Pysura - BETA

Hasura + Python = Pysura. An open source solution.

Because everyone should be a code wizard.

Requirements:

To deploy Hasura with Pysura with baked in Auth and a templated Flutter Frontend:

Python 3.9+
gcloud CLI
gcloud beta CLI
A billing account with Google Cloud
Docker (Make sure it's running in the background or your deployment will fail!)
Firebase CLI
npm
Dart
Flutter

pip install pysura
pysura
(pysura_cli) >>> setup_pysura

Note: The installer is doing a lot of things. Some of them take a long time, like creating databases, firewalls, and VPC networks. Sometimes it might look frozen, but give it some time to do its thing. It's automagically building an entire application for you. I promise doing this by hand takes longer. ;)

Useful Pysura commands:

deploy_frontend - Rebuild and Deploy the web frontend to firebase hosting for release
deploy_microservice - Regenerate routers and re-deploy the default microservice
print_pysura - Prints important pysura details like login credentials for the project
import_hasura_metadata - Import the hasura metadata from Hasura
export_hasura_metadata - Export the hasura metadata to Hasura
gcloud_deploy_hasura - Deploy the hasura microservice to GCloud
import_existing_hasura - Clone an existing Hasura empty Hasura instance. Full clone, from metadata and pg_dump including pulling ENUM values and cloning those.

Is this multi-platform compatible?

Mac - Yes, I developed it on a Mac!

Linux - It should work on Linux, it is untested. Let me know!

Windows - With minor updates it should work. I think the majority of the deployer works, but it might run into issues on the Firebase and Flutterfire CLI's.

What is Pysura?

Pysura is a CLI tool that's designed to make building and deploying multi-tenant enterprise grade applications as easy as a freshman year algorithms class. It's a highly opinionated way because it's the right way.

It's kind of like running npm init, if npm was for backends and frontends, it provides your backend, auth, and database for you in the cloud in a way that is infinitely scalable and uses zero-trust, with a type-safe GraphQL backend and a bring-your-own front-end approach with special built-in support for Flutter which is the default setup.

Why Flutter?

So that it doesn't matter what you are building, you can build it on Pysura. It's about damn time that Python developers have better mobile support. Let's bring python to mobile. Let's bring python to the web. Let's bring python to the edge. Let's bring python everywhere. And let's skin it with Flutter and feed it all the data it wants with GraphQL and Hasura.

Do I have to use Flutter for the frontend? No way! Pysura places firebase in front of your Hasura instance, so if your frontend of choice supports Firebase Auth integrations or libraries, you can use it, or even better open a PR and add a template for your provider. You do need Flutter for installation

What does the Pysura installer do?

Authenticates you with gcloud
Creates a new project in Google Cloud
Sets up a VPC network inside the project
Creates a new Cloud SQL instance running Postgres 14
Stores all Env variables in secret manager to be loaded from containers
Creates a new Hasura instance in Cloud Run with scaling settings always allocating memory for at least 1 instance
Attaches Firebase to the project
Deploys firebase functions triggering on user.create/delete managing Hasura user data each with min instances set to 1
Adds JWT authentication via Firebase to Hasura with phone sign in
Creates a new Flutter app with Firebase Auth using flutterfire
Configures Android signing keys (COMING SOON: Auto app icon configuration)
Configures iOS URL schemes (COMING SOON: Auto app icon configuration)
Updates the default flutter template to use the Pysura template for Flutter
Deploys the flutter template to Firebase hosting for the web. (COMING SOON: Auto app icon configuration)
Deploys a default microservice on Cloud Run using Pysura! Includes default event that handles JWT claim changes

You will have this exact app after running the installer, it will work on iOS, Android, Web, Mac, and as a Chrome web app. Untested on Linux/Windows native but it should work. ;)

https://pysura-demo-2.web.app/

What is a Pysura Microservice?

It's a wrapper around a FastAPI app that holds a collection of actions, events, and chron-jobs related to its function. I.e. A payment microservice might have all code related to payment processing. It bakes in Auth with an extra decorator which gives you easy access to the calling user each time a method is called. This works by using opaque tokens that pass through the headers into the microservice. It's a very simple way to do auth, and it's very secure. It gives full RBAC at a column level, and you can design rather complex auth rules using Hasura's permission system. Each action, event, and chron-job is placed in a separate file with routing already handled and the calling users information passed in the calling context. It makes it very easy to build your business logic in a clean, consistent way that is easy to test and maintain, and also easy to hand off to other developers.

Pysura uses middleware to handle the auth, and provide you access to app resources like the database, and the caller. It wraps FastAPI, and uses a router for each action, event, or chron-job. Separating the business logic into collections of microservices each containing folders of actions, events, and chron-jobs each delegated to their own file which makes it easy to scale and maintain.

Front-end clients should utilize the Hasura GraphQL API to access and mutate the data in the database. This is the recommended way to access the data. It is important to make sure you set up proper roles and permissions in Hasura to ensure that only the correct users can access the data meant for them. By default, Pysura includes a user and admin role

Actions are used to allow the front-end client a way to perform custom business logic, or do a task that requires calling third party APIs or services. Actions are called via the Hasura GraphQL API, and are executed on the microservice instance.

Events are used to trigger custom business logic when something occurs in the database. For example, when a user is created, you might want to send them a welcome email. Events are triggered by Hasura (via the Database), and are executed on the microservice instance.

Crons are used to trigger custom business logic on a schedule. For example, you might want to send a report to the admin every day at 8am. Crons are triggered by Hasura (via the Database), and are executed on the microservice instance.

By default, the application is configured with an event secret that must be passed to all microservices in the header from Hasura. This ensures that all traffic is coming from Hasura, and not from a malicious source. This makes things SUPER secure, with JWT auth, and event secrets. It's a very secure way to build an application.

The default microservice directory structure looks like this:

microservices/default
- microservices/default/Dockerfile
- microservices/default/README.md
microservices/default/actions
- microservices/default/actions/init.py
- microservices/default/actions/action_mutation_pong.py
- microservices/default/actions/action_query_ping.py
- microservices/default/actions/action_upload_file.py
- microservices/default/app.py
- microservices/default/app_secrets.py
microservices/default/crons
- microservices/default/crons/init.py
- microservices/default/crons/cron_update_app_message.py
- microservices/default/deploy.txt
microservices/default/events
- microservices/default/events/init.py
- microservices/default/events/event_update_user_cascade.py
- microservices/default/events/event_update_user_role.py
- microservices/default/generated_types.py
- microservices/default/pysura_metadata.json
- microservices/default/requirements.txt

Here, you can see app secrets are loaded from Secret manager in the app_secrets.py file:

app_secrets.py

from google.cloud import secretmanager

PROJECT_ID = "pysura-demo-2"


def get_secret_by_id(secret_id, version_id="latest"):
    client = secretmanager.SecretManagerServiceClient()
    name = f"projects/{PROJECT_ID}/secrets/{secret_id}/versions/{version_id}"
    response = client.access_secret_version(name=name)
    return response.payload.data.decode("UTF-8")


print("LOADING SECRETS")
HASURA_GRAPHQL_ADMIN_SECRET = get_secret_by_id("HASURA_GRAPHQL_ADMIN_SECRET")
HASURA_GRAPHQL_URL_ROOT = get_secret_by_id("HASURA_GRAPHQL_URL_ROOT")
HASURA_EVENT_SECRET = get_secret_by_id("HASURA_EVENT_SECRET")
HASURA_FIREBASE_SERVICE_ACCOUNT = get_secret_by_id("HASURA_FIREBASE_SERVICE_ACCOUNT")
HASURA_STORAGE_BUCKET = get_secret_by_id("HASURA_STORAGE_BUCKET")
HASURA_GRAPHQL_DATABASE_URL = get_secret_by_id("HASURA_GRAPHQL_DATABASE_URL")

In the app.py file, we inject the Pysura security middleware:

@app.middleware("http")
async def inject_security(request: Request, call_next):
    return await security_injection_middleware(request, call_next)

This is going to handle injecting the identity of the caller into the function directly from the JWT token.

The composition of all events, actions, and crons is as follows:

The PysuraSecurity callable class is used to inject the security middleware into the FastAPI app. It makes it easy to specify whether or not the route requires JWT authentication, what level of JWT authentication is required, and if the caller must send the backend event secret.

The PysuraProvider callable class providers the calling function with access to any resources that it might need.

This is what a full action looks like:

# (IMPORTS-START) - DO NOT DELETE THIS LINE!
import logging

from fastapi import APIRouter, Depends, Request
from pysura.faster_api.security import PysuraSecurity, PysuraProvider, Provider
from pysura.faster_api.enums import ApiResponse, ClientRole
from generated_types import *

# (IMPORTS-END) - DO NOT DELETE THIS LINE!

ROUTE = "/action_mutation_pong/"
ALLOWED_ROLES = [  # The roles allowed to call this action
    ClientRole.admin.name, ClientRole.user.name
]
action_mutation_pong_router = APIRouter(
    tags=["action_mutation_pong"]
)


@action_mutation_pong_router.post(ROUTE,
                                  dependencies=[
                                      Depends(PysuraSecurity(
                                          require_jwt=True,
                                          require_event_secret=True,
                                          allowed_roles=ALLOWED_ROLES
                                      ))
                                  ],
                                  response_model=ActionMutationPongOutput
                                  )
async def action_mutation_pong(_: Request,
                               action_mutation_pong_input: ActionMutationPongInput | None = None,
                               provider: Provider | None = Depends(PysuraProvider(
                                   # (DEPENDENCY-INJECTION-START) - DO NOT DELETE THIS LINE!
                                   provide_identity=True,
                                   provide_firebase=True,
                                   provide_graphql=True,
                                   provide_storage=True,
                                   provide_sql=True
                                   # (DEPENDENCY-INJECTION-END) - DO NOT DELETE THIS LINE!
                               ))):
    # (BUSINESS-LOGIC-START) - DO NOT DELETE THIS LINE!
    user_id = provider.user_identity.user_id
    provider.graphql.execute(
        query="""mutation Pong($user_id: String = "", $message: String = "", $number: Int = 10) {
  update_public_user_by_pk(pk_columns: {user_id: $user_id}, _set: {message: $message, number: $number}) {
    user_id
  }
}""",
        variables={
            "user_id": user_id,
            "message": action_mutation_pong_input.message,
            "number": action_mutation_pong_input.number
        }
    )
    response = ActionMutationPongOutput(
        data=ActionMutationPongOutputData(
            message=action_mutation_pong_input.message,
            number=action_mutation_pong_input.number
        ),
        nodes=None,
        response_name=ApiResponse.SUCCESS.name,
        response_value=ApiResponse.SUCCESS.value
    ).dict()
    return response
    # (BUSINESS-LOGIC-END) - DO NOT DELETE THIS LINE!

The provider is very powerful:

class Provider(BaseModel):
    user_identity: UserIdentity | None = None
    firebase_app: App | None = None
    graphql: PysuraGraphql | None = None
    storage: PysuraStorage | None = None
    sql: PysuraSQLClient | None = None

    class Config:
        arbitrary_types_allowed = True
        json_encoders = {
            App: lambda v: v if v is None else v.name,
            PysuraGraphql: lambda v: v if v is None else v.name,
            PysuraStorage: lambda v: v if v is None else v.name,
            PysuraSQLClient: lambda v: v if v is None else v.name,
        }


class UserIdentity(BaseModel):
    token: str | None = None
    role: str | None = None
    user_id: str | None = None
    allowed_roles: List[str] | None = None
    phone_number: str | None = None
    email: str | None = None
    iss: str | None = None
    aud: str | None = None
    iat: int | None = None
    exp: int | None = None
    auth_time: int | None = None
    sub: str | None = None


class PysuraGraphql(GraphqlClient):

    def __init__(self):
        self.name = "graphql_client"
        super().__init__(endpoint=HASURA_GRAPHQL_URL_ROOT)

    def execute(self):
        """Executes a query as Admin"""
        pass

    def execute_as_user(self):
        """Executes a query as the caller passing the JWT token"""
        pass

    async def execute_async(self):
        """Executes a query as Admin asynchronously"""
        pass

    async def execute_async_as_user(self):
        """Executes a query as the caller passing the JWT token asynchronously"""
        pass


class PysuraStorage(google_storage.Client):

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs, project=cred_dict['project_id'], credentials=creds)
        self.name = "storage_client"
        self.default_bucket = HASURA_STORAGE_BUCKET

    def upload_file(self, file_data: bytes | str, file_name: str, file_type: str, user_id: str):
        pass


class PysuraSQLClient:

    def get_conn(self):
        pass

    def get_engine(self):
        pass

    def get_db(self):
        pass

    def refresh_engine(self):
        pass

    def set_connection_url(self, connection_url: str):
        pass

Ex.

from pysura.faster_api.security import Provider

kwargs = {}
p = Provider(
    **kwargs
)

print(p.user_identity.user_id)
response = p.graphql.execute(
    query="""query SomeGraphql { some_graphql }""",
    variables={}
)
print(response["data"]["some_graphql"])

from firebase_admin import auth

user = auth.get_user(p.user_identity.user_id, app=p.firebase_app)
print(user)


response = p.storage.upload_file(
    file_data="b64_string",
    file_name="img2.png",
    file_type="image/png",
    user_id=p.user_identity.user_id
)

print(response["signed_url"])
print(response["user_id"])

Do I need to deploy Hasura with Pysura to use it?

As of right now yes. In the future, no!

Is this just a wrapper for the Hasura CLI?

Pysura does not use the Hasura CLI, and instead manages the metadata directly via retrieving it and overwriting it.

It's important to note that Pysura will add the local machine of its installer's IP to the allow-list for the database. This is to allow pysura to connect via pscopg2 for things like table management. You'll need to de-authorize the network if you want to remove the machine from the allow-list once setup is complete. TODO: Make a deauthorize command.

Limitations:

Currently, this only supports a Google-based deployment/stack. Easily fixable. Just need a fellow AWS/Azure/Etc. wizard who can translate the gcloud commands. I think the portability is pretty high since everything is done from the CLI. The setup wizard is just curling everything.

Neat! But it doesn't support my cloud provider, when will you be adding AWS/Azure/Etc. support?

I won't. You can. I built this because I needed it. If our stack changes providers, or potentially if someone was willing to pay for it, I'd be willing to add support for other providers. But as of now, the best bet would be to open a PR and add support for your provider of choice. I tried to design it to be pretty modular. The gcloud --format=json flag is what changed the game when I found it. Maybe try ChatGPT to translate the gcloud commands to AWS/Azure/Etc. commands.

FAQ:

Q: I broke the installer, something wasn't enabled properly. (I didn't have docker running, didn't have gcloud installed, etc.)

A: Best bet is to try to rerun the installer, it might throw some errors but it should recover for the most part. If it doesn't, then you should trash the project and rerun the setup. Just remember to go to billing and disable the old project, and shut it down.

Q: How do I contribute?

A: Read over the current code, and see if you can make it better. If you can, open a PR. If you can't, open an issue or a feature request for something that would make it better.

Q: Is this affiliated with Hasura, Google, or Firebase?

A: No, this has no affiliation with Hasura, Google, or Firebase. In no way should this be considered an official product of any of these companies. Although this makes heavy use of tools and API's provided it is in no way an official product of any company or tool used.

Q: I'm not sure what to enter for one of the command line prompts, what should I do?

A: If you aren't sure what to enter, most of the time the provided example should work.

Q: One of the choices in the list of options for a prompt doesn't make sense, or there is something missing!

A: Please open an issue or a feature request. I'm sure I missed some things. Stitching this together required a bit of everything, including working outside my area of expertise.

TODOS:

Remove pyscopg2 dependency to drop postgres req

DONE:

HIGH PRIORITY:

LOW PRIORITY:

Add support for AWS/Azure/Etc.
Add support for other databases (Postgres, MySQL, BigQuery, etc)
Add support for other frontends (React, Vue, etc)
Rewrite the CLI tool in a more modular way, with command plug-ins.
It should be possible to introspect the entirety of the gcloud CLI by parsing the output of gcloud help. And walking the CLI tool This would allow for a more dynamic CLI tool that could be used for other projects.

tristengoodz / pysura