app.py

"""
FastAPI Application for Multimodal RAG System
US Army Medical Research Papers Q&A
"""

"""
FastAPI Application for Multimodal RAG System
US Army Medical Research Papers Q&A
"""

import os
import io
import logging
from typing import List, Dict, Optional, Union
from contextlib import asynccontextmanager
from datetime import datetime

from fastapi import FastAPI, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import FileResponse, StreamingResponse
from fastapi.staticfiles import StaticFiles
from pydantic import BaseModel, Field

# Import from query_index (standalone)
from query_index import MultimodalRAGSystem

# PDF generation libraries
from reportlab.lib.pagesizes import letter
from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle
from reportlab.lib.units import inch
from reportlab.lib.enums import TA_CENTER, TA_JUSTIFY
from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer, PageBreak, Image as RLImage, Table, TableStyle
from reportlab.lib import colors
import requests

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
)
logger = logging.getLogger(__name__)

# Global variables
rag_system: Optional[MultimodalRAGSystem] = None

# Keep short conversation history 
chat_history: List[Dict[str, str]] = []
MAX_HISTORY_TURNS = 3

# Lifecycle management
@asynccontextmanager
async def lifespan(app: FastAPI):
    """Initialize and cleanup RAG system"""
    global rag_system
    
    logger.info("Starting RAG system initialization...")
    try:
        rag_system = MultimodalRAGSystem()
        logger.info("RAG system initialized successfully!")
    except Exception as e:
        logger.error(f"Error during initialization: {str(e)}")
        rag_system = None
    
    yield
    
    logger.info("Shutting down RAG system...")
    rag_system = None

# Create FastAPI app
app = FastAPI(
    title="Multimodal RAG API",
    description="Q&A system for US Army medical research papers (Text + Images)",
    version="2.0.0",
    lifespan=lifespan
)

# CORS middleware
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# Mount static files
app.mount("/static", StaticFiles(directory="static"), name="static")

# Mount extracted images
# This allows the frontend to load images via /extracted_images/filename.jpg
if os.path.exists("extracted_images"):
    app.mount("/extracted_images", StaticFiles(directory="extracted_images"), name="images")

# Mount PDF documents
if os.path.exists("WHEC_Documents"):
    app.mount("/documents", StaticFiles(directory="WHEC_Documents"), name="documents")

# Pydantic models
class QueryRequest(BaseModel):
    question: str = Field(..., min_length=1, max_length=1000, description="Question to ask")

class ImageSource(BaseModel):
    path: Optional[str]
    filename: Optional[str]
    score: Optional[float]
    page: Optional[Union[str, int]]
    file: Optional[str]
    link: Optional[str] = None

class TextSource(BaseModel):
    text: str
    score: float
    page: Optional[Union[str, int]]
    file: Optional[str]
    link: Optional[str] = None

class QueryResponse(BaseModel):
    answer: str
    images: List[ImageSource]
    texts: List[TextSource]
    question: str

class HealthResponse(BaseModel):
    status: str
    rag_initialized: bool

class ConversationItem(BaseModel):
    question: str
    answer: str
    images: List[ImageSource]
    texts: List[TextSource]
    timestamp: str

class ReportRequest(BaseModel):
    conversations: List[ConversationItem]

# API Endpoints

@app.get("/", tags=["Root"])
async def root():
    """Serve the frontend application"""
    return FileResponse('static/index.html')

@app.get("/health", response_model=HealthResponse, tags=["Health"])
async def health_check():
    """Health check endpoint"""
    return HealthResponse(
        status="healthy",
        rag_initialized=rag_system is not None
    )

@app.post("/query", response_model=QueryResponse, tags=["Query"])
async def query_rag(request: QueryRequest):
    global chat_history

    if not rag_system:
        raise HTTPException(status_code=503, detail="RAG system not initialized.")

    try:
        result = rag_system.ask(
            query_str=request.question,
            chat_history=chat_history
        )

        # Update history
        chat_history.append({"role": "user", "content": request.question})
        chat_history.append({"role": "assistant", "content": result["answer"]})

        # Trim history safely
        if len(chat_history) > MAX_HISTORY_TURNS * 2:
            chat_history = chat_history[-MAX_HISTORY_TURNS * 2:]

        return QueryResponse(
            answer=result["answer"],
            images=result["images"],
            texts=result["texts"],
            question=request.question
        )

    except Exception as e:
        logger.error(f"Error processing query: {str(e)}")
        raise HTTPException(status_code=500, detail=str(e))

@app.post("/generate-report", tags=["Report"])
async def generate_report(request: ReportRequest):
    """Generate a PDF report from conversation data"""
    try:
        buffer = io.BytesIO()
        
        # Create PDF document
        doc = SimpleDocTemplate(
            buffer,
            pagesize=letter,
            rightMargin=0.75*inch,
            leftMargin=0.75*inch,
            topMargin=1*inch,
            bottomMargin=0.75*inch
        )
        
        # Container for the 'Flowable' objects
        story = []
        
        # Define styles
        styles = getSampleStyleSheet()
        
        # Custom styles
        title_style = ParagraphStyle(
            'CustomTitle',
            parent=styles['Heading1'],
            fontSize=24,
            textColor=colors.HexColor('#3b82f6'),
            spaceAfter=12,
            alignment=TA_CENTER,
            fontName='Helvetica-Bold'
        )
        
        subtitle_style = ParagraphStyle(
            'CustomSubtitle',
            parent=styles['Normal'],
            fontSize=12,
            textColor=colors.HexColor('#64748b'),
            spaceAfter=20,
            alignment=TA_CENTER
        )
        
        question_style = ParagraphStyle(
            'QuestionStyle',
            parent=styles['Heading2'],
            fontSize=14,
            textColor=colors.HexColor('#3b82f6'),
            spaceAfter=10,
            fontName='Helvetica-Bold'
        )
        
        answer_style = ParagraphStyle(
            'AnswerStyle',
            parent=styles['Normal'],
            fontSize=11,
            alignment=TA_JUSTIFY,
            spaceAfter=12
        )
        
        source_title_style = ParagraphStyle(
            'SourceTitle',
            parent=styles['Heading3'],
            fontSize=11,
            textColor=colors.HexColor('#8b5cf6'),
            spaceAfter=6,
            fontName='Helvetica-Bold'
        )
        
        # Add title
        story.append(Paragraph("WHEC Research Assistant", title_style))
        story.append(Paragraph("Conversation Report", title_style))
        story.append(Spacer(1, 0.2*inch))
        
        # Add metadata
        current_time = datetime.now().strftime("%B %d, %Y at %I:%M %p")
        story.append(Paragraph(f"Generated: {current_time}", subtitle_style))
        story.append(Paragraph("Source: WHEC (Warrior Heat- and Exertion-Related Events Collaborative)", subtitle_style))
        story.append(Paragraph(f"Total Questions: {len(request.conversations)}", subtitle_style))
        
        story.append(Spacer(1, 0.3*inch))
        
        # Add horizontal line
        story.append(Spacer(1, 0.1*inch))
        
        # Process each conversation
        for idx, conv in enumerate(request.conversations, 1):
            # Question
            story.append(Paragraph(f"Question {idx}", question_style))
            story.append(Paragraph(conv.question, answer_style))
            story.append(Spacer(1, 0.15*inch))
            
            # Answer
            story.append(Paragraph("Answer", source_title_style))
            story.append(Paragraph(conv.answer, answer_style))
            story.append(Spacer(1, 0.15*inch))
            
            # Text Sources
            if conv.texts:
                story.append(Paragraph("Referenced Text Sources", source_title_style))
                for i, txt in enumerate(conv.texts, 1):
                    source_text = f"[{i}] {txt.file or 'Unknown Document'} (Page {txt.page or 'N/A'}, {round((txt.score or 0) * 100)}% match)"
                    story.append(Paragraph(source_text, styles['Normal']))
                    
                    excerpt = f'<i>"{txt.text[:300]}..."</i>'
                    story.append(Paragraph(excerpt, styles['Normal']))
                    story.append(Spacer(1, 0.1*inch))
                
                story.append(Spacer(1, 0.1*inch))
            
            # Image Sources
            if conv.images:
                relevant_images = [img for img in conv.images if (img.score or 0) >= 0.3]
                if relevant_images:
                    story.append(Paragraph("Referenced Images", source_title_style))
                    
                    for i, img in enumerate(relevant_images[:3], 1):  # Limit to 3 images
                        # Add image metadata
                        img_text = f"[{i}] {img.filename or 'Unknown'} from {img.file or 'Unknown Document'} (Page {img.page or 'N/A'}, {round((img.score or 0) * 100)}% match)"
                        story.append(Paragraph(img_text, styles['Normal']))
                        
                        # Try to add the actual image
                        if img.path:
                            try:
                                # Construct the full path to the image
                                img_path = img.path.replace('/extracted_images/', '')
                                full_img_path = os.path.join('extracted_images', img_path)
                                
                                if os.path.exists(full_img_path):
                                    # Add image with max width of 4 inches
                                    rl_img = RLImage(full_img_path, width=4*inch, height=3*inch, kind='proportional')
                                    story.append(rl_img)
                                    story.append(Spacer(1, 0.1*inch))
                            except Exception as e:
                                logger.warning(f"Could not add image to PDF: {e}")
                    
                    story.append(Spacer(1, 0.15*inch))
            
            # Add separator between conversations (except after last one)
            if idx < len(request.conversations):
                story.append(Spacer(1, 0.2*inch))
                # Add a horizontal line as separator
                line_table = Table([['']], colWidths=[6.5*inch])
                line_table.setStyle(TableStyle([
                    ('LINEABOVE', (0, 0), (-1, 0), 1, colors.HexColor('#334155')),
                ]))
                story.append(line_table)
                story.append(Spacer(1, 0.2*inch))
        
        # Build PDF
        doc.build(story)
        
        # Get PDF from buffer
        buffer.seek(0)
        
        # Return PDF as download
        return StreamingResponse(
            buffer,
            media_type="application/pdf",
            headers={
                "Content-Disposition": f"attachment; filename=WHEC_Report_{datetime.now().strftime('%Y-%m-%d')}.pdf"
            }
        )
        
    except Exception as e:
        logger.error(f"Error generating PDF report: {str(e)}")
        raise HTTPException(status_code=500, detail=f"Error generating report: {str(e)}")

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)