app.py

"""
RAG SaaS Platform - Streamlit UI
"""
import streamlit as st
import requests
import pandas as pd
import time

# Page Configuration - MUST be first Streamlit command
st.set_page_config(
    page_title="RAG SaaS Platform",
    page_icon="🔍",
    layout="wide",
    initial_sidebar_state="expanded",
)

# Minimal CSS - reduced for faster loading
st.markdown("""
<style>
    .stButton>button {
        width: 100%; border-radius: 5px; height: 3em;
        background-color: #1f6feb; color: white; font-weight: bold;
    }
    .source-badge {
        display: inline-block; padding: 5px 10px; border-radius: 5px;
        margin: 2px; font-size: 12px; font-weight: bold;
    }
    .source-file {background-color: #1f6feb; color: white;}
    .source-web {background-color: #238636; color: white;}
    .source-youtube {background-color: #da3633; color: white;}
    .source-gdrive {background-color: #34a853; color: white;}
    .source-onedrive {background-color: #0078d4; color: white;}
    .source-dropbox {background-color: #0061ff; color: white;}
</style>
""", unsafe_allow_html=True)

# Initialize session state
if 'tenant_id' not in st.session_state:
    st.session_state.tenant_id = None
if 'access_token' not in st.session_state:
    st.session_state.access_token = None
if 'api_url' not in st.session_state:
    st.session_state.api_url = "http://localhost:7860"

# Sidebar Configuration
with st.sidebar:
    st.title("⚙️ Configuration")
    
    # API URL
    api_url = st.text_input(
        "Backend API URL",
        value=st.session_state.api_url,
        help="URL of the RAG SaaS backend"
    )
    st.session_state.api_url = api_url
    
    st.divider()
    
    # Authentication Section
    st.subheader("🔐 Authentication")
    
    if not st.session_state.tenant_id:
        # Login/Register tabs
        tab1, tab2 = st.tabs(["Login", "Register"])
        
        with tab1:
            st.caption("Login with Tenant ID")
            login_tenant_id = st.text_input("Tenant ID", key="login_tenant_id")
            login_username = st.text_input("Username/Email", key="login_username")
            login_password = st.text_input("Password", type="password", key="login_password")
            
            if st.button("Login", key="login_btn"):
                try:
                    response = requests.post(
                        f"{api_url}/v1/auth/login",
                        json={
                            "tenant_id": login_tenant_id,
                            "username": login_username,
                            "password": login_password
                        }
                    )
                    if response.status_code == 200:
                        data = response.json()
                        st.session_state.tenant_id = data['tenant_id']
                        st.session_state.access_token = data['access_token']
                        st.success(f"✅ Logged in as {data['tenant_name']}")
                        st.rerun()
                    else:
                        st.error(f"Login failed: {response.json().get('detail', 'Unknown error')}")
                except Exception as e:
                    st.error(f"Connection error: {e}")
        
        with tab2:
            st.caption("Create New Tenant Account")
            reg_name = st.text_input("Organization Name", key="reg_name")
            reg_email = st.text_input("Admin Email", key="reg_email")
            reg_password = st.text_input("Admin Password", type="password", key="reg_password")
            
            if st.button("Create Tenant", key="register_btn"):
                try:
                    response = requests.post(
                        f"{api_url}/v1/auth/tenants",
                        json={
                            "name": reg_name,
                            "admin_email": reg_email,
                            "admin_password": reg_password
                        }
                    )
                    if response.status_code == 200:
                        data = response.json()
                        st.success(f"✅ Tenant created!")
                        st.info(f"**Your Tenant ID:** `{data['tenant_id']}`\n\n⚠️ Save this ID - you'll need it to login!")
                    else:
                        st.error(f"Registration failed: {response.json().get('detail', 'Unknown error')}")
                except Exception as e:
                    st.error(f"Connection error: {e}")
    else:
        # Logged in state
        st.success(f"✅ Logged in")
        st.caption(f"Tenant ID: `{st.session_state.tenant_id}`")
        
        if st.button("Logout"):
            st.session_state.tenant_id = None
            st.session_state.access_token = None
            st.rerun()

# Main Content
if not st.session_state.tenant_id:
    st.title("🔍 RAG SaaS Platform")
    st.info("👈 Please login or create a tenant account to continue")
    
    # Show platform features
    st.subheader("Platform Features")
    col1, col2, col3 = st.columns(3)
    
    with col1:
        st.markdown("### 📄 Multi-Source Retrieval")
        st.write("Unified search across documents, URLs, and YouTube videos")
    
    with col2:
        st.markdown("### 📊 Tenant Metrics")
        st.write("Isolated metrics tracking for each tenant")
    
    with col3:
        st.markdown("### ⚡ Optimized Performance")
        st.write("Model caching with 70-90% latency reduction")

else:
    # Logged in - show main interface
    st.title(f"🔍 RAG SaaS Platform")
    
    # Main tabs
    tab1, tab2, tab3, tab4, tab5 = st.tabs(["💬 Ask Questions", "📁 Manage Sources", "☁️ Cloud Connectors", "📊 Metrics", "ℹ️ About"])
    
    # Tab 1: Ask Questions
    with tab1:
        st.header("Ask Questions")
        
        # Show active sources
        try:
            response = requests.get(
                f"{api_url}/v1/sources/active",
                headers={"X-Tenant-ID": st.session_state.tenant_id}
            )
            if response.status_code == 200:
                data = response.json()
                active_sources = data.get('active_sources', [])
                source_types = data.get('source_types_present', [])
                
                if source_types:
                    st.success(f"✅ Active sources: {', '.join(source_types)}")
                    
                    # Show source badges
                    badges_html = ""
                    for source_type in source_types:
                        badge_class = f"source-{source_type}"
                        badges_html += f'<span class="source-badge {badge_class}">{source_type.upper()}</span>'
                    st.markdown(badges_html, unsafe_allow_html=True)
                else:
                    st.warning("⚠️ No active sources. Add sources in the 'Manage Sources' tab.")
        except Exception as e:
            st.error(f"Error fetching sources: {e}")
        
        st.divider()
        
        # Question input
        question = st.text_area("Your Question", placeholder="What would you like to know?", height=100)
        
        col1, col2 = st.columns([3, 1])
        with col1:
            top_k = st.slider("Number of results", 5, 20, 10)
        with col2:
            ask_btn = st.button("🔍 Ask", use_container_width=True, type="primary")
        
        if ask_btn and question:
            with st.spinner("🔍 Searching across all sources..."):
                try:
                    # Measure total request time
                    request_start_time = time.time()
                    
                    response = requests.post(
                        f"{api_url}/v1/ask/",
                        headers={"X-Tenant-ID": st.session_state.tenant_id},
                        json={"query": question, "top_k": top_k}
                    )
                    
                    # Calculate total elapsed time
                    total_elapsed_ms = (time.time() - request_start_time) * 1000
                    
                    if response.status_code == 200:
                        data = response.json()
                        
                        # Display answer
                        st.markdown("### 💡 Answer")
                        st.markdown(f"**{data['answer']}**")
                        
                        st.divider()
                        
                        # ===== ENHANCED LATENCY METRICS DISPLAY =====
                        st.markdown("### ⚡ Performance Metrics")
                        
                        # Get latency values from backend
                        retrieval_time = data.get('retrieval_time_ms', 0)
                        generation_time = data.get('generation_time_ms', 0)
                        backend_total = retrieval_time + generation_time
                        network_overhead = total_elapsed_ms - backend_total
                        
                        # Create 5 columns for detailed metrics
                        col1, col2, col3, col4, col5 = st.columns(5)
                        
                        with col1:
                            st.metric(
                                "🔍 Retrieval", 
                                f"{retrieval_time:.0f}ms",
                                help="Time to search across all sources"
                            )
                        
                        with col2:
                            st.metric(
                                "🤖 Generation", 
                                f"{generation_time:.0f}ms",
                                help="Time to generate the answer"
                            )
                        
                        with col3:
                            st.metric(
                                "⚙️ Backend Total", 
                                f"{backend_total:.0f}ms",
                                help="Total backend processing time"
                            )
                        
                        with col4:
                            st.metric(
                                "🌐 Network", 
                                f"{network_overhead:.0f}ms",
                                help="Network latency (request + response)"
                            )
                        
                        with col5:
                            # Color-code total time
                            total_color = "🟢" if total_elapsed_ms < 500 else "🟡" if total_elapsed_ms < 1000 else "🔴"
                            st.metric(
                                f"{total_color} Total Time", 
                                f"{total_elapsed_ms:.0f}ms",
                                help="End-to-end response time"
                            )
                        
                        # Visual latency breakdown
                        st.markdown("#### Latency Breakdown")
                        latency_data = {
                            "Retrieval": retrieval_time,
                            "Generation": generation_time,
                            "Network": max(0, network_overhead)
                        }
                        
                        # Create a simple bar chart representation
                        total_for_percentage = sum(latency_data.values())
                        if total_for_percentage > 0:
                            breakdown_html = '<div style="display: flex; width: 100%; height: 30px; border-radius: 5px; overflow: hidden;">'
                            
                            colors = {
                                "Retrieval": "#1f6feb",
                                "Generation": "#238636", 
                                "Network": "#da3633"
                            }
                            
                            for component, time_ms in latency_data.items():
                                percentage = (time_ms / total_for_percentage) * 100
                                color = colors.get(component, "#666")
                                breakdown_html += f'<div style="background-color: {color}; width: {percentage}%; display: flex; align-items: center; justify-content: center; color: white; font-size: 11px; font-weight: bold;" title="{component}: {time_ms:.0f}ms ({percentage:.1f}%)">{component[:3]}</div>'
                            
                            breakdown_html += '</div>'
                            st.markdown(breakdown_html, unsafe_allow_html=True)
                            
                            # Show percentage breakdown
                            breakdown_text = " | ".join([
                                f"{comp}: {(time_ms/total_for_percentage)*100:.1f}% ({time_ms:.0f}ms)" 
                                for comp, time_ms in latency_data.items()
                            ])
                            st.caption(breakdown_text)
                        
                        st.divider()
                        
                        # Display confidence
                        confidence = data.get('confidence', 0)
                        confidence_color = "🟢" if confidence > 0.7 else "🟡" if confidence > 0.4 else "🔴"
                        st.metric(
                            f"{confidence_color} Answer Confidence", 
                            f"{confidence:.1%}",
                            help="Model's confidence in the answer quality"
                        )
                        
                        # Display sources used
                        if 'sources_used' in data and data['sources_used']:
                            st.markdown("### 📚 Sources Used")
                            sources_html = ""
                            for source_type in data['sources_used']:
                                badge_class = f"source-{source_type}"
                                sources_html += f'<span class="source-badge {badge_class}">{source_type.upper()}</span>'
                            st.markdown(sources_html, unsafe_allow_html=True)
                            st.caption(f"Retrieved from {len(data['sources_used'])} different source type(s)")
                        
                        # Display citations
                        if data.get('citations'):
                            st.markdown("### 📖 Citations")
                            for i, citation in enumerate(data['citations'], 1):
                                with st.expander(f"Citation {i} - {citation.get('source_type', 'unknown').upper()}"):
                                    st.json(citation)
                    else:
                        st.error(f"❌ Error: {response.json().get('detail', 'Unknown error')}")
                except Exception as e:
                    st.error(f"❌ Request failed: {e}")
                    st.exception(e)
    
    # Tab 2: Manage Sources
    with tab2:
        st.header("Manage Sources")
        
        # Sub-tabs for different source types
        source_tab1, source_tab2, source_tab3, source_tab4 = st.tabs(
            ["📄 Documents", "🌐 URLs", "🎥 YouTube", "📋 All Sources"]
        )
        
        with source_tab1:
            st.subheader("Upload Documents")
            st.info("Upload PDF, DOCX, or TXT files")
            uploaded_file = st.file_uploader("Choose a file", type=['pdf', 'docx', 'txt'])
            
            if uploaded_file and st.button("Upload & Ingest"):
                with st.spinner("Uploading and ingesting..."):
                    try:
                        files = {"file": uploaded_file}
                        response = requests.post(
                            f"{api_url}/v1/ingest/upload",
                            headers={"X-Tenant-ID": st.session_state.tenant_id},
                            files=files
                        )
                        if response.status_code == 200:
                            st.success("✅ Document uploaded and ingested successfully!")
                        else:
                            st.error(f"Upload failed: {response.json().get('detail', 'Unknown error')}")
                    except Exception as e:
                        st.error(f"Upload error: {e}")
        
        with source_tab2:
            st.subheader("Add Web URLs")
            url = st.text_input("Enter URL", placeholder="https://example.com/article")
            url_name = st.text_input("Source Name (optional)", placeholder="Company Blog")
            
            if st.button("Add URL") and url:
                with st.spinner("Adding URL..."):
                    try:
                        response = requests.post(
                            f"{api_url}/v1/ingest/url",
                            headers={"X-Tenant-ID": st.session_state.tenant_id},
                            json={"url": url, "name": url_name or url}
                        )
                        if response.status_code == 200:
                            st.success("✅ URL added and ingestion started!")
                        else:
                            st.error(f"Failed: {response.json().get('detail', 'Unknown error')}")
                    except Exception as e:
                        st.error(f"Error: {e}")
        
        with source_tab3:
            st.subheader("Add YouTube Videos")
            st.info("✅ Captions-first strategy with transcript upload fallback")
            
            youtube_url = st.text_input("YouTube URL", placeholder="https://youtube.com/watch?v=...")
            video_name = st.text_input("Video Name (optional)")
            
            transcript_file = st.file_uploader(
                "Upload Transcript (optional)",
                type=['vtt', 'srt', 'txt'],
                help="If video has no captions, upload a transcript file"
            )
            
            owner_attestation = st.checkbox(
                "I attest that I have rights to use this content",
                help="Required: Confirm you own or have permission to use this video"
            )
            
            if st.button("Add YouTube Video") and youtube_url and owner_attestation:
                with st.spinner("Adding YouTube video..."):
                    try:
                        files = {}
                        if transcript_file:
                            files['transcript_file'] = transcript_file
                        
                        data = {
                            'youtube_url': youtube_url,
                            'name': video_name or youtube_url,
                            'owner_attestation': str(owner_attestation).lower()
                        }
                        
                        response = requests.post(
                            f"{api_url}/v1/sources/youtube",
                            headers={"X-Tenant-ID": st.session_state.tenant_id},
                            data=data,
                            files=files if files else None
                        )
                        
                        if response.status_code == 200:
                            result = response.json()
                            st.success(f"✅ YouTube source added! Strategy: {result.get('strategy', 'captions_api')}")
                        else:
                            st.error(f"Failed: {response.json().get('detail', 'Unknown error')}")
                    except Exception as e:
                        st.error(f"Error: {e}")
        
        with source_tab4:
            st.subheader("All Sources")
            
            try:
                response = requests.get(
                    f"{api_url}/v1/sources/",
                    headers={"X-Tenant-ID": st.session_state.tenant_id}
                )
                
                if response.status_code == 200:
                    sources = response.json()
                    
                    if sources:
                        for source in sources:
                            with st.expander(f"{source['name']} ({source['type']})"):
                                st.write(f"**ID:** {source['id']}")
                                st.write(f"**Type:** {source['type']}")
                                st.write(f"**URI:** {source['uri']}")
                                st.write(f"**Created:** {source.get('created_at', 'N/A')}")
                                
                                # Toggle source
                                col1, col2 = st.columns(2)
                                with col1:
                                    if st.button(f"Disable", key=f"disable_{source['id']}"):
                                        requests.put(
                                            f"{api_url}/v1/sources/{source['id']}/toggle?enabled=false",
                                            headers={"X-Tenant-ID": st.session_state.tenant_id}
                                        )
                                        st.rerun()
                                with col2:
                                    if st.button(f"Enable", key=f"enable_{source['id']}"):
                                        requests.put(
                                            f"{api_url}/v1/sources/{source['id']}/toggle?enabled=true",
                                            headers={"X-Tenant-ID": st.session_state.tenant_id}
                                        )
                                        st.rerun()
                    else:
                        st.info("No sources yet. Add sources using the tabs above.")
            except Exception as e:
                st.error(f"Error loading sources: {e}")
    
    # Tab 3: Cloud Connectors
    with tab3:
        st.header("☁️ Cloud Connectors")
        st.write("Connect to your enterprise cloud storage to sync documents directly.")
        
        # Connect New Account
        st.subheader("Connect New Account")
        col1, col2, col3 = st.columns(3)
        
        with col1:
            if st.button("🔌 Connect Google Drive", use_container_width=True):
                try:
                    res = requests.get(f"{api_url}/v1/connectors/gdrive/oauth/start?tenant_id={st.session_state.tenant_id}")
                    if res.status_code == 200:
                        st.markdown(f"[Click here to authorize Google Drive]({res.json()['auth_url']})")
                    else:
                        st.error("Failed to start GDrive OAuth")
                except Exception as e:
                    st.error(f"Error: {e}")
        
        with col2:
            if st.button("🔌 Connect OneDrive", use_container_width=True):
                try:
                    res = requests.get(f"{api_url}/v1/connectors/onedrive/oauth/start?tenant_id={st.session_state.tenant_id}")
                    if res.status_code == 200:
                        st.markdown(f"[Click here to authorize OneDrive]({res.json()['auth_url']})")
                    else:
                        st.error("Failed to start OneDrive OAuth")
                except Exception as e:
                    st.error(f"Error: {e}")
        
        with col3:
            if st.button("🔌 Connect Dropbox", use_container_width=True):
                try:
                    res = requests.get(f"{api_url}/v1/connectors/dropbox/oauth/start?tenant_id={st.session_state.tenant_id}")
                    if res.status_code == 200:
                        st.markdown(f"[Click here to authorize Dropbox]({res.json()['auth_url']})")
                    else:
                        st.error("Failed to start Dropbox OAuth")
                except Exception as e:
                    st.error(f"Error: {e}")
        
        st.divider()
        
        # Manage Existing Connectors
        st.subheader("Your Connected Accounts")
        try:
            response = requests.get(
                f"{api_url}/v1/connectors/",
                params={"tenant_id": st.session_state.tenant_id}
            )
            if response.status_code == 200:
                connectors = response.json()
                if connectors:
                    for conn in connectors:
                        with st.expander(f"{conn['provider'].upper()} - {conn['display_name']}"):
                            col_a, col_b, col_c = st.columns([2, 1, 1])
                            with col_a:
                                st.write(f"**Status:** {conn['sync_status'].upper()}")
                                if conn['last_sync']:
                                    st.write(f"**Last Sync:** {conn['last_sync']}")
                            
                            with col_b:
                                if st.button("🔄 Trigger Sync", key=f"sync_{conn['id']}"):
                                    sync_res = requests.post(f"{api_url}/v1/connectors/{conn['id']}/sync")
                                    if sync_res.status_code == 200:
                                        st.success("Sync triggered!")
                                    else:
                                        st.error("Failed to trigger sync")
                            
                            with col_c:
                                if st.button("📁 Manage Targets", key=f"targets_{conn['id']}"):
                                    st.session_state[f"show_targets_{conn['id']}"] = True
                            
                            # Target Selection Mock/Simplified
                            if st.session_state.get(f"show_targets_{conn['id']}"):
                                st.info("Loading roots...")
                                roots_res = requests.get(f"{api_url}/v1/connectors/{conn['id']}/roots")
                                if roots_res.status_code == 200:
                                    roots = roots_res.json()
                                    target_ids = st.multiselect(
                                        "Select Folders/Drives to sync",
                                        options=[r['id'] for r in roots],
                                        format_func=lambda x: next(r['name'] for r in roots if r['id'] == x),
                                        key=f"ms_{conn['id']}"
                                    )
                                    if st.button("Save Targets", key=f"save_{conn['id']}"):
                                        selected_targets = [{"id": tid, "type": "folder"} for tid in target_ids]
                                        save_res = requests.post(
                                            f"{api_url}/v1/connectors/{conn['id']}/targets",
                                            params={"tenant_id": st.session_state.tenant_id},
                                            json=selected_targets
                                        )
                                        if save_res.status_code == 200:
                                            st.success("Targets updated!")
                                            st.session_state[f"show_targets_{conn['id']}"] = False
                                            st.rerun()
                else:
                    st.info("No cloud accounts connected yet.")
        except Exception as e:
            st.error(f"Error fetching connectors: {e}")

    # Tab 4: Metrics
    with tab4:
        st.header("📊 Your Metrics")
        
        # Time range selector
        hours = st.selectbox("Time Range", [1, 6, 12, 24, 48, 168], index=3, format_func=lambda x: f"Last {x} hours")
        
        if st.button("Refresh Metrics"):
            try:
                response = requests.get(
                    f"{api_url}/v1/tenants/{st.session_state.tenant_id}/metrics?hours={hours}",
                    headers={"X-Tenant-ID": st.session_state.tenant_id}
                )
                
                if response.status_code == 200:
                    metrics = response.json()
                    counters = metrics.get('counters', {})
                    
                    # Display key metrics
                    col1, col2, col3, col4 = st.columns(4)
                    
                    with col1:
                        st.metric("Total Retrievals", counters.get('retrieval_total', 0))
                    with col2:
                        st.metric("Total Generations", counters.get('generation_total', 0))
                    with col3:
                        st.metric("Sources Created", counters.get('sources_created', 0))
                    with col4:
                        st.metric("Ingestions", counters.get('ingestion_total', 0))
                    
                    st.divider()
                    
                    # Show all counters
                    st.subheader("All Metrics")
                    if counters:
                        df = pd.DataFrame(list(counters.items()), columns=['Metric', 'Value'])
                        st.dataframe(df, use_container_width=True)
                    else:
                        st.info("No metrics data yet. Start using the platform to see metrics!")
                    
                    # Time series data
                    if metrics.get('time_series'):
                        st.subheader("Time Series Data")
                        for metric_name, data_points in metrics['time_series'].items():
                            if data_points:
                                st.write(f"**{metric_name}**")
                                df = pd.DataFrame(data_points)
                                st.line_chart(df.set_index('timestamp')['value'])
                else:
                    st.error(f"Failed to load metrics: {response.json().get('detail', 'Unknown error')}")
            except Exception as e:
                st.error(f"Error: {e}")
        
        st.divider()
        
        # Cache statistics
        st.subheader("🚀 Performance Metrics")
        try:
            response = requests.get(f"{api_url}/v1/metrics/cache-stats")
            if response.status_code == 200:
                cache_stats = response.json()
                
                col1, col2, col3 = st.columns(3)
                with col1:
                    st.metric("Cache Hit Rate", f"{cache_stats.get('hit_rate', 0):.1%}")
                with col2:
                    st.metric("Cache Size", f"{cache_stats.get('cache_size', 0)}/{cache_stats.get('cache_max_size', 0)}")
                with col3:
                    st.metric("Model Load Time", f"{cache_stats.get('model_load_time_seconds', 0):.1f}s")
        except Exception as e:
            st.warning(f"Cache stats unavailable: {e}")
    
    # Tab 5: About
    with tab5:
        st.header("ℹ️ About RAG SaaS Platform")
        
        st.markdown("""
        ### Features
        
        ✅ **Multi-Source Retrieval** - Unified search across documents, URLs, and YouTube videos  
        ✅ **Enterprise Cloud Connectors** - Native sync with Google Drive, OneDrive, and Dropbox  
        ✅ **Tenant Isolation** - Complete data and metrics separation per tenant  
        ✅ **Optimized Performance** - Model caching with 70-90% latency reduction  
        ✅ **Compliant YouTube** - Captions-first with transcript upload fallback  
        ✅ **Comprehensive Metrics** - Track all your usage and performance  
        
        ### How It Works
        
        1. **Add Sources** - Upload documents, add URLs, or YouTube videos
        2. **Ask Questions** - The system retrieves from ALL your sources
        3. **Get Answers** - Receive answers with citations from multiple sources
        4. **Track Metrics** - Monitor your usage and performance
        
        ### Technical Details
        
        - **Backend:** FastAPI with SQLite/PostgreSQL
        - **Vector Store:** Local (upgradeable to Qdrant/Pgvector)
        - **Models:** Local CPU-optimized (Qwen 0.5B + MiniLM)
        - **Retrieval:** RRF fusion with diversity constraints
        """)