transports 0.12.2

transports moves typed models across any wire as incremental patches.

A model is defined once as a pydantic model, stdlib dataclass, or msgspec struct. A Python process hosts it and mutates it normally, and transports sends only the patch needed to update a remote mirror. The same Rust core drives Python (PyO3) and JavaScript (wasm), so both sides share the same Value type, diff, patch application, and codecs.

import transports
from pydantic import BaseModel

class Device(BaseModel):
    name: str
    on: bool = False

session = transports.Session()
lamp = Device(name="lamp")
model_id = session.host(lamp)

lamp.on = True
print(session.drain())
# [(1, {'rev': 1, 'ops': [{'Set': {'path': [{'Key': 'on'}], 'value': {'Bool': True}}}]})]

Install

pip install transports
pip install "transports[connections]"   # WebSocket server/client adapters
pip install "transports[sse]"           # Server-Sent Events adapter
pip install "transports[jupyter]"       # Jupyter comm adapter

Documentation

Tutorial

Quickstart builds one hosted model and mirrors it with the in-process server/client protocol.

How-to guides

Model bridges covers pydantic, dataclasses, msgspec, schemas, and plain JavaScript objects. Connections covers WebSocket, SSE, Jupyter comm, and anywidget adapters. Codecs covers JSON, MessagePack, and custom codecs. Multi-tenancy and sharing covers tenant isolation, shared models, and merge strategies. Pyodide covers building and loading transports in a browser.

Reference

API reference lists the Python API. Wire protocol reference describes Value, patch operations, protocol messages, codecs, and model ids. Benchmark results reports recorded server fan-out latency, completion time, and resource use.

Explanation

Concepts explains why transports uses incremental patches, server-owned revisions, a shared Rust core, and transport-agnostic adapters.