ADR 0008: Gate GLPI-derived capabilities on measured user demand
- Status: accepted
- Date: 2026-08-21
Context
GLPI spans inventory/CMDB, assistance, management, tools, administration, and automation. K0smos already owns tickets, projects, clients, planning, deadlines, contracts, knowledge/documentation, notifications, import and audit concerns. Copying GLPI's feature or database model would duplicate these owners and create an unbounded product programme.
The capability matrix in doc/public/en/architecture/evaluations/glpi.md maps official GLPI
features to current k0smos ownership and identifies real gaps. The clearest gap
is generic asset lifecycle/CMDB data, but no user interviews, deployment target,
source-of-truth policy, or integration customer has been supplied.
Decision
Do not build a GLPI clone or an Asset module yet.
Run the documented user interview/ranking gate first. For every selected gap, choose one of native bounded implementation, GLPI integration/synchronization, or no action. Organizations already operating GLPI should default to integration assessment, not parallel authoritative records.
If demand confirms a first native increment, it may create one optional Asset module limited to typed classes/fields, location, ownership, lifecycle, relationships, attachments, provenance/history, and stable links to Client and Ticket. SLA/OLA, Problem, Change, reservations, discovery agents, and software licenses remain independent future decisions.
Consequences
The current increment delivers evidence and an interview instrument, not new schema, routes, ACL, menu entries, or runtime dependencies. This is a deliberate product gate: technical feasibility is insufficient authority to choose what real users need or whether GLPI remains the source of truth.
Re-evaluation input
The gate needs anonymized interview scorecards from representative requesters, technicians, service owners, asset/finance owners, and administrators, plus an inventory of existing GLPI instances/integrations and data ownership.