dnd5e-engine 0.1.0

Pure-Python D&D 5e SRD rules engine — combat, checks, effects. Host-agnostic, zero I/O.

dnd5e-engine

Pure-Python D&D 5e SRD rules engine — host-agnostic, zero I/O. Combat, checks, effects, on a zone graph or a 2-D grid.

Status

Working engine. It resolves combat against the typed 2024-SRD corpus shipped by dnd5e-srd-data (the BundledAssetLoader reads the bundled canonical/ data — no network, no DB). The engine is edition-agnostic: it consumes whatever typed content it is handed.

Two spatial backends are supported, selected at start_combat:

• Zone graph — pass scene_zones=SceneTopology(zones=..., edges=...); range/reach are resolved as shortest-path distance over a weighted, undirected zone graph.
• 2-D grid — pass grid_scene=GridScene(width, height); positions are "col,row" cell ids (cell_id(col, row)), distance is Chebyshev (8-direction, one cell = cell_size_ft).

Install

Dev (editable, from this directory):

cd packages/dnd5e-engine
uv venv && uv pip install -e '.[dev]'
uv run --extra dev pytest -q
uv run --extra dev ruff check src/ tests/ scripts/
uv run --extra dev mypy src/

Standalone wheel:

uv build          # builds dist/dnd5e_engine-*.whl (+ sdist)

A clean-room install smoke builds both wheels, installs them into a throwaway venv with no editable path deps, and runs a real grid combat through the public API:

bash scripts/smoke_clean_install.sh   # ends with "==> SMOKE PASSED"

Quickstart

A minimal grid combat: open it, move a PC one cell, then close it.

import asyncio

from dnd5e_engine import (
    EncounterMemberSpec,
    GridScene,
    PartyMemberSpec,
    PlayerIntent,
    cell_id,
    end_combat,
    start_combat,
    submit_player_intent,
)


async def main() -> None:
    start = await start_combat(
        session_id="demo",
        party=[
            PartyMemberSpec(
                entity_id="char:hero",
                name="Hero",
                initiative=20,
                hp_current=12,
                hp_max=12,
                ac=12,
                zone_id=cell_id(0, 0),
            )
        ],
        encounter=[
            EncounterMemberSpec(
                entity_id="mon:foe",
                entity_type="Monster",
                name="Foe",
                initiative=1,
                hp_current=7,
                hp_max=7,
                zone_id=cell_id(5, 0),
            )
        ],
        grid_scene=GridScene(width=10, height=10),
        rng_seed=1,
    )

    # The hero won initiative; move one cell diagonally.
    await submit_player_intent(
        start.handle,
        actor_id="char:hero",
        intent=PlayerIntent(intent_type="move", target_zone_id=cell_id(1, 1)),
    )

    result = await end_combat(start.handle)
    print("ended reason:", result.outcome.ended_reason)


asyncio.run(main())

An attack instead of a move: submit PlayerIntent(intent_type="attack", target_id="mon:foe", weapon_id="longsword") — the engine fetches the typed weapon from the bundled corpus and walks its activities.

Public API

All exported names live in __all__ in src/dnd5e_engine/__init__.py. The key entry points:

• start_combat(*, session_id, party, encounter, scene_zones=|grid_scene=, rng_seed, ...) — open a combat, materialize runtime state, return a StartCombatResult (.handle, opening .events).
• submit_player_intent(handle, actor_id, intent) — validate and resolve a PC's PlayerIntent for the current turn.
• advance_monster_turn(handle) — resolve the active monster's turn via its typed action repertoire + behavior gambits.
• end_combat(handle) — close a combat and return an EndCombatResult (.outcome, .events, .final_active_effects).
• narration_events(handle) — async iterator streaming the CombatEvent union for the narrator.
• get_actor_active_effects(handle, entity_id) — read-only snapshot of one combatant's active effects.
• resolve_check(spec) — resolve an out-of-combat ability check / saving throw (CheckSpec → CheckResult).
• build_party_member(spec, ...) — project a CharacterBuildSpec into a combat-ready PartyMemberSpec.
• make_build_spec(...) — assemble a CharacterBuildSpec (ability scores, class, species).

Spatial helpers and spec types:

• GridScene / cell_id(col, row) / parse_cell(cell_id) — 2-D grid scene + cell-id codec.
• SceneTopology / ZoneEdge — zone-graph scene description.
• PartyMemberSpec / EncounterMemberSpec — combatant inputs to start_combat.
• PlayerIntent — a PC's submitted intent (move / attack / cast_spell / use_item / ...).
• CombatHandle, StartCombatResult, EndCombatResult, CombatOutcome, CombatEvent, ActiveEffect, CheckSpec/CheckResult/CheckKind, CharacterBuildSpec, AbilityScores, and the IntentType / ActionType literals.

Layout

packages/dnd5e-engine/
├── pyproject.toml          hatchling build; pydantic + d20 + dnd5e-srd-data deps
├── LICENSE                 MIT (engine code)
├── scripts/
│   ├── _smoke_grid_combat.py     clean-room smoke program (runs in a fresh venv)
│   └── smoke_clean_install.sh    builds wheels + installs them + runs the smoke
├── src/dnd5e_engine/
│   ├── __init__.py         public API (__all__)
│   ├── orchestrator.py     start/submit/advance/end combat seam + live state
│   ├── spatial.py          grid + zone topologies, cell_id / parse_cell
│   ├── specs.py            GridScene, SceneTopology, party/encounter specs
│   ├── check.py            out-of-combat ability check / saving throw resolver
│   ├── build_party.py / build_spec.py   character build → combat spec projection
│   ├── lib_loader.py       BundledAssetLoader singleton (typed SRD corpus)
│   ├── events.py / outcome.py / results.py   event union, outcome, result envelopes
│   ├── activities/         typed-Activity resolvers (attack / save / monster actions)
│   ├── rules/              dice, conditions, gambits, ...
│   └── types/              combat / effects / conditions / intent value types
└── tests/                  pytest suite

License

• Engine code: MIT — see LICENSE. The engine ships no SRD data.
• SRD content (the typed corpus consumed via dnd5e-srd-data): CC-BY-4.0 — see the dnd5e-srd-data package's LICENSE and NOTICE for the dataset license and attribution chain.