User-provided source archive
The certified current input is a LinkedIn data export supplied by the user. It is parsed to derive relationship facts.
The current architecture keeps IndexedDB as the local performance and cache layer while securely persisting the factual graph to the cloud. A fresh device can restore those facts and recompute the same intelligence without requiring the original LinkedIn ZIP.
RelationshipsIQ separates the source archive, the derived factual graph and the intelligence calculated from that graph. Each layer has a different job.
The certified current input is a LinkedIn data export supplied by the user. It is parsed to derive relationship facts.
IndexedDB remains the local performance layer for fast graph access, restore and offline behavior.
Cloud state stores the factual inputs required to rebuild the network under tenant isolation and row-level security.
Radar, Runway, Companies & Paths and the broader intelligence layer are recalculated from restored facts instead of trusting stale saved scores.
The relationship model uses communication metadata while avoiding retention of private message body content.
Your explicit corrections and relationship context remain first-class inputs rather than cosmetic labels.
A packaged local/desktop distribution remains a valid product surface, but the public website should not present a downloadable executable until there is a signed, updateable distribution to offer. The current usable product path is the web application architecture with local-first storage and cloud continuity.
That requirement can shape packaging without changing the core relationship-intelligence model.