backend
FastAPI, SQLModel, async services for Pulse Radar backend development.
When & Why to Use This Skill
This Claude skill provides a comprehensive framework for building high-performance backends using FastAPI, SQLModel, and asynchronous services. It leverages a Hexagonal Architecture (Ports and Adapters) pattern to ensure a clean separation of concerns between API endpoints, business logic, and data persistence, making it ideal for scalable and maintainable enterprise-grade applications.
Use Cases
- Architecting scalable RESTful APIs: Utilizing FastAPI and async patterns to handle high-concurrency requests with minimal latency.
- Implementing Hexagonal Architecture: Decoupling business logic from external dependencies like databases and third-party APIs to improve testability and flexibility.
- Database Modeling and ORM Management: Defining complex data structures and relationships using SQLModel for seamless integration with PostgreSQL and other SQL databases.
- AI-Driven Backend Development: Managing LLM provider integrations, vector embeddings, and automated data classification workflows within a structured service layer.
- Standardizing Backend Workflows: Enforcing strict type checking with mypy and code quality with ruff to maintain a professional-grade codebase.
| name | backend |
|---|---|
| description | FastAPI, SQLModel, async services for Pulse Radar backend development. |
Backend Development Skill
Architecture (Hexagonal)
backend/app/
├── api/v1/ # 23 routers (~100 endpoints)
├── models/ # 21 SQLModel entities
├── services/ # 35 services (business logic)
├── tasks/ # TaskIQ workers
└── ws/ # WebSocket endpoint
Layer Pattern
Router (API) → Service (logic) → CRUD (data)
↓ ↓ ↓
Validation Business rules Database
API Pattern
@router.get("/atoms/{atom_id}")
async def get_atom(
atom_id: UUID,
service: AtomService = Depends()
) -> AtomRead:
return await service.get_by_id(atom_id)
Service Pattern
class AtomService:
def __init__(self, crud: AtomCRUD = Depends()):
self.crud = crud
async def get_by_id(self, atom_id: UUID) -> AtomRead:
atom = await self.crud.get(atom_id)
if not atom:
raise HTTPException(404, "Atom not found")
return AtomRead.model_validate(atom)
Key Models
- Message: content, embedding (1536), importance_score, noise_classification
- Topic: name, icon, color, keywords
- Atom: type, title, content, confidence, user_approved
- LLMProvider: type (ollama/openai), api_key_encrypted, validation_status
Key Enums
AnalysisStatus: pending, analyzed, spam, noise
AtomType: problem, solution, decision, question, insight, pattern, requirement
ProviderType: ollama, openai
ValidationStatus: pending, validating, connected, error
Verification
just typecheck # mypy strict
just fmt # ruff format
just test # pytest suite
References
- @references/models.md — All 21 models with fields
- @references/services.md — All 35 services
- @references/taskiq.md — Background task patterns