Skip to main content

Django backend for TiDB

Project description

TiDB dialect for Django

PyPI PyPI - Python Version PyPI - Downloads .github/workflows/ci.yml

This adds compatibility for TiDB to Django.

Installation Guide


Before installing django-tidb, ensure you have a MySQL driver installed. You can choose either mysqlclient(recommended) or pymysql(at your own risk).

Install mysqlclient (Recommended)

Please refer to the mysqlclient official guide

Install pymysql (At your own risk)

django-tidb has not been tested with pymysql

pip install pymysql

Then add the following code at the beginning of your Django's

import pymysql


Installing django-tidb

To install django-tidb, you need to select the version that corresponds with your Django version. Please refer to the table below for guidance:

The minor release number of Django doesn't correspond to the minor release number of django-tidb. Use the latest minor release of each.

django django-tidb install command
v5.0.x v5.0.x pip install 'django-tidb>=5.0.0,<5.1.0'
v4.2.x v4.2.x pip install 'django-tidb>=4.2.0,<4.3.0'
v4.1.x v4.1.x pip install 'django-tidb>=4.1.0,<4.2.0'
v3.2.x v3.2.x pip install 'django-tidb>=3.2.0,<3.3.0'


Set 'ENGINE': 'django_tidb' in your settings to this:

    'default': {
        'ENGINE': 'django_tidb',
        'NAME': 'django',
        'USER': 'root',
        'PASSWORD': '',
        'HOST': '',
        'PORT': 4000,
DEFAULT_AUTO_FIELD = 'django.db.models.AutoField'
USE_TZ = False
SECRET_KEY = 'django_tests_secret_key'


AUTO_RANDOM is a feature in TiDB that generates unique IDs for a table automatically. It is similar to AUTO_INCREMENT, but it can avoid write hotspot in a single storage node caused by TiDB assigning consecutive IDs. It also have some restrictions, please refer to the documentation.

To use AUTO_RANDOM in Django, you can do it by following two ways:

  1. Declare globally in as shown below, it will affect all models:

    DEFAULT_AUTO_FIELD = 'django_tidb.fields.BigAutoRandomField'
  2. Manually declare it in the model as shown below:

    from django_tidb.fields import BigAutoRandomField
    class MyModel(models.Model):
        id = BigAutoRandomField(primary_key=True)
        title = models.CharField(max_length=200)

BigAutoRandomField is a subclass of BigAutoField, it can only be used for primary key and its behavior can be controlled by setting the parameters shard_bits and range. For detailed information, please refer to the documentation.


  1. Check if the original column is BigAutoField(bigint), if not, migrate it to BigAutoField(bigint) first.

  2. In the database configuration (, define SET @@tidb_allow_remove_auto_inc = ON in the init_command. You can remove it after completing the migration.

        'default': {
            'ENGINE': 'django_tidb',
            'OPTIONS': {
                'init_command': 'SET @@tidb_allow_remove_auto_inc = ON',
  3. Finnaly, migrate it to BigAutoRandomField(bigint).


AUTO_RANDOM is supported after TiDB v3.1.0, and only support define with range after v6.3.0, so range will be ignored if TiDB version is lower than v6.3.0


AUTO_ID_CACHE allow users to set the cache size for allocating the auto-increment ID, as you may know, TiDB guarantees that AUTO_INCREMENT values are monotonic (always increasing) on a per-server basis, but its value may appear to jump dramatically if an INSERT operation is performed against another TiDB Server, This is caused by the fact that each server has its own cache which is controlled by AUTO_ID_CACHE. But from TiDB v6.4.0, it introduces a centralized auto-increment ID allocating service, you can enable MySQL compatibility mode by set AUTO_ID_CACHE to 1 when creating a table without losing performance.

To use AUTO_ID_CACHE in Django, you can specify tidb_auto_id_cache in the model's Meta class as shown below when creating a new table:

class MyModel(models.Model):
    title = models.CharField(max_length=200)

    class Meta:
        tidb_auto_id_cache = 1

But there are some limitations:

  • tidb_auto_id_cache can only affect the table creation, after that it will be ignored even if you change it.
  • tidb_auto_id_cache only affects the AUTO_INCREMENT column.

Vector (Beta)

Now only TiDB Cloud Serverless cluster supports vector data type, see Integrating Vector Search into TiDB Serverless for AI Applications.

VectorField is still in beta, and the API may change in the future.

To use VectorField in Django, you need to install django-tidb with vector extra:

pip install 'django-tidb[vector]'

Then you can use VectorField in your model:

from django.db import models
from django_tidb.fields.vector import VectorField

class Test(models.Model):
    embedding = VectorField(dimensions=3)

Create a record

Test.objects.create(embedding=[1, 2, 3])

Get instances with vector field

TiDB Vector support below distance functions:

  • L1Distance
  • L2Distance
  • CosineDistance
  • NegativeInnerProduct

Get instances with vector field and calculate distance to a given vector:

Test.objects.annotate(distance=CosineDistance('embedding', [3, 1, 2]))

Get instances with vector field and calculate distance to a given vector, and filter by distance:

Test.objects.alias(distance=CosineDistance('embedding', [3, 1, 2])).filter(distance__lt=5)

Supported versions

  • TiDB 5.0 and newer
  • Django 3.2, 4.1, 4.2 and 5.0
  • Python 3.6 and newer(must match Django's Python version requirement)


create your virtualenv with:

$ virtualenv venv
$ source venv/bin/activate

you can use the command deactivate to exit from the virtual environment.

run all integration tests.

$ DJANGO_VERSION=3.2.12 python

Migrate from previous versions

Releases on PyPi before 3.0.0 are published from repository This repository is a new implementation and released under versions from 3.0.0. No backwards compatibility is ensured. The most significant points are:

  • Engine name is django_tidb instead of django_tidb.tidb.

Known issues

  • TiDB before v6.6.0 does not support FOREIGN KEY constraints(#18209).
  • TiDB before v6.2.0 does not support SAVEPOINT(#6840).
  • TiDB has limited support for default value expressions, please refer to the documentation.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

django-tidb-5.0.0.tar.gz (22.4 kB view hashes)

Uploaded Source

Built Distribution

django_tidb-5.0.0-py3-none-any.whl (24.3 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page