supabase
Handles the full Supabase workflow from schema changes to deployment, with built-in security guardrails that catch common traps like RLS...
fastapi-production
Production-grade FastAPI patterns for async APIs, SQLAlchemy 2.0, Pydantic v2, and robust error handling. Use when building API endpoints, handling database operations, implementing middleware, or optimizing performance.
Full skill instructions
Production-ready patterns for FastAPI applications with async SQLAlchemy, Pydantic validation, and enterprise-grade error handling.
backend/app/
├── api/
│ ├── deps.py # Dependency injection
│ └── routes/ # API endpoints
├── controllers/ # Request/response handling
├── services/ # Business logic
├── repositories/ # Data access (optional)
├── models/ # SQLAlchemy ORM
├── schemas/ # Pydantic schemas
├── core/ # Config, security
└── db/ # Database session
Route (thin) → Parse request, call controller/service, return response
Controller → Orchestrate service calls, transform data
Service → Business logic, validation, transactions
Repository → Database queries (optional abstraction)
Model → ORM definitions, relationships
Schema → Request/response validation
from fastapi import APIRouter, Depends, HTTPException, status
from sqlalchemy.ext.asyncio import AsyncSession
from app.api.deps import get_db, get_current_user
from app.schemas.schedule import ScheduleCreate, ScheduleResponse, ScheduleList
from app.services import schedule_service
from app.models.user import User
router = APIRouter(prefix="/schedules", tags=["schedules"])
@router.get("", response_model=ScheduleList)
async def list_schedules(
skip: int = 0,
limit: int = 100,
db: AsyncSession = Depends(get_db),
current_user: User = Depends(get_current_user),
) -> ScheduleList:
"""
List all schedules with pagination.
- **skip**: Number of records to skip
- **limit**: Maximum number of records to return
"""
schedules = await schedule_service.get_multi(db, skip=skip, limit=limit)
total = await schedule_service.count(db)
return ScheduleList(items=schedules, total=total, skip=skip, limit=limit)
@router.get("/{schedule_id}", response_model=ScheduleResponse)
async def get_schedule(
schedule_id: str,
db: AsyncSession = Depends(get_db),
current_user: User = Depends(get_current_user),
) -> ScheduleResponse:
"""Get a specific schedule by ID."""
schedule = await schedule_service.get(db, id=schedule_id)
if not schedule:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND,
detail="Schedule not found",
)
return schedule
@router.post("", response_model=ScheduleResponse, status_code=status.HTTP_201_CREATED)
async def create_schedule(
schedule_in: ScheduleCreate,
db: AsyncSession = Depends(get_db),
current_user: User = Depends(get_current_user),
) -> ScheduleResponse:
"""Create a new schedule."""
schedule = await schedule_service.create(db, obj_in=schedule_in, created_by=current_user.id)
return schedule
from typing import List
@router.post("/bulk", response_model=List[ScheduleResponse])
async def create_schedules_bulk(
schedules_in: List[ScheduleCreate],
db: AsyncSession = Depends(get_db),
current_user: User = Depends(get_current_user),
) -> List[ScheduleResponse]:
"""Create multiple schedules in a single transaction."""
if len(schedules_in) > 100:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail="Maximum 100 schedules per request",
)
async with db.begin():
schedules = await schedule_service.create_multi(
db, objs_in=schedules_in, created_by=current_user.id
)
return schedules
from pydantic import BaseModel, Field, field_validator
from datetime import date, datetime
from typing import Optional, List
# Base schema with shared fields
class ScheduleBase(BaseModel):
name: str = Field(..., min_length=1, max_length=255)
start_date: date
end_date: date
@field_validator("end_date")
@classmethod
def end_date_after_start(cls, v: date, info) -> date:
if "start_date" in info.data and v < info.data["start_date"]:
raise ValueError("end_date must be after start_date")
return v
# Create schema - fields required for creation
class ScheduleCreate(ScheduleBase):
description: Optional[str] = None
# Update schema - all fields optional
class ScheduleUpdate(BaseModel):
name: Optional[str] = Field(None, min_length=1, max_length=255)
start_date: Optional[date] = None
end_date: Optional[date] = None
description: Optional[str] = None
# Response schema - includes DB fields
class ScheduleResponse(ScheduleBase):
id: str
description: Optional[str]
created_at: datetime
updated_at: Optional[datetime]
created_by: str
model_config = {"from_attributes": True}
# List response with pagination
class ScheduleList(BaseModel):
items: List[ScheduleResponse]
total: int
skip: int
limit: int
class AssignmentResponse(BaseModel):
id: str
person_id: str
block_id: str
rotation_type: str
model_config = {"from_attributes": True}
class ScheduleWithAssignments(ScheduleResponse):
assignments: List[AssignmentResponse] = []
from typing import Optional, List, TypeVar, Generic
from sqlalchemy.ext.asyncio import AsyncSession
from sqlalchemy import select, func
from pydantic import BaseModel
from app.models.schedule import Schedule
from app.schemas.schedule import ScheduleCreate, ScheduleUpdate
class ScheduleService:
"""Service for schedule business logic."""
async def get(self, db: AsyncSession, *, id: str) -> Optional[Schedule]:
"""Get a schedule by ID."""
result = await db.execute(select(Schedule).where(Schedule.id == id))
return result.scalar_one_or_none()
async def get_multi(
self,
db: AsyncSession,
*,
skip: int = 0,
limit: int = 100,
) -> List[Schedule]:
"""Get multiple schedules with pagination."""
result = await db.execute(
select(Schedule)
.offset(skip)
.limit(limit)
.order_by(Schedule.created_at.desc())
)
return list(result.scalars().all())
async def count(self, db: AsyncSession) -> int:
"""Count total schedules."""
result = await db.execute(select(func.count()).select_from(Schedule))
return result.scalar() or 0
async def create(
self,
db: AsyncSession,
*,
obj_in: ScheduleCreate,
created_by: str,
) -> Schedule:
"""Create a new schedule."""
db_obj = Schedule(
**obj_in.model_dump(),
created_by=created_by,
)
db.add(db_obj)
await db.flush()
await db.refresh(db_obj)
return db_obj
async def update(
self,
db: AsyncSession,
*,
db_obj: Schedule,
obj_in: ScheduleUpdate,
) -> Schedule:
"""Update an existing schedule."""
update_data = obj_in.model_dump(exclude_unset=True)
for field, value in update_data.items():
setattr(db_obj, field, value)
db.add(db_obj)
await db.flush()
await db.refresh(db_obj)
return db_obj
async def delete(self, db: AsyncSession, *, id: str) -> bool:
"""Delete a schedule by ID."""
obj = await self.get(db, id=id)
if obj:
await db.delete(obj)
await db.flush()
return True
return False
# Singleton instance
schedule_service = ScheduleService()
from app.scheduling.acgme_validator import ACGMEValidator
class AssignmentService:
"""Service with business logic and validation."""
def __init__(self):
self.validator = ACGMEValidator()
async def create_assignment(
self,
db: AsyncSession,
*,
obj_in: AssignmentCreate,
created_by: str,
) -> Assignment:
"""Create assignment with ACGME validation."""
# Check existing assignments for conflicts
conflicts = await self._check_conflicts(db, obj_in)
if conflicts:
raise ValueError(f"Assignment conflicts with: {conflicts}")
# Validate ACGME compliance
violations = await self.validator.validate_assignment(db, obj_in)
if violations:
raise ACGMEViolationError(violations)
# Create assignment
assignment = Assignment(**obj_in.model_dump(), created_by=created_by)
db.add(assignment)
await db.flush()
return assignment
async def _check_conflicts(
self,
db: AsyncSession,
obj_in: AssignmentCreate,
) -> List[Assignment]:
"""Check for conflicting assignments."""
result = await db.execute(
select(Assignment)
.where(Assignment.person_id == obj_in.person_id)
.where(Assignment.block_id == obj_in.block_id)
)
return list(result.scalars().all())
from sqlalchemy.orm import selectinload, joinedload
async def get_schedule_with_assignments(
db: AsyncSession, schedule_id: str
) -> Optional[Schedule]:
"""Get schedule with all assignments eagerly loaded."""
result = await db.execute(
select(Schedule)
.options(
selectinload(Schedule.assignments).selectinload(Assignment.person),
selectinload(Schedule.assignments).selectinload(Assignment.block),
)
.where(Schedule.id == schedule_id)
)
return result.scalar_one_or_none()
from sqlalchemy import and_, or_, func
from datetime import date, timedelta
async def get_upcoming_assignments(
db: AsyncSession,
person_id: str,
days_ahead: int = 7,
) -> List[Assignment]:
"""Get assignments for the next N days."""
today = date.today()
end_date = today + timedelta(days=days_ahead)
result = await db.execute(
select(Assignment)
.join(Block)
.where(
and_(
Assignment.person_id == person_id,
Block.date >= today,
Block.date <= end_date,
)
)
.order_by(Block.date, Block.session)
)
return list(result.scalars().all())
async def get_coverage_stats(
db: AsyncSession,
schedule_id: str,
) -> dict:
"""Get coverage statistics for a schedule."""
result = await db.execute(
select(
Block.date,
func.count(Assignment.id).label("assigned_count"),
)
.outerjoin(Assignment, Assignment.block_id == Block.id)
.where(Assignment.schedule_id == schedule_id)
.group_by(Block.date)
)
return {row.date: row.assigned_count for row in result.all()}
from sqlalchemy import select
from sqlalchemy.orm import with_for_update
async def update_with_lock(
db: AsyncSession,
assignment_id: str,
new_status: str,
) -> Assignment:
"""Update with row lock to prevent race conditions."""
result = await db.execute(
select(Assignment)
.where(Assignment.id == assignment_id)
.with_for_update() # Lock the row
)
assignment = result.scalar_one_or_none()
if not assignment:
raise ValueError("Assignment not found")
assignment.status = new_status
await db.flush()
return assignment
# app/core/exceptions.py
from typing import Any, Optional
class AppException(Exception):
"""Base application exception."""
def __init__(
self,
message: str,
code: str = "APP_ERROR",
details: Optional[dict] = None,
):
self.message = message
self.code = code
self.details = details or {}
super().__init__(message)
class NotFoundError(AppException):
"""Resource not found."""
def __init__(self, resource: str, id: str):
super().__init__(
message=f"{resource} not found",
code="NOT_FOUND",
details={"resource": resource, "id": id},
)
class ValidationError(AppException):
"""Validation failed."""
def __init__(self, errors: list):
super().__init__(
message="Validation failed",
code="VALIDATION_ERROR",
details={"errors": errors},
)
class ACGMEViolationError(AppException):
"""ACGME compliance violation."""
def __init__(self, violations: list):
super().__init__(
message="ACGME compliance violation",
code="ACGME_VIOLATION",
details={"violations": violations},
)
# app/api/exception_handlers.py
from fastapi import Request, status
from fastapi.responses import JSONResponse
from app.core.exceptions import AppException, NotFoundError, ValidationError
async def app_exception_handler(request: Request, exc: AppException) -> JSONResponse:
"""Handle application exceptions."""
status_map = {
"NOT_FOUND": status.HTTP_404_NOT_FOUND,
"VALIDATION_ERROR": status.HTTP_400_BAD_REQUEST,
"ACGME_VIOLATION": status.HTTP_422_UNPROCESSABLE_ENTITY,
}
return JSONResponse(
status_code=status_map.get(exc.code, status.HTTP_500_INTERNAL_SERVER_ERROR),
content={
"error": {
"code": exc.code,
"message": exc.message,
"details": exc.details,
}
},
)
# Register in main.py
app.add_exception_handler(AppException, app_exception_handler)
# app/api/deps.py
from typing import AsyncGenerator
from sqlalchemy.ext.asyncio import AsyncSession
from app.db.session import async_session
async def get_db() -> AsyncGenerator[AsyncSession, None]:
"""Dependency for database session."""
async with async_session() as session:
try:
yield session
await session.commit()
except Exception:
await session.rollback()
raise
from fastapi import Depends, HTTPException, status
from fastapi.security import OAuth2PasswordBearer
from app.core.security import decode_token
from app.models.user import User
oauth2_scheme = OAuth2PasswordBearer(tokenUrl="/api/auth/login")
async def get_current_user(
db: AsyncSession = Depends(get_db),
token: str = Depends(oauth2_scheme),
) -> User:
"""Get current authenticated user."""
credentials_exception = HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
detail="Could not validate credentials",
headers={"WWW-Authenticate": "Bearer"},
)
payload = decode_token(token)
if payload is None:
raise credentials_exception
user_id = payload.get("sub")
if user_id is None:
raise credentials_exception
user = await user_service.get(db, id=user_id)
if user is None:
raise credentials_exception
return user
async def get_current_active_user(
current_user: User = Depends(get_current_user),
) -> User:
"""Ensure user is active."""
if not current_user.is_active:
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN,
detail="Inactive user",
)
return current_user
from functools import wraps
from typing import List
def require_roles(roles: List[str]):
"""Dependency factory for role-based access."""
async def role_checker(
current_user: User = Depends(get_current_active_user),
) -> User:
if current_user.role not in roles:
raise HTTPException(
status_code=status.HTTP_403_FORBIDDEN,
detail="Insufficient permissions",
)
return current_user
return role_checker
# Usage
@router.delete("/{schedule_id}")
async def delete_schedule(
schedule_id: str,
db: AsyncSession = Depends(get_db),
current_user: User = Depends(require_roles(["ADMIN", "COORDINATOR"])),
) -> dict:
"""Delete a schedule (admin/coordinator only)."""
await schedule_service.delete(db, id=schedule_id)
return {"status": "deleted"}
from fastapi import BackgroundTasks
async def send_notification(user_id: str, message: str):
"""Background task to send notification."""
# Implementation
pass
@router.post("/assignments")
async def create_assignment(
assignment_in: AssignmentCreate,
background_tasks: BackgroundTasks,
db: AsyncSession = Depends(get_db),
current_user: User = Depends(get_current_user),
):
"""Create assignment and notify affected users."""
assignment = await assignment_service.create(db, obj_in=assignment_in)
# Add background task
background_tasks.add_task(
send_notification,
assignment.person_id,
f"New assignment: {assignment.rotation_type}",
)
return assignment
from app.core.celery import celery_app
@celery_app.task(bind=True, max_retries=3)
def generate_schedule_task(self, schedule_id: str, params: dict):
"""Celery task for schedule generation."""
try:
# Long-running operation
result = generate_schedule(schedule_id, **params)
return {"status": "success", "result": result}
except Exception as exc:
self.retry(exc=exc, countdown=60)
@router.post("/schedules/{schedule_id}/generate")
async def trigger_generation(
schedule_id: str,
params: GenerationParams,
current_user: User = Depends(get_current_user),
):
"""Trigger async schedule generation."""
task = generate_schedule_task.delay(schedule_id, params.model_dump())
return {"task_id": task.id, "status": "queued"}
from functools import lru_cache
from fastapi_cache import FastAPICache
from fastapi_cache.decorator import cache
@router.get("/stats")
@cache(expire=300) # Cache for 5 minutes
async def get_stats(db: AsyncSession = Depends(get_db)):
"""Get cached statistics."""
return await stats_service.calculate(db)
from fastapi.responses import StreamingResponse
import csv
import io
@router.get("/export/csv")
async def export_csv(
db: AsyncSession = Depends(get_db),
current_user: User = Depends(get_current_user),
):
"""Stream large CSV export."""
async def generate_csv():
output = io.StringIO()
writer = csv.writer(output)
writer.writerow(["ID", "Name", "Date"])
async for batch in get_data_batches(db):
for row in batch:
writer.writerow([row.id, row.name, row.date])
yield output.getvalue()
output.seek(0)
output.truncate(0)
return StreamingResponse(
generate_csv(),
media_type="text/csv",
headers={"Content-Disposition": "attachment; filename=export.csv"},
)
cd /home/user/Autonomous-Assignment-Program-Manager/backend
# Development
uvicorn app.main:app --reload
# Testing
pytest
pytest --cov=app
# Linting
ruff check app/
ruff format app/
# Type checking
mypy app/
# Database
alembic upgrade head
alembic revision --autogenerate -m "description"
For schema changes:
For auth/security code:
After creating endpoints:
Escalate to human when:
Handles the full Supabase workflow from schema changes to deployment, with built-in security guardrails that catch common traps like RLS...
Comprehensive guides and best practices for Neon Serverless Postgres, covering setup, connection methods, authentication, and platform APIs.
Guide for setting up and using Firebase Authentication. Use this skill when the user's app requires user sign-in, user management, or secure data access using auth rules.
A skill to evaluate how secure Firestore security rules are. Use this when Firestore security rules are updated to ensure that the generated rules are extremely secure and robust.
Official skill for integrating Firebase AI Logic (Gemini API) into web applications. Covers setup, multimodal inference, structured output, and security.
Complete Better Auth server and client setup with database adapters, session management, plugins, and security configuration.
Deploy and manage projects on Vercel using token-based authentication. Use when working with Vercel CLI using access tokens rather than interactive login — e.g. "deploy to vercel", "set up vercel", "add environment variables to vercel".
Complete Cloudflare platform integration with decision trees for compute, storage, AI, networking, security, and infrastructure-as-code.
Run Azure compliance and security audits with azqr plus Key Vault expiration checks. Covers best-practice assessment, resource review, policy/compliance validation, and security posture checks. WHEN: compliance scan, security audit, BEFORE running azqr (compliance cli tool), Azure best practices, Key Vault expiration check, expired certificates, expiring secrets, orphaned resources, compliance assessment.
Send, read, and manage Gmail messages, drafts, labels, and account settings.
Comprehensive website auditing across 230+ rules in 21 categories including SEO, performance, security, and accessibility.
Shared authentication, CLI syntax, and output formatting patterns for gws Google Workspace commands.
Discover your celebrity doppelgänger with StarByFace!
GeoSpy: Pricing, Features, FAQs, and Alternatives for AI Teams
Detect AI-generated voices to protect against audio fraud.
Ensure Your Content's Originality with AI Plagiarism Checker
Upscale images by 400% without quality loss
Protect your social media from copyright disputes
Accurately Detect AI-Generated Content with TheChecker.AI
Chrome extension that detects and flags AI-generated content