Hunter NPC

Creates a simple hunter NPC that follows the player and in a 2D environment. The hunter will pick the next step that minimizes the distance to the player.

Quickstart

• Dependency management is handled by pdm, install pdm

curl -sSL https://pdm-project.org/install-pdm.py | python3 -

• Install dependencies. Note that this will create a virtual environment for you

pdm install

• Activate python environment

# use pdm venve activate to get the command to activate virtual env
source .venv/bin/activate

• Open up jupyter, and navigate to the url provided if it doesn't already open up the browser session

jupyter-lab

# then go to the link provided via the CLI

You can then run the notebook in notebooks/npc.ipynb the outputs of this will be

• prover and verifier key pairs for the zkSNARK of the hunter NPC
• EVM contract code for the hunter NPC verifier

You can find sample contracts (and corresponding abi) in the outputs directory.

Lifecycle of a proof

1. call ezkl.gen_witness when the state on chain updates (eg. the player moves and now you need to update the hunter) to create a new proposed_output.
2. call ezkl.prove to generate a new proof for that data.
3. call some (not inplemented here) wrapper contract with an updateState(proof, proposed_move) method which governs new moves for the NPC on-chain.
4. this function will make a subcall to the ezkl verifier's verifyProof method where calldata proof is the proof it has received and instances is the concatenation of current_state and proposed_move into a flat array (instances = [player_x, player_y, hunter_x, hunter_y, proposed_move]). The verifier will return a boolean value indicating whether the proof is valid or not.

This lifecycle can easily be automated with a bot that listens to the chain and updates the hunter NPC state accordingly. We can do this using the web3 library in python, the ezkl lilith cluster, or directly in the player's browser using the ezkl JS bindings (i.e the player when they move have to also compute a proof for the hunter's move -- they sort of run their own enemy as well).

TODOs:

• build a sample bot using web3
• build a sample bot using the ezkl lilith cluster
• build a flow for the player to also compute a proof for the hunter's move in the browser