KOTOR Modding Knowledge Base

This repository is a source corpus and a knowledge-base rebuild in progress. It is not yet a verified modding wiki.

The previous release described AI-synthesized documents as curated facts. An audit found that its Discord parser merged follow-up messages into author fields, omitted messages and media, lost exact message IDs, and fed that damaged corpus into later AI summaries. The resulting articles also mixed original KOTOR behavior with reone, KotOR.js, xoreos, tools, ports, and sometimes the wrong game.

Those articles are now clearly quarantined in docs/legacy/. Start with the documentation status and scope and evidence model.

Current trustworthy baseline

• 12,132 unique Discord messages from six OpenKotOR channels
• preserved message IDs, author IDs/names, timestamps, reply relationships, and attachment metadata
• 1,034 attachment/media references, including 2 voice messages and 125 video references
• readable Markdown mirrors plus normalized JSONL records
• a validator that compares each normalized record with a fresh HTML parse and checks attachment, analysis, and chunk coverage

The two voice messages are preserved in their exact conversation context but are marked as not transcribed. One unrelated 57 MB third-party source map is intentionally excluded from redistribution while its attachment metadata remains recorded.

Trust boundaries

Path	Status
sources/discord/raw-export/	Primary HTML source snapshot and its same-directory attachment namespace
extracted_channels/*.jsonl	Reproducible normalized source and preferred citation target
extracted_channels/*.md	Human-readable mirror of the normalized source
analysis/	Reproducible discovery indexes, not factual authority
chunks/	Reproducible processing input, not factual authority
generated/legacy-ai-briefs/	Invalid legacy AI briefs generated from the lossy corpus; do not use
generated/ai-drafts/	Optional, ignored, explicitly unverified discovery output
docs/legacy/	Unverified synthesized articles retained only for audit and salvage
docs/legacy-audit.md	Audit evidence explaining quarantine; not technical authority
Other documents outside docs/legacy/	Process and scope documentation until verified articles are rebuilt

Why scope is mandatory

The following are separate subjects and must never be collapsed into “the engine”:

• the original KOTOR 1 Odyssey implementation;
• the original KOTOR 2 / TSL Odyssey implementation;
• platform- and release-specific ports/builds;
• reone;
• KotOR.js;
• xoreos;
• format libraries and editors such as PyKotor and Holocron Toolset;
• patching systems such as Kotor-Patch-Manager, KPatcher, and TSLPatcher.

Every verified claim will separately identify content game, runtime, tool/project, artifact domain, platform, build/release, revision, evidence, claim status, issue lifecycle where relevant, reviewer, and verification date.

Reproducible corpus pipeline

From the repository root:

python tools/extract_html.py
python tools/analyze_channels.py
python tools/chunk_channels.py
python tools/validate_corpus.py

The final command must pass before derived data is accepted. The old AI briefs are not regenerated or promoted automatically.

Repository state

KotorModdingKnowledgeBase/
├── docs/
│   ├── README.md                 documentation status
│   ├── verification-model.md    required scope and evidence fields
│   └── legacy/                  quarantined AI synthesis (21 articles)
├── extracted_channels/          normalized JSONL and readable Markdown
├── analysis/                    reproducible inventory output
├── chunks/                      reproducible conversation segments
├── sources/discord/
│   ├── raw-export/              preserved flat HTML/attachment snapshot; treat as immutable
│   └── attachment-exclusions.json
├── generated/
│   └── legacy-ai-briefs/        invalid legacy AI briefs
├── tools/                       extraction, indexing, and validation scripts

The repository root now contains only the six project entry files. The raw export directory intentionally remains flat because the HTML snapshots use same-directory relative links to roughly 1,300 files / 2 GiB of images, videos, models, archives, code, and miscellaneous uploads. The generated ASSET-INDEX.md is an inventory aid, not a statement that every item is relevant, safe to redistribute, or technically verified.

Attribution and licensing

The corpus originates primarily from OpenKotOR Discord exports and includes third-party projects, game-derived assets, community uploads, and standalone documents. See ATTRIBUTION.md for the current provenance record.

The repository license must not be read as relicensing Discord messages, third-party code/docs, game files, or uploaded assets. Those materials retain their respective rights and require an item-level provenance and licensing review before redistribution.

This is a fan-created project. Knights of the Old Republic is a trademark of LucasArts/Disney. The project is not affiliated with or endorsed by LucasArts, BioWare, Obsidian, Aspyr, or Disney.