Praxish

Steps toward an open-source reimplementation of the Praxis logic language, which was used for content authoring in the ambitious interactive drama framework Versu.

To use Praxish in your own project, you'll first want to import the core library files in the following order:

• util.js – General-purpose utility functions employed by the other files
• db.js – Exclusion logic database
• praxish.js – Functions for defining social practices, querying for possible actions, etc.

We also provide a couple of (optional, experimental) core modules that implement different decision-making strategies for autonomous characters:

• planner.js – Forward-looking agents that use planning for goal-directed action selection
• swaygent.js – "Swayable agents" that use Ensemble-like volition and influence rules to choose desirable actions opportunistically

Then you can look at the various demos for examples of how to integrate Praxish functionality into a game loop:

• demos/test – Basic functionality tests, which demonstrate the gradual build-up of a social world
• demos/pwim – A slightly more complicated example game that allows the player to pick actions for their character to perform, while other characters act autonomously
• demos/sway – An interactive example project that demonstrates Swaygent decision-making logic
• Bauprobe – Praxish-driven agents in continuous time and (user-editable) space

The original Praxish paper, published at AIIDE 2023, is probably the best starting point if you want to read more:

@inproceedings{dameris2023praxish,
  title={Praxish: a rational reconstruction of a logic-based DSL for modeling social practices},
  author={Dameris, James and Roman, Rosaura Hernandez and Kreminski, Max},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment},
  volume={19},
  pages={407--415},
  year={2023}
}

We archived the version of Praxish described in this paper as a GitHub release, then continued developing new features atop the basic Praxish substrate.

Our next major Praxish-based project (intended to demonstrate an interaction technique called "play what I mean", or "PWIM" for short) introduced several improvements to the Praxish library:

• An interactive demo project that allows you to explore a small storyworld simulated by Praxish
• A more modularized codebase, with key Praxish functions mostly attached to a top-level Praxish module object rather than dumped unceremoniously into the global JavaScript namespace
• A planner module that allows non-player characters to act with greater foresight

For a bit more on PWIM, see our FDG 2024 demo paper; the demos/pwim subfolder of this repo; or the separate repo that we used for PWIM development:

@article{kreminski2024cheap,
  title={Cheap and Easy Open-Ended Text Input for Interactive Emergent Narrative},
  author={Kreminski, Max},
  journal={arXiv preprint arXiv:2406.00942},
  year={2024}
}

After the PWIM release, Praxish development paused for a couple of years, then resumed in 2026. So far, this latest wave of development has introduced:

• The Swaygent module, which allows practices to be augmented with Ensemble-like influence rules and volition rules that drive a smarter form of reactive utility-based decision-making for NPCs
• Basic arithmetic in queries, via numeric comparators (gt, gte, lt, lte) and the calc operator
• Experimental support for subqueries, which enable both aggregates (e.g., count) and compound negation (similar to not in Datomic)
• A new debugging affordance (impossibleActions) that makes it easier to determine what Praxish actions are currently impossible for a particular actor and why
• Another new debugging affordance (DB.diff) that makes it easier to figure out what changed between two different versions of an exclusion logic database
• An interactive demo project that shows off Swaygents, weighted random NPC action selection, new debugging features, and allows player control of any character to be toggled on and off
• Bauprobe, a Praxish-based social simulation environment that demonstrates agents wandering around user-created settings in continuous time and space

Praxish development continues today! Stay tuned for more...