Skip to content

Python 3 Guide


WIP: The iroha-python SDK only works with the iroha2-lts for now. It applies both to the iroha2-edge and the iroha2 branches. Our team recommends using the iroha2-edge branch while we update the iroha2 one.

1. Iroha 2 Client Setup

There are two versions of Iroha Python to choose from. In theory, the Iroha 1 version of Iroha Python (that also has the best documentation) should be compatible with an Iroha 2 deployment.

Thus we should build and install the Iroha 2 compatible version of Iroha-python, using (for now) its GitHub repository.

Let's create a separate folder for Iroha Python and clone its GitHub repository into it:

$ cd ~/Git/
$ git clone --branch iroha2
$ cd iroha-python

Iroha Python is written in Rust using the PyO3 library. Thus, unlike most Python packages, you must build it first:

$ pip install maturin
$ maturin build

After the build is complete, install it into your system:

$ pip install ./target/wheels/iroha_python-*.whl

Finally, you will need a working client configuration:

$ cp -vfr ~/Git/iroha/configs/client/config.json example/config.json


You can also use the provided config.json in the example folder if you also call docker compose up from that same folder. This has to do with the fact that the configuration for the Docker files in Iroha Python is slightly different.

2. Configuring Iroha 2

Unlike iroha_client_cli, finding the configuration file in a scripting language is the responsibility of the person writing the script. The easiest method is to de-serialise a dictionary type from the provided config.json.

This is an example of how you could do that in Python:

import json
from iroha2 import Client

cfg = json.loads(open("config.json").read())
cl = Client(cfg)

If the configuration file is malformed, you can expect an exception to notify you. However, the client doesn't do any verification: if the account used in config.json is not in the blockchain or has the wrong private key, you won't know that until you try and execute a simple instruction. More on that in the following section.


It should also be noted that Iroha Python is under heavy development. It severely lacks in documentation and its API can be made more idiomatically Python.

3. Registering a Domain

It is important to remember that Iroha Python is wrapping Rust code. As such, many of Python idioms are not yet accommodated; for example, there's no duck-typing of the Register instruction.

from iroha2.data_model.isi import *
from iroha2.data_model.domain import *

domain = Domain("looking_glass")
register = Register(Expression(Value(Identifiable(domain))))

Instead, we are creating a domain and wrapping it in multiple type-erasing constructs. A domain is wrapped in Identifiable (which would be a trait in Rust), which is wrapped in Value, which is wrapped in Expression, which finally is wrapped in the Register instruction. This is not entirely against Python conventions (it is strongly typed, after all), and not entirely counter-intuitive, once you see the corresponding Rust code.

The instruction to register must be submitted, in order for anything to happen.

hash = cl.submit_isi(register)

Note that we also keep track of the hash of the transaction. This will become useful when you visualize the output.

4. Registering an Account

Registering an account is similar to the process of registering a domain, except the wrapping structures are different. There are a couple of things to watch out for.

First of all, we can only register an account to an existing domain. The best UX design practices dictate that you should check if the requested domain exists now, and if it doesn't, suggest a fix to the user.

from iroha2.data_model.isi import *
from iroha2.data_model.account import *

public_key =# Get this from white_rabbit.
bunny = Account("white_rabbit@looking_glass", signatories=[public_key])
register = Register(Expression(Value(Identifiable(bunny))))

Second, you should provide the account with a public key. It is tempting to generate both the public and the private key at this time, but it isn't the brightest idea. Remember that the white_rabbit trusts you, alice@wonderland, to create an account for them in the domain looking_glass, but doesn't want you to have access to that account after creation.

If you gave white_rabbit a key that you generated yourself, how would they know if you don't have a copy of their private key? Instead, the best way is to ask white_rabbit to generate a new key-pair, and then give you the public half of it.

After putting all of this together, we submit it as before:

hash = cl.submit_isi(register)

5. Registering and minting assets

Iroha has been built with few underlying assumptions about what the assets need to be in terms of their value type and characteristics (fungible or non-fungible, mintable or non-mintable).

Asset creation is by far the most cumbersome:

import iroha2.data_model.asset as asset
from iroha2.sys.iroha_data_model import Value

time = asset.Definition(
    id=asset.DefinitionId(name="time", domain_name="looking_glass"),
    metadata={"a": Value.U32(10)},

Note the following; First, we used the **kwargs syntax to make everything more explicit.

We have a value_type which must be specified. Python is duck-typed, while Rust isn't. Make sure that you track the types diligently, and make use of mypy annotations.

The Quantity value type is an internal 32-bit unsigned integer. Your other options are BigQuantity, which is a 128-bit unsigned integer, and Fixed. All of these are unsigned. Any checked operation with a negative Fixed value (one that you got by converting a negative floating-point number), will result in an error.

Continuing the theme of explicit typing, the asset.DefinitionId is its own type. We could have also written asset.DefinitionId.parse("time#looking_glass"), but making sure that you know what's going on is more useful in this case.

Finally, we have mintable. By default this is set to True, however, setting it to False means that any attempt to mint more of time#looking_glass is doomed to fail. Unfortunately, since we didn't add any time at genesis, the white_rabbit will never have time. There just isn't any in his domain, and more can't be minted.

OK. So how about a mint demonstration? Fortunately, alice@wonderland has an asset called roses#wonderland, which can be minted. For that we need to do something much simpler.

amount = Expression(Value(U32(42)))
destination = Expression(Value(Identifiable(asset.DefinitionId.parse("rose#wonderland"))))
mint_amount = Mint(amount, destination)

This would add 42 to the current tally of roses that Alice has.

6. Visualizing outputs

The paradigm that Iroha chose to allow monitoring some events is the filter-map paradigm. Let's look at what we need to do in order to know e.g. what happened to a submitted instruction.

First, we'll need to remember the hash of the transaction that we want to track, next we create a filter:

filter = EventFilter.Pipeline(

And add a listener on that filter. Don't worry, the Rust side of the process is asynchronous, so barring issues with the GIL, you won't lock up your interpreter.

Note the types. The EventFilter is a type that filters out anything that isn't an event (and non-event types are beyond the scope of this tutorial). The pipeline module helps us by providing a concrete type of EventFilter, namely one that listens for transactions. Note that we haven't used the hash here.

Finally, we add a listening filter to the client:

listener = cl.listen(filter)

Now we must actually listen for events:

for event in listener:

    if event["Pipeline"]["status"] == "Committed" \
        and event["Pipeline"]["hash"] == hash:

And now, we have an infinite loop that will not quit until the event gets committed.


Nobody should do this in production code, and instead monitor the event queue for (at least) the possibility that the transaction gets Rejected.