django-nova 0.2.0

Next-generation Django toolkit: typed ORM, unified validation, async-first, smart caching

🚀 Django Nova

Typed, unified, and async-first toolkit for Django 5+

Django Nova eliminates fundamental architectural flaws in Django that lead to data corruption, runtime errors, and maintainability issues in scientific and enterprise software.

---

🔑 Key Innovations

• ✅ Single Source of Truth: Define validation once in Pydantic. Django Models, Forms, and APIs automatically read from it. No more duplication.
• 🔒 Strict Type Safety: Full pyright --strict compatibility for ORM, QuerySets, and Models using modern PEP 695 syntax.
• ⚡ Smart QuerySet Cache: Automatic O(1) cache invalidation on write via Django signals. Zero percent stale data in research pipelines.
• 🔄 Zero-Downtime Migrations: Native PostgreSQL CONCURRENTLY operations for locked tables containing millions of rows.
• 📊 Structured Observability: Built-in structlog integration emitting machine-readable JSON logs with ISO-timestamps for Datadog/ELK.
• 🔍 Distributed Tracing: OpenTelemetry spans for Model.save() and Cache operations. Zero overhead if OTEL is not installed.
• 🔌 DRF Auto-Serializer: Generate Django Rest Framework Serializers dynamically from Pydantic schemas. Pydantic validation overrides DRF.
• 🚀 FastAPI Auto-Router: Generate fully documented FastAPI routers with native OpenAPI/Swagger from Django models.

---

🚀 Quick Start

Installation

# Core library
pip install django-nova

# With DRF support
pip install django-nova[drf]

# With FastAPI support
pip install django-nova[fastapi]

# With full enterprise stack (tracing, logging)
pip install django-nova[tracing,observability]

---

💡 Usage Example

Define your rules once, use them everywhere:

# models.py
from pydantic import BaseModel, field_validator
from django.db import models
from nova import NovaModel, NovaConfig

# 1. Define validation rules (ONCE)
class ResearcherSchema(BaseModel):
    name: str
    h_index: int = 0

    @field_validator("h_index")
    @classmethod
    def validate_h_index(cls, v: int) -> int:
        if v < 0:
            raise ValueError("h-index cannot be negative")
        return v

# 2. Link to Django
class Researcher(NovaModel):
    name = models.CharField(max_length=300)
    h_index = models.IntegerField(default=0)

    _nova_config = NovaConfig(
        pydantic_schema=ResearcherSchema,
        cache_enabled=True,
        strict_validation=True
    )

Now, any attempt to save invalid data is blocked at the ORM level, and the schema is automatically reused in DRF and FastAPI!

---

🔗 Ecosystem Integration

Django Nova acts as a universal hub between Python frameworks.

Django Rest Framework

from nova.ecosystem.drf import to_drf_serializer

# Dynamically generates a ModelSerializer that delegates business logic to Pydantic
ResearcherSerializer = to_drf_serializer(Researcher)

FastAPI

from fastapi import FastAPI
from nova.ecosystem.fastapi import to_fastapi_router

app = FastAPI()
# Generates GET/POST endpoints with native OpenAPI/Swagger documentation
app.include_router(to_fastapi_router(Researcher, prefix="/api/researchers"))

---

🏗️ Architecture

Django Nova intercepts standard processes at the core level:

Request -> View -> Model.save() -> [Pydantic Validation -> Django Fields -> Business Logic] -> DB
                |
                +-> Cache Invalidation Signal -> Evict stale QuerySets
                |
                +-> OpenTelemetry Span -> Metrics & Traces
                |
                +-> Structlog -> JSON Logs to Datadog/ELK

Core Tech Stack:

• PEP 562: Lazy imports bypassing AppRegistryNotReady.
• PEP 695: Modern generic syntax (class Cache[T]:).
• SQL Compiler: Deterministic cache hash key generation (safe across any Django version).

---

🧪 Testing

The project is tested on the bleeding-edge stack (Python 3.14 + Django 5.2).

pip install -e ".[dev]"
pytest tests/ -v  # 39 passed

---

👤 Author

Artem Alimpiev

• ORCID: 0009-0007-6740-7242
• DOI: 10.5281/zenodo.20057443
• DOI: 10.5281/zenodo.20659647

---

📄 License

This project is licensed under the terms of the MIT License. See the LICENSE file for details.