Extending SeedBank#

SeedBank can be extended to support new random number generators. Right now this only be done by adding to the SeedBank code; we intend to expose an API to register additional RNGs in the future (see issue 2).

RNG Interface#

The RNG seeding interface consists of two functions:

is_available() bool#

This function should return a boolean indicating whether or not the RNG is available for seeding.

seed(state: seedbank.SeedState)#

This function takes an RNG seed state (implemented by SeedState) and seeds the specified RNG. It is allowed to fail if is_avialable() returns False.

RNG Modules#

RNG seeding support is implemented by Python modules implementing the RNG interface. There are several of these in the seedbank package.

The seedbank.SEED_INITIALIZERS variable contains a list of module names that will be initialized by seedbank.initialize().


This attribute stores a list of strings naming modules that implement the RNG interface and will be seeded when SeedBank’s global seed is set with initialize().

The initialization logic works as follows:

for seed_mod in SEED_INITIALIZERS:
    if isinstance(mod, str):
        mod = import_module(mod)
    if mod.is_available():

Seed State#

Internally, SeedBank uses the SeedState class to track seeds and obtain both NumPy entropy and integer seeds for other RNGs.

class seedbank.SeedState(seed=None)#

Manage a root seed and facilities to derive seeds.


seed (Optional[np.random.SeedSequence]) – The initial seed for this seed state. If None, a default SeedSequence is initialized.

property seed: SeedSequence#

Get the seed sequence for this seed state.

property int_seed#

Get this seed as an integer.


Get n words of entropy as a NumPy array.


words (int) – the number of words to return.


the entropy.

Return type:

ndarray[int, dtype[uint32 | uint64]]

derive(base, keys=None)#

Derive a new seed state.


The derived seed state.

Return type: