,
}
for k, v in defaults.items():
    if k not in st.session_state:
        st.session_state[k] = v

def reset_chat():
    st.session_state.messages = []
    st.session_state.pending_followups = []
    gc.collect()

def full_reset():
    for k in list(defaults.keys()):
        st.session_state[k] = defaults[k]
    gc.collect()

def display_pdf(file_bytes, file_name):
    b64 = base64.b64encode(file_bytes).decode("utf-8")
    st.markdown(f'<iframe src="data:application/pdf;base64,{b64}" width="100%" height="480px" style="border:1px solid var(--border);border-radius:var(--radius-md);"></iframe>', unsafe_allow_html=True)

def web_search(query: str) -> str:
    api_key = os.environ.get("SERPER_API_KEY", "")
    if not api_key:
        return ""
    try:
        response = requests.post(
            "https://google.serper.dev/search",
            headers={"X-API-KEY": api_key, "Content-Type": "application/json"},
            json={"q": query, "num": 3}, timeout=10,
        )
        data = response.json()
        results = []
        for r in data.get("organic", [])[:3]:
            results.append(f"{r.get('title','')}: {r.get('snippet','')}")
        return "\n".join(results)
    except:
        return ""

def detect_source(result_text):
    web_signals = ["according to", "website", "http", "search result", "online"]
    return "web" if any(s in result_text.lower() for s in web_signals) else "pdf"

def estimate_confidence(result_text):
    if "i'm sorry" in result_text.lower() or "couldn't find" in result_text.lower(): return 20
    if len(result_text) > 400: return 88
    if len(result_text) > 150: return 72
    return 55

def generate_followups(query, answer):
    keywords = [w for w in query.lower().split() if len(w) > 4]
    return [
        f"Can you elaborate on {keywords[0] if keywords else 'this'} in more detail?",
        "What are the practical applications of this?",
        "What are the limitations or challenges here?",
    ]

def run_query(prompt):
    """Direct Groq call — no CrewAI overhead, much faster."""
    client = Groq(api_key=os.environ.get("GROQ_API_KEY"))

    # Step 1: Try PDF search
    context = ""
    source = "web"
    if st.session_state.pdf_tool:
        try:
            pdf_result = st.session_state.pdf_tool._run(prompt)
            if pdf_result and "No relevant" not in pdf_result:
                context = f"Document context:\n{pdf_result}\n\n"
                source = "pdf"
        except:
            pass

    # Step 2: If no PDF context, search web
    if not context:
        web_result = web_search(prompt)
        if web_result:
            context = f"Web search results:\n{web_result}\n\n"
            source = "web"

    # Step 3: Ask Groq directly
    system_prompt = "You are a helpful research assistant. Answer questions concisely based on the provided context. If no context is provided, answer from your knowledge. Keep answers under 200 words."
    user_message = f"{context}Question: {prompt}"

    response = client.chat.completions.create(
        model="openai/gpt-oss-20b",
        messages=[
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": user_message}
        ],
        max_tokens=1000,
        temperature=0.1,
        reasoning_effort="low",
    )
    return response.choices[0].message.content, source

# ── Sidebar ───────────────────────────────────────────────────────────────────
with st.sidebar:
    st.markdown("""
    <div style="padding:0.2rem 0 1rem;">
        <div style="display:flex;align-items:center;gap:10px;margin-bottom:4px;">
            <div style="width:30px;height:30px;background:linear-gradient(135deg,#00e5ff,#7c5cfc);border-radius:8px;display:flex;align-items:center;justify-content:center;font-weight:800;color:#000;font-size:14px;">🧠</div>
            <span style="font-family:'Syne',sans-serif;font-size:1rem;font-weight:700;background:linear-gradient(135deg,#00e5ff,#7c5cfc);-webkit-background-clip:text;-webkit-text-fill-color:transparent;">NeuralDocs</span>
        </div>
        <div style="font-family:'JetBrains Mono',monospace;font-size:0.6rem;letter-spacing:2px;color:#3d5a75;text-transform:uppercase;">Agentic RAG · Groq + LLaMA 3</div>
    </div>
    """, unsafe_allow_html=True)

    st.markdown('<div class="sidebar-section-label">📄 Document</div>', unsafe_allow_html=True)
    uploaded_file = st.file_uploader("Upload PDF", type=["pdf"], label_visibility="collapsed")
    if uploaded_file is not None:
        if st.session_state.pdf_tool is None or st.session_state.pdf_name != uploaded_file.name:
            with tempfile.TemporaryDirectory() as tmp:
                path = os.path.join(tmp, uploaded_file.name)
                with open(path, "wb") as f:
                    f.write(uploaded_file.getvalue())
                with st.spinner("Indexing document…"):
                    st.session_state.pdf_tool = DocumentSearchTool(file_path=path)
                    st.session_state.pdf_name = uploaded_file.name
            st.session_state.messages = []
            st.session_state.pending_followups = []
        st.markdown(f'<div class="status-pill status-ready"><span class="status-dot"></span>{uploaded_file.name[:28]}{"…" if len(uploaded_file.name)>28 else ""}</div>', unsafe_allow_html=True)
        with st.expander("Preview PDF", expanded=False):
            display_pdf(uploaded_file.getvalue(), uploaded_file.name)
    else:
        st.markdown('<div class="status-pill status-waiting"><span class="status-dot pulse"></span>No document loaded</div>', unsafe_allow_html=True)

    st.markdown('<div class="sidebar-section-label">📊 Session Stats</div>', unsafe_allow_html=True)
    total = st.session_state.total_queries or 0
    pdf_h = st.session_state.pdf_hits or 0
    web_h = st.session_state.web_hits or 0
    pdf_pct = int((pdf_h/total)*100) if total else 0
    web_pct = int((web_h/total)*100) if total else 0
    st.markdown(f"""
    <div class="metrics-row">
        <div class="metric-card"><div class="metric-value">{total}</div><div class="metric-label">Queries</div></div>
        <div class="metric-card"><div class="metric-value" style="color:var(--success)">{pdf_pct}%</div><div class="metric-label">PDF hits</div></div>
    </div>
    <div class="metrics-row">
        <div class="metric-card"><div class="metric-value" style="color:var(--accent2)">{web_pct}%</div><div class="metric-label">Web hits</div></div>
        <div class="metric-card"><div class="metric-value" style="color:var(--warning)">{len(st.session_state.messages)//2}</div><div class="metric-label">Exchanges</div></div>
    </div>
    """, unsafe_allow_html=True)

    if st.session_state.query_log:
        st.markdown('<div class="sidebar-section-label">🕓 Query History</div>', unsafe_allow_html=True)
        for item in reversed(st.session_state.query_log[-6:]):
            src_icon = "🌐" if item["source"] == "web" else "📄"
            st.markdown(f'<div class="history-item"><div class="history-q">{src_icon} {item["q"][:48]}{"…" if len(item["q"])>48 else ""}</div><div class="history-time">{item["ts"]}</div></div>', unsafe_allow_html=True)

    st.markdown('<div class="sidebar-section-label">⚙ Controls</div>', unsafe_allow_html=True)
    col1, col2 = st.columns(2)
    with col1: st.button("Clear Chat", on_click=reset_chat, use_container_width=True)
    with col2: st.button("Full Reset", on_click=full_reset, use_container_width=True)

# ── Main ──────────────────────────────────────────────────────────────────────
st.markdown("""
<div class="app-header">
    <div class="app-logo">🧠</div>
    <div>
        <div class="app-title">NeuralDocs</div>
        <div class="app-subtitle">Agentic RAG · PDF + Web Intelligence</div>
    </div>
    <div class="header-badge">GROQ · LLAMA 3</div>
</div>
""", unsafe_allow_html=True)

if not st.session_state.messages:
    st.markdown("""
    <div class="empty-state">
        <div class="empty-icon">🧠</div>
        <div class="empty-text">Upload a PDF in the sidebar, then ask anything.<br>The agent searches your document first — then the web.</div>
    </div>
    """, unsafe_allow_html=True)

for i, message in enumerate(st.session_state.messages):
    with st.chat_message(message["role"]):
        st.markdown(message["content"])
        if message["role"] == "assistant" and i == len(st.session_state.messages) - 1 and st.session_state.pending_followups:
            conf = estimate_confidence(message["content"])
            src = message.get("source", "pdf")
            st.markdown(f"""
            <div class="answer-meta">
                <span style="font-family:var(--font-mono);font-size:0.65rem;color:var(--accent);">{'🌐 Web' if src=='web' else '📄 PDF'}</span>
                <div class="confidence-bar-wrap" style="display:flex;align-items:center;gap:6px;font-family:var(--font-mono);font-size:0.65rem;color:var(--text-muted);">
                    <span>Confidence</span>
                    <div class="confidence-bar"><div class="confidence-fill" style="width:{conf}%"></div></div>
                    <span>{conf}%</span>
                </div>
            </div>
            """, unsafe_allow_html=True)
            st.markdown('<div style="font-family:var(--font-mono);font-size:0.6rem;color:var(--text-muted);margin:8px 0 4px;">💡 FOLLOW-UP SUGGESTIONS</div>', unsafe_allow_html=True)
            cols = st.columns(len(st.session_state.pending_followups))
            for j, (col, q) in enumerate(zip(cols, st.session_state.pending_followups)):
                with col:
                    if st.button(q, key=f"followup_{i}_{j}", use_container_width=True):
                        st.session_state.pending_prompt = q

def handle_query(prompt):
    st.session_state.pending_followups = []
    st.session_state.messages.append({"role": "user", "content": prompt})
    with st.chat_message("user"):
        st.markdown(prompt)
    with st.chat_message("assistant"):
        with st.spinner("Researching…"):
            result, src = run_query(prompt)
        st.markdown(result)
    ts = datetime.now().strftime("%H:%M")
    st.session_state.total_queries += 1
    if src == "pdf": st.session_state.pdf_hits += 1
    else: st.session_state.web_hits += 1
    st.session_state.query_log.append({"q": prompt, "ts": ts, "source": src})
    st.session_state.messages.append({"role": "assistant", "content": result, "source": src})
    st.session_state.pending_followups = generate_followups(prompt, result)
    st.rerun()

if st.session_state.pending_prompt:
    prompt = st.session_state.pending_prompt
    st.session_state.pending_prompt = None
    handle_query(prompt)

prompt = st.chat_input("Ask anything about your document or search the web…")
if prompt:
    handle_query(prompt)