Pipelines

Four pipelines, one user-facing entry point each. See AGENTS.md in bgg-research for the canonical reference.

Sync is local-only.
Run npm run dev on your machine and click the button there. It reads data/bgg.sqlite and pushes to GitHub, which auto-deploys here.

1 · BGG scraping

scripts/refresh_bgg.py

BGG XML API → SQLite. Rate-limited at 2 req/sec. Slow + ban-sensitive.

Games scraped
10,035
10,029 ranked
Last scrape
1d ago
2026-04-25
view details →

2 · LLM enrichment

scripts/enrich_llm.py

Two depth tiers — wide (terse, ~5 fields) and deep (paragraphs + theme axis).

Wide (v2)
522
1d ago
Deep (v3)
40
13h ago
view prompts + samples →

3 · Analysis

scripts/analyze.py

Two tracks: BGG-mechanism regression (track A) + LLM-primitive ranking (track B).

Outputs: 7 CSVs + briefs markdown + multi-sheet analysis.xlsx.

view details →

4 · Candidate deep dives

scripts/deep_dive.py

7-section design document per top candidate. Heat-style.

Currently 8 published.

view details →

Pipeline run log

PipelineNotesRowsStarted
syncsync-db.ts run10,0354h ago