feat: add document summary generation feature

backend/alembic/versions/875afd8e0b2e_add_summary_to_document.py

@@ -0,0 +1,32 @@
+"""Add summary to document
+
+Revision ID: 875afd8e0b2e
+Revises: 6e2d1d2b7c4a
+Create Date: 2026-03-28 16:00:35.361232
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+# revision identifiers, used by Alembic.
+revision: str = '875afd8e0b2e'
+down_revision: Union[str, Sequence[str], None] = '6e2d1d2b7c4a'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    """Upgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.add_column('documents', sa.Column('summary', sa.Text(), nullable=True))
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_column('documents', 'summary')
+    # ### end Alembic commands ###

backend/app/api/document.py

@@ -10,8 +10,9 @@
 from app.utils.utils import get_page_count
 from app.core.deps import get_current_user
 from app.repositories.document_repository import DocumentRepository
-from app.schemas.document import DocumentOut
+from app.schemas.document import DocumentOut, DocumentSummaryOut
 from app.utils.utils import process_document_task
+from app.services.document_service import DocumentService
 
 
 router = APIRouter(prefix="/documents", tags=["document"])
@@ -88,3 +89,36 @@ def delete_document(
     if DocumentRepository.delete_document(db, document_id):
         return {"message": "Document and all related data deleted"}
 
+@router.post("/{document_id}/summary", response_model=DocumentSummaryOut)
+def generate_document_summary_endpoint(
+    document_id: UUID,
+    db: Session = Depends(get_db),
+    user: User = Depends(get_current_user)
+):
+    document = DocumentRepository.get_document_by_id(db, document_id)
+    if not document or document.user_id != user.id:
+        raise HTTPException(status_code=404, detail="Document not found")
+
+    if document.summary:
+        return {"summary": document.summary, "message": "Summary already created"}
+
+    try:
+        summary = DocumentService.generate_document_summary(db, document_id)
+        return {"summary": summary, "message": "Summary generated successfully"}
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+@router.get("/{document_id}/summary", response_model=DocumentSummaryOut)
+def get_document_summary_endpoint(
+    document_id: UUID,
+    db: Session = Depends(get_db),
+    user: User = Depends(get_current_user)
+):
+    document = DocumentRepository.get_document_by_id(db, document_id)
+    if not document or document.user_id != user.id:
+        raise HTTPException(status_code=404, detail="Document not found")
+
+    if not document.summary:
+        raise HTTPException(status_code=404, detail="Summary not generated yet")
+
+    return {"summary": document.summary, "message": "Summary retrieved from storage"}

backend/app/models/document.py

@@ -2,7 +2,7 @@
 
 import enum
 
-from sqlalchemy import Column, ForeignKey, Index, Integer, String,  DateTime, Enum, func
+from sqlalchemy import Column, ForeignKey, Index, Integer, String, DateTime, Enum, func, Text
 from sqlalchemy.orm import relationship
 from sqlalchemy.dialects.postgresql import UUID
 
@@ -44,6 +44,8 @@ class Document(Base):
     nullable=False
   )
 
+  summary = Column(Text, nullable=True)
+
   user = relationship("User", back_populates="documents")
   chunks = relationship("Chunk", back_populates="document", cascade="all, delete")
   ai_interactions = relationship("AIInteraction", back_populates="document", cascade="all, delete-orphan")

backend/app/repositories/document_repository.py

@@ -3,6 +3,7 @@
 from uuid import UUID
 
 from app.models.document import Document
+from app.models.chunk import Chunk
 
 
 class DocumentRepository:
@@ -54,3 +55,18 @@ def delete_document(db: Session, document_id: UUID) -> bool:
         db.delete(document)
         db.commit()
         return True
+
+    @staticmethod
+    def update_document_summary(db: Session, document_id: UUID, summary: str) -> None:
+        document = db.query(Document).filter(Document.id == document_id).first()
+
+        if not document:
+            raise ValueError(f"Document with id {document_id} not found")
+
+        document.summary = summary
+        db.commit()
+        db.refresh(document)
+
+    @staticmethod
+    def get_document_chunks(db: Session, document_id: UUID) -> list[Chunk]:
+        return db.query(Chunk).filter(Chunk.document_id == document_id).order_by(Chunk.chunk_index.asc()).all()

backend/app/schemas/document.py

@@ -10,6 +10,11 @@ class DocumentOut(BaseModel):
   filename: str
   pages: int
   status: str
+  summary: str | None = None
   created_at: datetime
 
   model_config = ConfigDict(from_attributes=True)
+
+class DocumentSummaryOut(BaseModel):
+  summary: str
+  message: str | None = None

backend/app/services/document_service.py

@@ -1,6 +1,10 @@
 import fitz  # PyMuPDF
 from pptx import Presentation
 from io import BytesIO
+from sqlalchemy.orm import Session
+from uuid import UUID
+from app.repositories.document_repository import DocumentRepository
+from app.services.llm_service import LLMService
 
 
 class DocumentService:
@@ -49,3 +53,28 @@ def _clean_text(text: str) -> str:
         text = text.replace("\x00", "")
         text = text.replace("\n\n", "\n")
         return text.strip()
+
+    @staticmethod
+    def generate_document_summary(db: Session, document_id: UUID) -> str:
+        # Check if already generated
+        document = DocumentRepository.get_document_by_id(db, document_id)
+        if not document:
+            raise ValueError(f"Document {document_id} not found")
+        if document.summary:
+            return document.summary
+
+        chunks = DocumentRepository.get_document_chunks(db, document_id)
+        if not chunks:
+            raise ValueError(f"No text extracted for document {document_id}")
+
+        # Combine all parts
+        combined_text = "\n\n".join([chunk.text for chunk in chunks])
+
+        # Generate summary using LLM service
+        summary_markdown = LLMService.generate_summary(combined_text)
+
+        # Save to DB
+        DocumentRepository.update_document_summary(db, document_id, summary_markdown)
+
+        return summary_markdown
+

backend/app/services/llm_service.py

@@ -32,6 +32,37 @@ def generate(cls, prompt: str) -> str:
 
         response.raise_for_status()
 
-        data = response.json()
+        return data.get("choices", [{}])[0].get("message", {}).get("content", "")
+
+    @classmethod
+    def generate_summary(cls, text: str) -> str:
+        system_prompt = (
+            "You are an expert tutor. Provide a comprehensive, structured, and interactive summary of the provided text. "
+            "Use Markdown formatting, bold key terms, use bullet points, and clear headings. "
+            "Ensure everything a user needs to know to ace an exam on this topic is included. Be concise to save tokens but highly informative."
+        )
+
+        # Limiting input text length to avoid excessive token usage
+        max_chars = 120000
+        truncated_text = text[:max_chars] if len(text) > max_chars else text
 
+        payload = {
+            "model": cls.MODEL,
+            "messages": [
+                {"role": "system", "content": system_prompt},
+                {"role": "user", "content": f"Summarize this document comprehensively into key points:\n\n{truncated_text}"}
+            ],
+            "temperature": 0.3,
+            "max_tokens": 1500
+        }
+
+        response = requests.post(
+            cls.API_URL,
+            headers=cls.HEADERS,
+            json=payload,
+            timeout=180
+        )
+
+        response.raise_for_status()
+        data = response.json()
         return data.get("choices", [{}])[0].get("message", {}).get("content", "")

frontend/src/components/study-room/DocumentSummaryView.tsx

@@ -0,0 +1,109 @@
+import React, { useEffect, useState } from 'react';
+import ReactMarkdown from 'react-markdown';
+import remarkGfm from 'remark-gfm';
+import { FileText, Check, Copy } from 'lucide-react';
+import { documentService, type DocumentSummary } from '../../services/document.service';
+import { Button } from '../ui/Button';
+import { LoadingView } from './LoadingView';
+import { Skeleton } from '../ui/Skeleton';
+
+interface DocumentSummaryViewProps {
+  documentId: string;
+}
+
+export const DocumentSummaryView: React.FC<DocumentSummaryViewProps> = ({ documentId }) => {
+  const [summary, setSummary] = useState<DocumentSummary | null>(null);
+  const [isLoading, setIsLoading] = useState(false);
+  const [error, setError] = useState<string | null>(null);
+  const [isCopied, setIsCopied] = useState(false);
+
+  useEffect(() => {
+    if (!documentId) return;
+
+    const fetchOrGenerateSummary = async () => {
+      setIsLoading(true);
+      setError(null);
+      try {
+        // Try getting the existing summary first
+        try {
+          const data = await documentService.getSummary(documentId);
+          setSummary(data);
+        } catch (e: unknown) {
+          const err = e as { response?: { status?: number }, status?: number };
+          // If not found (404), generate it
+          if (err.response?.status === 404 || err.status === 404) {
+            const generatedData = await documentService.generateSummary(documentId);
+            setSummary(generatedData);
+          } else {
+            throw e;
+          }
+        }
+      } catch (e: unknown) {
+        console.error('Failed to load summary:', e);
+        setError('Failed to generate or retrieve the document summary. Please try again later.');
+      } finally {
+        setIsLoading(false);
+      }
+    };
+
+    if (!summary) {
+      fetchOrGenerateSummary();
+    }
+  }, [documentId, summary]);
+
+  const handleCopy = () => {
+    if (summary) {
+      navigator.clipboard.writeText(summary.summary);
+      setIsCopied(true);
+      setTimeout(() => setIsCopied(false), 2000);
+    }
+  };
+
+  return (
+    <div className="flex flex-col h-full bg-surface">
+      <div className="flex-1 overflow-y-auto px-6 py-6 scrollbar-hide">
+        {isLoading ? (
+          <div className="flex flex-col items-center justify-center h-full space-y-6 animate-in fade-in duration-700">
+            <LoadingView text="AI is reading and summarizing your document..." />
+            <div className="w-full max-w-sm mt-8 space-y-3 opacity-50">
+              <Skeleton className="h-4 w-full rounded-md" />
+              <Skeleton className="h-4 w-[90%] rounded-md" />
+              <Skeleton className="h-4 w-[95%] rounded-md" />
+              <Skeleton className="h-4 w-[80%] rounded-md" />
+              <Skeleton className="h-4 w-full mt-6 rounded-md" />
+              <Skeleton className="h-4 w-[85%] rounded-md" />
+              <Skeleton className="h-4 w-[90%] rounded-md" />
+            </div>
+          </div>
+        ) : error ? (
+          <div className="flex flex-col items-center justify-center h-full py-12 text-center animate-fade-in">
+            <div className="w-12 h-12 bg-destructive/10 rounded-2xl flex items-center justify-center mb-4">
+              <FileText className="w-6 h-6 text-destructive" />
+            </div>
+            <p className="text-sm font-medium text-foreground">{error}</p>
+          </div>
+        ) : summary ? (
+          <div className="prose prose-sm md:prose-base prose-invert max-w-none animate-fade-in-up pb-8 selection:bg-primary/30">
+            <ReactMarkdown remarkPlugins={[remarkGfm]}>
+              {summary.summary}
+            </ReactMarkdown>
+          </div>
+        ) : null}
+      </div>
+
+      {/* Footer */}
+      {summary && !isLoading && !error && (
+        <div className="px-6 py-4 border-t border-border/10 bg-surface/80 backdrop-blur-md shrink-0 flex justify-end">
+          <Button
+            variant="secondary"
+            size="sm"
+            onClick={handleCopy}
+            icon={isCopied ? <Check className="w-4 h-4" /> : <Copy className="w-4 h-4" />}
+          >
+            {isCopied ? 'Copied to Clipboard' : 'Copy Summary'}
+          </Button>
+        </div>
+      )}
+    </div>
+  );
+};