feat: Implement Sample Paper Generator with format detection and question generation

backend/alembic/versions/983ad2983d5c_add_selected_resource_ids_to_chat_.py

@@ -0,0 +1,32 @@
+"""add selected_resource_ids to chat_sessions
+
+Revision ID: 983ad2983d5c
+Revises: d3dcf6b6fe3f
+Create Date: 2026-05-18 20:55:14.773697
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '983ad2983d5c'
+down_revision: Union[str, Sequence[str], None] = 'd3dcf6b6fe3f'
+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('chat_sessions', sa.Column('selected_resource_ids', sa.JSON(), nullable=True))
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_column('chat_sessions', 'selected_resource_ids')
+    # ### end Alembic commands ###

backend/alembic/versions/ab47c0990b86_add_question_pdf_url_to_paper_outputs.py

@@ -0,0 +1,32 @@
+"""add question_pdf_url to paper_outputs
+
+Revision ID: ab47c0990b86
+Revises: 983ad2983d5c
+Create Date: 2026-05-18 21:35:55.165368
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = 'ab47c0990b86'
+down_revision: Union[str, Sequence[str], None] = '983ad2983d5c'
+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('paper_outputs', sa.Column('question_pdf_url', sa.String(), nullable=True))
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    """Downgrade schema."""
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_column('paper_outputs', 'question_pdf_url')
+    # ### end Alembic commands ###

backend/app/config.py

@@ -30,6 +30,8 @@ class Settings(BaseSettings):
 
     FRONTEND_URL: str = "http://localhost:5173"
     QUESTIONS_LIMIT: int = 30
+    PAPERS_LIMIT: int = 3
+    RESOURCES_LIMIT: int = 3
     class Config:
         env_file = ".env"
 

backend/app/llm/client.py

@@ -9,6 +9,7 @@ def __init__(self):
         self.base_url = "https://openrouter.ai/api/v1/chat/completions"
 
     async def stream_chat(self, messages: list, model: str = "openrouter/owl-alpha") -> AsyncGenerator[str, None]:
+    # async def stream_chat(self, messages: list, model: str = "nvidia/nemotron-3-super-120b-a12b:free") -> AsyncGenerator[str, None]:
         headers = {
             "Authorization": f"Bearer {self.api_key}",
             "Content-Type": "application/json",

backend/app/llm/prompts.py

@@ -20,3 +20,42 @@
 Question:
 {question}
 """
+
+DETECT_FORMAT_PROMPT = """
+Analyse this past year paper and extract the question format.
+Return ONLY a JSON object with this exact structure, no explanation:
+
+{
+  "mcq": <count>,
+  "short": <count>,
+  "long": <count>,
+  "mcq_marks": <marks each>,
+  "short_marks": <marks each>,
+  "long_marks": <marks each>,
+  "total_marks": <total>,
+  "duration_minutes": <duration or null>
+}
+"""
+
+GENERATE_PAPER_PROMPT = """
+You are generating a sample exam paper for a student.
+
+Format config: {format_config}
+Subject context from student's material:
+{context_chunks}
+
+Generate exactly the number of questions specified in the format config.
+Return ONLY a JSON array of question objects. Each object must have:
+  - type (mcq | short | long)
+  - marks (integer)
+  - topic (string)
+  - question_text (string)
+  - For MCQ: also include options (array of 4 strings) and answer (correct option text)
+  - For short/long: also include answer (model answer string) and explanation (string)
+
+Rules:
+- Distribute questions across different topics evenly.
+- No repeated questions.
+- Difficulty should match a real exam for this level.
+- JSON array only. No preamble, no explanation, no markdown fences.
+"""

backend/app/main.py

@@ -1,6 +1,6 @@
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
-from .routers import auth, resources, solver
+from .routers import auth, resources, solver, papers
 from .config import settings
 
 app = FastAPI(title="PYQ Solver API")
@@ -18,6 +18,7 @@
 app.include_router(auth.router, prefix="/api")
 app.include_router(resources.router, prefix="/api")
 app.include_router(solver.router, prefix="/api")
+app.include_router(papers.router, prefix="/api")
 
 @app.get("/")
 async def root():

backend/app/models/chat_session.py

@@ -1,4 +1,4 @@
-from sqlalchemy import Column, String, DateTime, UUID, ForeignKey
+from sqlalchemy import Column, String, DateTime, UUID, ForeignKey, JSON
 from sqlalchemy.orm import relationship
 from datetime import datetime
 import uuid
@@ -10,6 +10,7 @@ class ChatSession(Base):
     id = Column(UUID(as_uuid=True), primary_key=True, default=uuid.uuid4)
     user_id = Column(UUID(as_uuid=True), ForeignKey("users.id", ondelete="CASCADE"), nullable=False)
     title = Column(String, default="New Chat")
+    selected_resource_ids = Column(JSON, default=list)
     created_at = Column(DateTime, default=datetime.utcnow)
     updated_at = Column(DateTime, default=datetime.utcnow, onupdate=datetime.utcnow)
 

backend/app/models/paper_output.py

@@ -12,7 +12,8 @@ class PaperOutput(Base):
     questions = Column(JSON, nullable=False) # Array of question objects
     include_answers = Column(Boolean, default=True)
     include_explanations = Column(Boolean, default=True)
-    pdf_url = Column(String, nullable=True)
+    pdf_url = Column(String, nullable=True) # Full version (with answers/explanations as per toggles)
+    question_pdf_url = Column(String, nullable=True) # Clean question paper only
     created_at = Column(DateTime, default=datetime.utcnow)
 
     paper = relationship("Paper", back_populates="output")