PyForge

Rust-accelerated serialization and validation for Django — drop-in, zero rewrite.

What Is PyForge

PyForge moves Django REST Framework's serialization and validation hot paths from Python to Rust. It reads your existing Django model definitions, compiles a typed schema once at startup, and processes field values through Rust-native code on every request. The result is 3-8x faster API responses for list views, bulk operations, and any endpoint that touches more than a handful of records.

It is designed for Django developers running production APIs on DRF who need better throughput without migrating to a different framework.

Benchmarks

Measured on CPython 3.12, Rust 1.93, macOS ARM64. Model: 9-field RentalApplication (CharField, DecimalField, DateTimeField, UUIDField, etc.)

Small batches (<10 records) show minimal gain due to the Python/Rust bridge overhead (~5-8µs per call).

Full methodology and per-field-type breakdown: BENCHMARKS.md

Installation

pip install pyforge-django

Add the app to your Django settings:

INSTALLED_APPS = [
    ...
    "django_pyforge",
]

Quickstart

from django_pyforge import ModelSchema, serialize_instance, validate_instance
from django_pyforge.serializers import RustSerializerMixin
from rest_framework import serializers

# 1. Compile a schema once (e.g., in AppConfig.ready)
schema = ModelSchema(RentalApplication)

# 2. Serialize a single instance — one Rust call, all fields
result = serialize_instance(instance, schema)

# 3. Validate an instance against the schema
report = validate_instance(instance, schema)
if not report["is_valid"]:
    print(report["errors"])

# 4. Or use the DRF mixin — zero changes to your existing code
class ApplicationSerializer(RustSerializerMixin, serializers.ModelSerializer):
    class Meta:
        model = RentalApplication
        fields = "__all__"

Supported Django Field Types

Requirements

• Python 3.11 or newer
• Django 4.2 LTS or 5.x
• Rust 1.75+ (only needed when building from source)
• Platforms: Linux, macOS, Windows (manylinux wheels published to PyPI)

Architecture

PyForge has three layers. ModelSchema is a Rust #[pyclass] that reads Django's _meta API once and compiles field descriptors into a cached struct. serialize_instance takes a Django model object and the cached schema, extracts every field value via getattr in a single Rust call, converts each to a native Rust type (Decimal, DateTime, UUID, etc.), serializes to JSON-compatible output, and returns a Python dict. RustSerializerMixin sits on top as a DRF mixin that automatically classifies which fields are simple model fields (Rust-accelerated) and which are computed fields (delegated back to DRF). For validation batches above 64 fields, Rayon distributes work across CPU cores while the GIL is released.

Crates

License

MIT — Abdulwahed Mansour