irminsul-corpus

Autonomous knowledge base pipeline for Irminsul — a production RAG assistant for Genshin Impact.

Deterministic corpus builder that pulls peer-reviewed theorycrafting (KQM TCL), exact game stats (genshin-db API), and canonical lore (Fandom Wiki). Runs weekly via GitHub Actions. Each vector carries 7 metadata fields for smart filtering by tier, character, content type, and trust score.

---

Corpus

Tier	Source	Files	Trust
TCL	KQM Theorycrafting Library	305	Ground truth
Structured	genshin-db API	405	Ground truth
Wiki	Genshin Impact Fandom Wiki	130	Canonical

Total: 840 files · 6,876 Pinecone vectors · 7 metadata fields per chunk

---

Pipeline

discover → tcl → structured → wiki → ingest

Runs every Sunday 2am UTC via GitHub Actions. Auto-discovers new characters when patches drop. Incremental updates only.

---

Architecture

• Smart name resolver with persistent cache — handles full wiki names automatically
• Incremental timestamp-based updates — only re-fetches pages changed since last run
• Self-healing structured data — validate_file() detects broken API fields and auto-re-fetches

---

Setup

git clone https://github.com/MukulRay1603/irminsul-corpus
cd irminsul-corpus
pip install -r requirements.txt
cp .env.example .env

Required secrets (GitHub Actions):

• PINECONE_API_KEY
• PINECONE_INDEX

Optional (local only):

• GEMINI_API_KEY (legacy, not used in pipeline)

Run:

python run_pipeline.py

---

Attribution & Legal

• KQM Theorycrafting Library — © KQM contributors, CC BY-NC-SA 4.0
  https://github.com/KQM-git/TCL

• genshin-db — open source game data by theBowja
  https://github.com/theBowja/genshin-db

• Genshin Impact Fandom Wiki — community wiki data
  https://genshin-impact.fandom.com

• Genshin Impact is the intellectual property of HoYoverse (miHoYo Co., Ltd.)
  This project is not affiliated with or endorsed by HoYoverse.
  All game content, characters, and lore belong to HoYoverse.