LightHouse Context Graph Service
One memory, shared by every part of the platform. Foundry writes to it while software is being built, Enterprise Intelligence writes to it while the business runs, and either can read what the other decided.
What people override
It watches what people override. When the same correction is made again and again, that pattern is put to a person as a proposed rule change — not applied on its own, and not lost either.
Decisions that lead to other decisions
A decision is rarely alone. One implements another, one caused a third, one was contradicted later. Those links are kept as the decisions are, so a question about a system returns the chain rather than a single note — and you can see what a change would disturb before you make it.
Not every question deserves the same search
A query about a settled matter goes first to the curated record — the architecture decisions, the playbooks, the procedures — and comes back with the authoritative answer rather than the popular one. Only an open-ended question goes wide, across the whole decision history, following the links between decisions to what caused what.
A record that survives an audit
The record is written once and not edited afterwards, personal data is stripped as it is captured, and each client’s history is separated from every other. When an auditor asks what was decided about a given subject in a given quarter, that is a query, not an archaeology project.
Asked from where the work happens
The history answers over an open protocol, so a question can be asked from the tools your engineers already use rather than only from a console of ours. That avoids the reason most knowledge systems go quiet: one more place to remember to look.
Tell us about a decision nobody can explain any more. A person reads every message and replies.