Curate Skyrim Oghma factory catalog

[RANGROO]

Summary

• curate Skyrim's bundled Oghma catalog using the conservative editorial workflow established for ALMSIVI
• reduce 1,849 factory rows to 1,562 canonical articles by retiring or merging deterministic duplicates, repetitive variants, and walkthrough-only entries while preserving safe spoken names as aliases
• consolidate 237 generic equipment variants into 29 material/faction families and 49 generic bottled potion/poison effects into 11 gameplay-effect families
• preserve all 169 ingredient articles that explicitly list Alchemy effects: and retain 23 distinctive named concoctions as individual articles
• add 86 durable lore and consolidation articles, repair malformed metadata/prose, and expand conservative spoken-name coverage to 819 aliases across 314 articles
• add 13,203 reviewed related-concept tag assignments across 4,612 case-insensitive values, with 6-12 tags per article (8.45 average)
• make Oghma UI free-text search cover topics, aliases, and tags and explain the tag field in edit/create forms
• keep tags out of the topic/alias identity lexicon and legacy full-text vector; draft PR Ground Oghma retrieval in catalog entities #695 consumes them through a guarded contextual fallback

Tag generation and review

Tags were generated outside this repository with a resumable, checkpointed GLM 5.1 workflow. Each request included the canonical article, aliases, category, both knowledge tiers, detected related topics, and exact ALMSIVI overlap tags. Responses had to pass a strict schema and deterministic validation before an atomic checkpoint was accepted.

The completed run covered all 1,562 articles, cost $2.62150021, and produced a review manifest, HTML/CSV review views, frequency report, and independent audit. Export required the complete catalog hash and an explicit approval token; provider output never edited the repository or database directly.

The independent audit reported zero hard errors. Its 1,402 soft review flags are retained outside the repository: 1,322 semantic relationships without verbatim token overlap, 79 gameplay concepts needing semantic review, and one high-frequency alchemy concept. Category-spanning samples and flagged gameplay rows were reviewed; unsupported examples found during review were corrected before export.

Validation

• deterministic catalog rebuild: identical SQL, alias CSV, and documentation SHA-256 hashes
• temporary PostgreSQL import: 1,562 rows, 1,562 unique topics, all 1,562 tagged, zero empty categories/descriptions, zero ambiguous topic/alias phrases
• focused candidate retrieval: 41/41; ingredient-effect preservation passed; p50 10.03 ms, p95 13.96 ms, p99 14.96 ms
• tag-policy A/B on identical 500 fixtures:
  • previous product library: 260/500 (all 240 tag-positive cases absent), p95 16.91 ms
  • tag-aware product library: 500/500, p50 10.26 ms, p95 16.95 ms, p99 19.07 ms
  • covers unique multiword tags, corroborated shared tags, ambiguity, one-word tags, speaker labels, corruptions, and canonical precedence
• unchanged conversational/STT holdout: 500/500; 100% recall, precision, and primary ordering; zero forbidden/extra articles; p95 26.14 ms end-to-end
• OghmaAliasesTest.php: 4 tests, 9 assertions
• PHP syntax checks and git diff --check
• deployed local Oghma UI: tag-only metal core search returned adamantine_tower; edit helper/value rendered; narrow layout had no horizontal overflow; no browser warnings/errors

Data lifecycle

Existing databases and custom Oghma entries are not automatically truncated, migrated, or overwritten. The curated seed applies to new installs or an administrator's explicit Oghma Factory Reset. The local test runtime was explicitly reset after a recoverable table/file backup; it now has 1,562 tagged articles, 314 aliased rows, 13,203 tag assignments, and 4,612 unique tag values.

Limits

• no Skyrim/in-game testing
• no microphone, upstream STT, or live MiniMe calls
• local harness, GLM scripts, provider checkpoints, review artifacts, credentials, and API responses remain outside this repository

Server parity

No StobeServer or DialecticServer port is proposed: this seed is Skyrim-specific lore and those products maintain different game knowledge.

[RANGROO]

Superseded by consolidated draft #701, which contains this work plus the completed Oghma catalog, retrieval, knowledge-class, biography, parity, and UI changes. Please review the combined draft there.