Source first.
Events over claims.
ACG is built the way an analyst would build it — every edge traces to a reported event, every event to a URL, and every tool output declares what the graph does not know.
Every record has a source
Gold-standard events are hand-verified against primary reporting; each carries source_url plus a confidence grade (verified / confirmed / reported / rumored). If we can't source it, we don't assert it.
金标事件逐条人工核验并附来源 URL 与置信度分级;没有来源就不入库。
Event-sourced, not entity-sourced
The atomic unit is the event — a funding round, an acquisition, an IPO, a regulation — with a timestamp, not a static 'net worth' claim. Narratives are derived from event chains, so they can be replayed and audited.
原子单位是事件(融资/并购/上市/法规),带时间戳;叙事由事件链推导,可回放、可审计。
Stated gaps, not hidden ones
The graph currently has no ownership percentages, and SPV / offshore layers are unlinked. Tool outputs carry data_gaps fields saying exactly what is missing instead of implying completeness.
图谱暂无股比数据、离岸层未连通;工具输出以 data_gaps 字段明示缺失,不假装完整。
Two-tier policy layer
9 gold-standard policy events (OJK sandbox, UU PDP, MAS PSA…) sit alongside 396 LexNusa-fed regulations. Gold events are year-verified; feed events carry LexNusa status (berlaku / dicabut / diubah) at year precision.
政策层分两级:9 条金标(逐条核验)+ 396 条 LexNusa 供给(年份精度,带生效/废止/修订状态)。
Neo4j (events + entities)
└─ pipelines/export_graph.py --apply → web/public/data/graph.json
└─ astro build → static site (this page)
└─ Cloudflare Pages / GitHub Pages Query access: 11 MCP tools (TypeScript, stdio) — see the repository README.