aprisma 0.1.4

A production-grade Python Prisma client

Aprisma

A Python Prisma client backed by SQLAlchemy — no Node.js or Rust at runtime.

Aprisma reads a standard schema.prisma file and generates a fully-typed async Python client. All queries run through SQLAlchemy Core — SQLite, PostgreSQL, and any other SQLAlchemy dialect work out of the box.

db = Prisma()
await db.connect("sqlite+aiosqlite:///dev.db")

alice = await db.user.create(data={"email": "[e-mail]", "name": "Alice"})

posts = await db.post.find_many(
    where={"author": {"email": "[e-mail]"}, "published": True},
    include={"author": True},
    order={"createdAt": "desc"},
)

await db.disconnect()

---

Why Aprisma?

The original prisma-client-py is no longer maintained and requires a Rust query engine and a Node.js subprocess at every query. Aprisma replaces that with a pure-Python SQLAlchemy layer — the Prisma CLI is only used during development to manage migrations, never at runtime.

	prisma-client-py	Aprisma
Runtime dependency	Rust engine + Node.js	None
Query layer	Prisma Query Engine	SQLAlchemy Core
Schema format	Prisma schema	Prisma schema
Async support	Yes	Yes
Type safety	Yes	Yes (Pydantic v2)

---

Features

• Pure Python runtime — no Node.js process, no Rust binary, no subprocess at query time
• Async-first — every method is async; built for asyncio
• Fully typed — Pydantic v2 models, TypedDict query inputs, typed relation attributes
• Prisma v7 schema — use the standard schema format and the official Prisma CLI for migrations
• SQLAlchemy backend — JOIN-based relation filtering, correlated EXISTS subqueries, connection pooling
• Rich query API — filtering, ordering, pagination (offset and cursor), distinct, include, select, transactions, raw queries, group_by aggregations

---

Installation

pip install aprisma

For PostgreSQL:

pip install "aprisma[postgresql]"

Aprisma uses the Prisma CLI (a Node.js tool) for schema migrations. It is only needed during development — see Prisma CLI Setup for a one-time setup guide.

---

Quick start

1. Write a schema

# schema.prisma

generator client {
  provider  = "aprisma"
  output    = "./prisma"
  interface = "asyncio"
}

datasource db {
  provider = "sqlite"
}

model User {
  id    Int    @id @default(autoincrement())
  email String @unique
  name  String
  posts Post[]
}

model Post {
  id        Int     @id @default(autoincrement())
  title     String
  content   String
  published Boolean @default(false)
  authorId  Int
  author    User    @relation(fields: [authorId], references: [id])
}

2. Push the schema and generate the client

# Push schema to the database (creates tables)
npx prisma db push --url "file:./dev.db"

# Generate the Python client
aprisma generate --schema schema.prisma

3. Use it

import asyncio
from prisma import Prisma

async def main():
    db = Prisma()
    await db.connect("sqlite+aiosqlite:///dev.db")

    # Create a user
    alice = await db.user.create(
        data={"email": "[e-mail]", "name": "Alice"}
    )

    # Create a post linked to that user
    post = await db.post.create(
        data={"title": "Hello world", "content": "My first post.", "authorId": alice.id},
        include={"author": True},
    )
    print(post.author.name)  # Alice

    # Filter posts through a related model
    posts = await db.post.find_many(
        where={"author": {"email": "[e-mail]"}},
        include={"author": True},
        order={"title": "asc"},
    )

    # Update
    await db.post.update(
        where={"id": post.id},
        data={"published": True},
    )

    # Count
    n = await db.post.count(where={"published": True})
    print(f"{n} published post(s)")

    # Delete
    await db.post.delete(where={"id": post.id})

    await db.disconnect()

asyncio.run(main())

---

Database support

Database	Driver	Install
SQLite	aiosqlite	included
PostgreSQL	asyncpg	pip install "aprisma[postgresql]"
Any SQLAlchemy dialect	bring your own	pass the URL to connect()

---

Documentation

Full documentation: xlaszlo.github.io/aprisma

---

Contributing

Found a bug or have a feature idea? Please open an issue on GitHub Issues with a minimal reproducible description of the problem or feature. Please do not open pull requests — a clear issue description is more useful than a PR without prior discussion.