mariax 1.1

Django ORM & Python library for MariaDB Vector support

mariax

MariaDB extended helpers focused on Vector support.

Architecture

                 ┌───────────────────────┐
                 │   Python App          │
                 │(Django / Raw Scripts) │
                 └─────────┬─────────────┘
                           │
             ┌─────────────▼─────────────┐
             │       Library Layer       │
             │       (Mariax)            │
             └─────────────┬─────────────┘
                           │
           ┌───────────────┴───────────────┐
           │                               │
    ┌──────▼──────┐                 ┌──────▼──────┐
    │ Django ORM  │                 │ SQL/CLI     │
    │ Integration │                 │ Utilities   │
    │ (VectorField│                 │ (DBClient,  │
    │ + Manager)  │                 │ DDL helpers │
    │ + QuerySet) │                 │ query.py)   │
    └──────┬──────┘                 └──────┬──────┘
           │                               │
           │                               │
           ▼                               ▼
    ┌───────────────┐                ┌───────────────┐
    │ Vector Utils  │                │ MariaDB 11.x  │
    │ (serialize /  │<──────────────>│ Vector Column │
    │ deserialize)  │                │ + Vector Index│
    └───────────────┘                └───────────────┘

TL;DR

• Low-level DB client (client.py)
• Vector utilities (vector.py)
• DDL helpers (ddl.py)
• Raw vector query function (query.py)
• CLI tool for index management (cli.py)
• Django ORM integration (VectorField, VectorManager)
• SQLAlchemy integration (VectorType)

Installation

pip install mariax

Example DDL

from mariax.django_integration.fields import VectorField
from mariax.django_integration.managers import VectorManager

class Document(models.Model):
    title = models.CharField(max_length=255)
    embedding = VectorField(dim=768)

    objects = VectorManager()

Example Query

similarity_search(db_client, table="docs", embedding_field="embedding", vector=[...], top_k=5)
# OR
top_docs = Document.objects.similarity_search(query_vector, top_k=5)

Example CLI

mariax create-index --table docs --name idx_docs_embedding --fields embedding --distance cosine