Graph exploration

[apognu]

This adds the ability to explore the network of connections around a party, account, or transaction — across all of an organization's ingested data — without manually cross-referencing tables by hand.

• Same-field relations (new configuration). Organizations can now declare that two fields on two different tables represent the same real-world value, even when there's no formal link between them — for example, two accounts sharing an IBAN, or two logins sharing an IP address. This is pure configuration; it doesn't touch or move any data, it just tells the system which shared values are meaningful.

• Adjacency index (background worker). A periodic per-organization job rebuilds an internal index of which records share which values, combining the data model's existing relationships with the newly configured same-field relations. The index is rebuilt in the background and swapped in atomically, so live queries are never blocked or slowed down during a rebuild.

• Graph walk. Starting from a chosen record, the system explores outward for a configurable number of degrees (2 by default, up to 5, a degree usually covers multiple hops), following both existing relationships and same-field matches in either direction. Instead of dumping every intermediate record it passes through, it collapses the result down to the meaningful end points (parties, by default) and explains why they're connected — e.g. "these two parties share the same IBAN" or "this party's account sent money to that one." If a shared value turns out to be extremely common (e.g. a generic identifier shared by thousands of records), it's reported as a capped, approximate count instead of being expanded in full, to keep results fast and readable.