Hello World dApp

By following this tutorial, you'll be able to get more confident about:

• Preparing a Docker image for your app with the ultimate purpose of deploying it on Golem.
• Converting the VM image to Golem and uploading it to Golem's repository.
• Creating a descriptor reflecting your app using YAML syntax similar to that used by docker-compose.
• Deploying your app to Golem using dapp-runner.

To follow this tutorial in full, you need to have Docker installed on your machine. If you don't have it installed, please refer to the instructions on Docker's website.

The example that we're going to show you here, uses a very simple Python HTTP server.

On the other hand, if you're more acquainted with any other language or platform that enables you to easily create a simple HTTP server and pack it into a Docker image, go for it.

Just as well - if the setup and construction of such a simple app is obvious to you, feel free to jump to Converting the image to Golem

For the sake of completeness, we're including the steps that prepare our Python environment. Again, you're free to do it your way and skip to Hello World dApp:

python3 -m venv ~/.envs/hello-dapps
source ~/.envs/hello-dapps/bin/activate
pip install -U pip poetry

mkdir -p hello_golem/server_app
cd hello_golem/server_app/
poetry init --no-interaction --python="^3.9"

In our little example here, we'll use Flask, a very robust and minimal Python back-end framework.

poetry add Flask

There, we're ready to start coding our app.

git clone https://github.com/golemfactory/dapp-experiments.git
cd dapp-experiments/05_hello_world

Fire up your favourite editor or IDE and navigate to the hello_golem/server_app directory that we have set up above.

If you have configured the app using poetry like we did above, you should already see the pyproject.toml and poetry.lock files.

In the directory, add hello_golem.py and just paste the following few lines which are a slightly modified version of Flask's original, minimal example:

from flask import Flask
app = Flask(__name__)
@app.route('/')
def hello_golem():
    return 'Hello from Golem!'
app.run(host="0.0.0.0")

The above code achieves the following:

1. Import the Flask engine
2. Define our Flask app
3. Set up the root route for the app
4. Launch the Flask server

Let's test this app locally, before we start preparing it for Golem. Launch the app:

python hello_golem.py

and then point your browser to http://127.0.0.1:5000.

You should see "Hello from Golem!" which proves that our minimal app is working correctly when launched completely stand-alone. Good!

We need to start with a Docker image since the only currently-supported way of creating a GVMI (Golem Virtual Machine Image) file is by providing a Docker image for conversion.

If you'd like to know more about GVMI images and about the conversion process, please refer to: Golem Images explained

Let's back-up a little and ensure that we're one directory above server_app.

cd ..

To build a docker image, we'll need a file called Dockerfile with the following contents:

FROM python:3.9-slim
RUN pip install -U pip poetry
RUN mkdir /app
COPY server_app/* /app/
WORKDIR /app
RUN poetry install --no-root
ENTRYPOINT poetry run python hello_golem.py

Let's now go through what happens there.

1). Take a stock Python Docker image (the slim version, we won't need anything more here):

FROM python:3.9-slim

2). Install the needed tools:

RUN pip install -U pip poetry

3). Copy the app contents to the image:

RUN mkdir /app
COPY server_app/* /app/

4). Install the app's requirements:

WORKDIR /app
RUN poetry install --no-root

5). Finally, set up the Docker's entrypoint:

ENTRYPOINT poetry run python hello_golem.py

With the Dockerfile ready, we can test whether the app works inside the container:

docker build -t hello-dapps .
docker run -it -p 5000:5000 hello-dapps

Once again, after the image launches, you may connect to http://127.0.0.1:5000/ with your browser, and you should be able to see the "Hello..." message, which confirms that the app has been correctly packaged and is working as it should.

To be able to use our newly-created image on Golem, we need to convert it to Golem's custom format and then make the image available to providers. The easiest way to achieve the latter is to upload the image to our image repository.

pip install gvmkit-build

gvmkit-build golem-example --push --nologin

Once the command completes, you should get a line containing the image hash, e.g.:

image already generated
on 102777844: success. hash link 3032b6e97914eb5ee87d71188180d271f04eb9472b6da0d308943b2f

The hash there: 3032b6e97914eb5ee87d71188180d271f04eb9472b6da0d308943b2f is what we're interested in.

If the command fails because the image has already been uploaded, feel free to use the hash above and follow through with the tutorial.

Now that we have the image ready, uploaded into the Golem repo, and its hash in hand, we can prepare the descriptor of the application that we're going to launch on Golem.

For more in-depth information about the dApp descriptors, please consult the appropriate section of the "Creating Golem dApps" article. Here, we're just covering the bare minimum.

Here's what it looks like:

payloads:
  hello:
    runtime: "vm"
    params:
      image_hash: "3032b6e97914eb5ee87d71188180d271f04eb9472b6da0d308943b2f"
nodes:
  hello:
    payload: "hello"
    init:
      - run:
          args: ["/bin/sh", "-c", "poetry run python hello_golem.py > /dev/null &"]
    http_proxy:
      ports:
        - "5000"

Let's add it as hello_golem.yaml in the hello_golem directory.

There are two obligatory elements in it, the payload and nodes.

The payload is the definition of what kind of activity you'd like the providers to run on your behalf.

In this case, it's just the reference to the VM image we just uploaded.

payloads:
  hello:
    runtime: "vm"
    params:
      image_hash: "3032b6e97914eb5ee87d71188180d271f04eb9472b6da0d308943b2f"

A node entry defines the parameters of the specific instances of services that we want launched on Golem.

nodes:
  hello:
    payload: "hello"
    init:
      - run:
          args: ["/bin/sh", "-c", "poetry run python hello_golem.py > /dev/null &"]
    http_proxy:
      ports:
        - "5000"

A couple of important details here.

Firstly, given the fact that Docker's ENTRYPOINT is not yet supported by the dapp-runner, our service definition must contain any and all commands that will start our service in the container.

Moreover, as the commands included in the init must finish before the service can be treated as started, we need to put the service in the background. For this reason, we're launching our service using the shell and adding the ampersand (&) to the end of the command.

One more caveat is that we need to redirect the output stream of the launched app so that it stays running after we exit the shell.

Lastly, we're adding the http_proxy element because our service is an HTTP app which we wish to be able to talk to. There's currently no way for the service to be exposed directly on the provider's own address but we can use the local HTTP proxy functionality to expose a port on our own machine that will forward traffic to the app through the Golem Network.

There's one more piece of data that we'll need to run our application. It's the configuration of our Golem requestor that we need to supply for the dapp-runner.

Unless you want to customize your set-up, it'll be easiest to just use the default that comes with the dapp-runner, which you can get with:

curl https://raw.githubusercontent.com/golemfactory/dapp-runner/main/configs/default.yaml > golem_config.yaml

First, let's make sure that you have your yagna service up and running and if not, execute this in another console session:

yagna service run

Please, run:

yagna app-key list

and copy the value listed in the key column.

If the above command doesn't give you any 32-char keys, just create your app key:

shell yagna app-key create requestor

and copy the value output by this command.

export YAGNA_APPKEY=<your key>

python3 -m venv --clear ~/.envs/dapp-runner
source ~/.envs/dapp-runner/bin/activate
pip install dapp-runner

Now you're ready to start the app on Golem.

dapp-runner start --config golem_config.yaml hello_golem.yaml

Once the app launches, you should see some status messages describing various stages of the deployment. And finally, you should see:

{"hello": {"local_proxy_address": "http://localhost:8081"}}

The port may be different on your machine, since it's assigned automatically. Copy the address that you get there and paste that into your browser.

Assuming everything went well, you've just managed to create and deploy your own decentralized application on Golem.