An fast and powerful RPC framework based on ASGI/WSGI. Based on WSGI/ASGI, you can deploy the rpc.py server to any server and use http2 to get better performance. And based on httpx's support for multiple http protocols, the client can also use http/1.0, http/1.1 or http2.
You can freely use ordinary functions and asynchronous functions for one-time response. You can also use generator functions or asynchronous generator functions to stream responses.
Install from PyPi:
pip install rpc.py
# need use client
pip install rpc.py[client]
# need use pydantic type hint or OpenAPI docs
pip install rpc.py[type]
# need use msgpack to serializer
pip install rpc.py[msgpack]
# need use CBOR to serializer
pip install rpc.py[cbor]
# or install all dependencies
pip install rpc.py[full]
Install from github:
pip install git https://github.com/abersheeran/[email protected]
Use ASGI
mode to register async def
...
from typing import AsyncGenerator
from typing_extensions import TypedDict
import uvicorn
from rpcpy import RPC
app = RPC(mode="ASGI")
@app.register
async def none() -> None:
return
@app.register
async def sayhi(name: str) -> str:
return f"hi {name}"
@app.register
async def yield_data(max_num: int) -> AsyncGenerator[int, None]:
for i in range(max_num):
yield i
D = TypedDict("D", {"key": str, "other-key": str})
@app.register
async def query_dict(value: str) -> D:
return {"key": value, "other-key": value}
if __name__ == "__main__":
uvicorn.run(app, interface="asgi3", port=65432)
OR
Use WSGI
mode to register def
...
from typing import Generator
from typing_extensions import TypedDict
import uvicorn
from rpcpy import RPC
app = RPC()
@app.register
def none() -> None:
return
@app.register
def sayhi(name: str) -> str:
return f"hi {name}"
@app.register
def yield_data(max_num: int) -> Generator[int, None, None]:
for i in range(max_num):
yield i
D = TypedDict("D", {"key": str, "other-key": str})
@app.register
def query_dict(value: str) -> D:
return {"key": value, "other-key": value}
if __name__ == "__main__":
uvicorn.run(app, interface="wsgi", port=65432)
Notice: Regardless of whether the server uses the WSGI mode or the ASGI mode, the client can freely use the asynchronous or synchronous mode.
Use httpx.Client()
mode to register def
...
from typing import Generator
from typing_extensions import TypedDict
import httpx
from rpcpy.client import Client
app = Client(httpx.Client(), base_url="http://127.0.0.1:65432/")
@app.remote_call
def none() -> None:
...
@app.remote_call
def sayhi(name: str) -> str:
...
@app.remote_call
def yield_data(max_num: int) -> Generator[int, None, None]:
yield
D = TypedDict("D", {"key": str, "other-key": str})
@app.remote_call
def query_dict(value: str) -> D:
...
OR
Use httpx.AsyncClient()
mode to register async def
...
from typing import AsyncGenerator
from typing_extensions import TypedDict
import httpx
from rpcpy.client import Client
app = Client(httpx.AsyncClient(), base_url="http://127.0.0.1:65432/")
@app.remote_call
async def none() -> None:
...
@app.remote_call
async def sayhi(name: str) -> str:
...
@app.remote_call
async def yield_data(max_num: int) -> AsyncGenerator[int, None]:
yield
D = TypedDict("D", {"key": str, "other-key": str})
@app.remote_call
async def query_dict(value: str) -> D:
...
You can also write two copies of code in one place. Just make sure that server.register
is executed before client.remote_call
.
import httpx
from rpcpy import RPC
from rpcpy.client import Client
server = RPC()
client = Client(httpx.Client(), base_url="http://127.0.0.1:65432/")
@client.remote_call
@server.register
def sayhi(name: str) -> str:
return f"hi {name}"
if __name__ == "__main__":
import uvicorn
uvicorn.run(app, interface="wsgi", port=65432)
If you need to deploy the rpc.py server under example.com/sub-route/*
, you need to set RPC(prefix="/sub-route/")
and modify the Client(base_path=https://example.com/sub-route/)
.
Currently supports three serializers, JSON, Msgpack and CBOR. JSON is used by default. You can override the default JSONSerializer
with parameters.
from rpcpy.serializers import MsgpackSerializer, CBORSerializer
RPC(
...,
response_serializer=MsgpackSerializer(),
)
# Or
Client(
...,
request_serializer=CBORSerializer(),
)
Thanks to the great work of pydantic, which makes rpc.py allow you to use type annotation to annotate the types of function parameters and response values, and perform type verification and JSON serialization . At the same time, it is allowed to generate openapi documents for human reading.
If you want to open the OpenAPI document, you need to initialize RPC
like this RPC(openapi={"title": "TITLE", "description": "DESCRIPTION", "version": "v1"})
.
Then, visit the "{prefix}openapi-docs"
of RPC and you will be able to see the automatically generated OpenAPI documentation. (If you do not set the prefix
, the prefix
is "/"
)
Currently, file upload is not supported, but you can do this by passing a bytes
object.