NatsAPI is a framework to develop Python3 applications that uses nats as communication medium instead of http. With nats you have a smaller footprint, faster req/s, pub/sub and better observability. NatsAPI is highly inspired by FastAPI and has the same development style. NatsAPI produces an AsyncAPI schema out of the box, this schema is not fully compatible with the standard waiting for version 3.0.0 to support request/reply pattern.
This library has support for python > 3.9 and pydantic v1 and v2.
Optional: Run a nats server in docker so you are able to connect:
docker run --rm --name nats -dp 4222:4222 nats:latest -js -DV
$ git clone git@github.com:wegroupwolves/natsapi.git && cd natsapi
$ poetry install
$ poetry run python examples/minimal.py
$ nats request natsapi.persons.greet '{"params": {"person": {"first_name": "Foo", "last_name": "Bar"}}}'
the output should look as follows:
18:19:00 Sending request on "natsapi.persons.greet"
18:19:00 Received on "_INBOX.dpBgTyG9XC5NhagdqRHTcp.eMkVkru8" rtt 1.052463ms
{"jsonrpc": "2.0", "id": "c2bc2d20-dbd5-4e39-a22d-c22a8631c5a3", "result": {"message": "Greetings Foo Bar!"}, "error": null}
UNDER CONSTRUCTION
from natsapi import NatsAPI, SubjectRouter
from pydantic import BaseModel
class Message(BaseModel):
message: str
class Person(BaseModel):
first_name: str
last_name: str
app = NatsAPI("natsapi")
router = SubjectRouter()
@router.request("persons.greet", result=Message)
async def greet_person(app: NatsAPI, person: Person):
return {"message": f"Greetings {person.first_name} {person.last_name}!"}
app.include_router(router)
if __name__ == "__main__":
app.run()Run as follows:
$ python app.pyDocs will be rendered as:
Send a request:
from natsapi import NatsAPI
import asyncio
async def main():
app = await NatsAPI("client").startup()
params = {"person": {"first_name": "Foo", "last_name": "Bar"}}
r = await app.nc.request("natsapi.persons.greet", params=params, timeout=5)
print(r.result)
asyncio.run(main())
#> {'message': 'Greetings Foo Bar!'}or on the command line
$ nats request natsapi.persons.greet '{"params": {"person": {"first_name": "Foo", "last_name": "Bar"}}}'
18:19:00 Sending request on "natsapi.persons.greet"
18:19:00 Received on "_INBOX.dpBgTyG9XC5NhagdqRHTcp.eMkVkru8" rtt 1.052463ms
{"jsonrpc": "2.0", "id": "c2bc2d20-dbd5-4e39-a22d-c22a8631c5a3", "result": {"message": "Greetings Foo Bar!"}, "error": null}from natsapi import NatsAPI, SubjectRouter
import logging
from pydantic import ValidationError
from sentry_sdk import configure_scope
from natsapi.models import JsonRPCRequest, JsonRPCError
from pydantic import BaseModel
class StatusResult(BaseModel):
status: str
app = NatsAPI("natsapi-example")
router = SubjectRouter()
@router.request("healthz", result=StatusResult)
async def handle_user(app: NatsAPI):
return {"status": "OK"}
app.include_router(router)
def configure_sentry(auth):
with configure_scope() as scope:
scope.user = {
"email": auth.get("email"),
"id": auth.get("uid"),
"ip_address": auth.get("ip_address"),
}
@app.exception_handler(ValidationError)
async def handle_validation_exception(exc: ValidationError, request: JsonRPCRequest, subject: str) -> JsonRPCError:
auth = request.params.get("auth") or {}
configure_sentry(auth)
logging.error(
exc,
exc_info=True,
stack_info=True,
extra={"auth": auth, "json": request.dict(), "subject": subject, "NATS": True, "code": -32003},
)
return JsonRPCError(code=-90001, message="VALIDATION_ERROR", data={"error_str": str(exc)})
if __name__ == "__main__":
app.run(reload=False)To see the documentation, you can use the binary to run the server. Root path is natsapi-example so:
$ ./nats-redoc 4222 master.trinity-testing
Server running
Docs can be found on localhost:8090
connected to nats on port 4222When surfing to localhost:8090, the documentation should look like this:
Plugins can be added and are found in natsapi/plugin.py.
- natsapi_mock: A handy mock fixture to intercept nats requests and to fake nats responses for any subject.
- Add Request/Reply AsyncApi support
- Hot reloading (when saving source code, application should be reloaded)
- Fancy readme
- Better benchmark
- Add support for 'side effect' testing so that you can have more than 1 response for same mocked route, based on ordering. See respx
- Better CI/CD -> with Python 3.13
- Make
nkeysoptional - Pydantic V2 support (nice to have)

