Skip to content

STRATEGY.md

Latest commit

 

History

History
67 lines (39 loc) · 4.3 KB

STRATEGY.md

File metadata and controls

67 lines (39 loc) · 4.3 KB
name PyKotor
last_updated 2026-05-23

PyKotor Strategy

Target problem

KotOR modding still runs on fragmented, Windows-centric legacy tools and ad-hoc format parsers. The crux is that mod authors cannot reliably move from edit → diff → patch → install on modern platforms without engine-fidelity gaps breaking real mods.

Our approach

Build one typed, cross-platform Python toolchain anchored on reverse-engineered K1+TSL fidelity — shared library, CLI, Holocron Toolset, HoloPatcher, KotorDiff, and KotorMCP — so every format and install path is game-verified before it ships.

Who it's for

Primary: Mod author — they're hiring PyKotor to author, package, and install KotOR I/II mods without legacy Windows-only tools.

Secondary: Developer / integrator — they're hiring PyKotor to automate mod pipelines, CI, and AI-assisted modding via library and MCP APIs.

Key metrics

  • TSLPatcher install parity — mods install identically to classic TSLPatcher; tracked via Libraries/PyKotor/tests/tslpatcher/parity/ (manifest + parametrized pytest), docs/solutions/testing/tslpatcher-parity-harness-mvp.md (prefer/defer/avoid), HoloPatcher regression tests when submodules are initialized, and open parity issues (#83, #67, #59)
  • Format round-trip fidelity — read→write→read produces game-compatible output; tracked via scoped pytest in Libraries/PyKotor/tests
  • Cross-platform install success — HoloPatcher runs on Windows, macOS, Linux, and Steam Deck; tracked via CI matrix and platform issue resolution (#54, #77)
  • Module Designer performance — ≥120 FPS (≤8.33 ms frame budget); tracked via GL performance benchmarks and docs/plans/2026-03-12-feat-module-designer-performance-bottleneck-plan.md
  • Module Designer self-contained workflow — level design without external tools; tracked via docs/LEVEL_EDITOR_CHECKLIST.md success criteria
  • Test signal quality — high-value regression coverage without ~2600-test noise; tracked via docs/plans/pykotor-test-suite-consolidation-plan.md and CI green on master via .github/workflows/python-package.yml (Python 3.9–3.11 primary gate; Python 3.8 minimum supported per AGENTS.md)

Tracks

Engine fidelity & format library

Reverse-engineered, unified K1+TSL format I/O with strict typing — the moat every tool surface depends on.

Why it serves the approach: Without game-verified formats, the rest of the toolchain cannot earn mod author trust.

End-to-end mod workflow

Holocron authoring → KotorDiff patch data → HoloPatcher install, with TSLPatcher compatibility as the acceptance bar.

Why it serves the approach: The product wins when a mod author completes the full loop on Linux without legacy tools.

Module Designer convergence

Current: Indoor Map Builder (Layout workflow) runs inside Module Designer. Target: one editor surface with contextual modes (Layout, walkmesh, module editing) per docs/LEVEL_EDITOR_CHECKLIST.md; close gaps (transform gizmos, walkmesh editing, inspector).

Why it serves the approach: Duplicate editor surfaces drain maintenance and block the north-star level-design UX.

Agent-native tooling

KotorMCP (emerging; submodule/PyPI packaging in progress) plus CLI action parity so every Toolset workflow is reachable by agents and automation.

Why it serves the approach: Differentiates PyKotor from xoreos, reone, and Kotor.NET in the emerging AI-assisted modding lane.

Not working on

  • KotOR engine reimplementation (reone / KotOR.js territory)
  • Multi-mod load-order orchestration (KotorModSync's lane)
  • Panda3D game engine as a shipping product surface (experimental research only)

Marketing

One-liner: A modern, cross-platform Python ecosystem for reading, modifying, and installing Knights of the Old Republic mods — with game-engine fidelity built in.

Key message: PyKotor replaces the legacy Windows toolchain with one typed library and integrated tools (Holocron, HoloPatcher, KotorDiff, CLI, MCP). Mod authors get authoring, packaging, and installation on any platform; developers get automation APIs grounded in reverse-engineered K1+TSL behavior.