How Ithildin Works
Ithildin is an investigative tooling platform built around one constraint: a reviewer should be able to inspect what a claim depends on. The system ingests records, tracks findings, preserves evidence provenance, and publishes reader-facing pages that keep public artifacts distinct from metadata-only source records.
High-level flow from source intake to portfolio-grade static output.
flowchart LR A["Providers
Registries, released docs, public records"] --> B["Trackers
Leads, findings, infra tasks"] B --> C["Queue + Workers
Enrichment, curation, exports"] C --> D["Evidence Layer
Citation resolution + source records"] D --> E["Publish Layer
Dossiers, articles, network, financial views"] E --> F["Quality Gates
Lint, snapshots, support coverage, build"]
- Dossiers combine findings, timelines, and curation prose with evidence mode.
- Articles turn finding clusters into reviewable case studies.
- Network and financial pages expose the underlying structure rather than only the prose output.
- Shared path/config handling for local, CI, and demo-mode builds.
- A canonical CLI for trackers, providers, queue actions, and builds.
- Static artifacts backed by lint, snapshot, and support-coverage checks.
Claims are expected to resolve either to a public artifact, a hosted copy, or a metadata-rich source record. Metadata-only records must still expose provenance, locator context, access notes, and every public occurrence where the source is used.
The public demo fixture is intentionally small: two dossiers, one article, a network graph, a financial view, and mixed citation types. The point is reproducibility and legibility, not hiding the engineering inside a giant private corpus.