Skip to content

solsim


Discord Chat Downloads

Tests Lint

Introduction

solsim is the Solana complex systems simulator. It simulates behavior of dynamical systemsβ€”DeFi protocols, DAO governance, cryptocurrencies, and moreβ€”built on the Solana blockchain.

Philosophy

Define your system how you see fit.

solsim will simulate its behavior and collect its results in a structured, straightforward manner.

Usage

  1. Implement initial_step and step methods.
  2. From each, return the current state, i.e. a dictionary mapping variables to current values.
  3. Specify the variables you'd like to "watch."
  4. Instantiate a Simulation, call .run().
  5. Receive a pandas DataFrame containing values of "watched" variables at each step in time.

With Solana

from anchorpy import Context
from solana.keypair import Keypair
from solsim.simulation import Simulation

class SomeSolanaSystem(BaseSolanaSystem):
    def __init__(self):
        super().__init__("path/to/workspace")
        self.account = Keypair()
        self.pubkey = self.account.public_key
        self.program = self.workspace["my_anchor_program"]  # solsim gives a Anchor program workspace (self.workspace).

    async def initial_step(self):
        self.program.rpc["initialize"]()  # Make RPC calls to your Anchor program.
        await self.client.request_airdrop(self.pubkey, 10)  # solsim gives you a Solana API client (self.client).
        return {"balance": await self.client.get_balance(self.pubkey)}

    async def step(self, state, history):
        self.program.rpc["submit_uniswap_trade"](
            ctx=Context(accounts={"account": self.pubkey}, signers=[self.account])
        )
        return {"balance": await self.client.get_balance(self.account)}


simulation = Simulation(system=SomeSolanaSystem(), watchlist=("balance"))
results = simulation.run(steps_per_run=5)  # Returns pandas DataFrame of results.

Without Solana

class SomeSystem(BaseSystem):
    def __init__(self, population):
        self.pop = population

    def initial_step(self):
        return {"population": self.pop}

    def step(self, state, history):
        return {"population": state["population"] * 1.1}


simulation = Simulation(system=SomeSystem(), watchlist=("population"))
results = simulation.run(steps_per_run=5)

CLI

Optionally run your simulations via CLI. Instead of calling simulation.run() in your code:

  1. Call simulation.cli().
  2. Run your simulation as e.g. python path/to/file.py run --num-runs 3.

Results Explorer

solsim gives you a streamlit app to explore results, e.g.

To automatically start this app following simulation, invoke one of the following:

  • simulation.run(visualize_results=True)
  • --viz-results flag in the CLI runner, e.g. python path/to/file.py run --viz-results

Installation

First, install Anchor.

Library

pip install solsim

Development

Install poetry. Then,

git clone --recurse-submodules https://github.com/cavaunpeu/solsim.git
cd solsim
poetry install
poetry shell

Detailed Usage

With Solana

First, write your Solana program. solsim prefers you do this in Anchor. Then,

  1. Write a system class that inherits from BaseSolanaSystem.
  2. Call super().__init__("path/to/program") in its __init__.
  3. Implement initial_step and step methods. (Since you'll interact with Solana asynchronously, these methods should be async.)

In 2., solsim exposes the following attributes to your system instance:

  • self.workspace: IDL clients for the Solana programs that comprise your system (via anchorpy).

For example, these clients let you interact with your respective programs' RPC endpoints.

  • self.client: a general Solana client (via solana-py).

This client lets you interact with Solana's RPC endpoints. Documentation here.

Finally,

  1. Define a watchlist: variables (returned in initial_step and step) you'd like to "watch."
  2. Instantiate and run your simulation, e.g. Simulation(MySystem(), watchlist).run(steps_per_run=10).

Without Solana

  1. Write a system class that inherits from BaseSystem.
  2. Implement initial_step and step methods.
  3. Define a watchlist.
  4. Instantiate and run your simulation.

Examples

Drunken Escrow

Agents are randomly paired to exchange random amounts of foo_coin and bar_coin via an Anchor escrow contract in each timestep.

  • Run: python -m examples.drunken_escrow.
  • Code: here.
  • Expected output (numbers may vary):
(.venv) ➜  solsim git:(main) python -m examples.drunken_escrow
πŸ‘† Starting Solana localnet cluster (~5s) ...
🟒 run: 0 | step: 100%|β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ| 4/4 [00:34<00:00,  8.60s/it]
πŸ‘‡ Terminating Solana localnet cluster ...
   step  run  mean_balance_spread  mean_swap_amount  num_swaps
0     0    0            17.333333         41.333333          3
1     1    0            22.000000          4.666667          3
2     2    0            40.000000         25.666667          3
3     3    0            44.000000         14.666667          3

Lotka-Volterra

The Lotka-Volterra model is a classic dynamical system in the field of ecology that tracks the evolution of interdependent predator and prey populations.

  • Run: python -m examples.lotka_volterra.
  • Code: here.
  • Expected output:
(.venv) ➜  solsim git:(main) python -m examples.lotka_volterra
🟒 run: 0 | step: 100%|β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ| 4/4 [00:00<00:00, 70492.50it/s]
   step  run  food_supply  population_size
0     0    0     1000.000            50.00
1     1    0      995.000            60.00
2     2    0      989.000            69.95
3     3    0      982.005            79.84

This example is inspired by cadCAD Edu.

Inspiration

solsim humbly builds on the shoulders of the giants that are cadCAD and tokenspice, among others.