SpectreIntel is built to separate fact, inference, legal meaning and risk. It does not accept a case narrative as proof. It reconstructs the case through traceable sources, documents, digital traces and explicit uncertainty.
SpectreIntel works under a cold rule: source beats impression, document beats assumption, evidence beats narrative. Facts are separated from inference. Risk is separated from proof.
The system is not a demo shell and not a dashboard. It is a working case structure with discovery, next-step planning, case graph persistence and an operational UI shell.
Finds candidate sources, candidate documents, candidate entities and candidate contradictions, then validates only the top-ranked outputs selectively.
Fixes the current objective, next step, case state and blocking conditions in a deterministic rule-based way.
Stores useful nodes, edges, provenance and cross-artifact links for one case without inflating into a decorative graph.
Lets a real user inspect sources, entities, evidence, gaps, risk and the next valid step without reading raw repo files.