Engineering documentation

This area is for the team that builds Trovensa. It holds what the code must agree on and why it was decided that way; the platform area holds how the product is meant to be used.

• Specifications: the contract summary condenses PRODUCT-CONTRACT.md including sections 10 and 11 (taxonomy and pilot positioning, 87 permissions and personas at version 3, 189 designed routes with the thread and bound-user rules, the 60-table data model and audit rule, the 61-value tone map, providers, worker, the public site routes, the R3 twin names, ownership) and the spec schema is the five-part skeleton a specification version follows, one block per stage id.
• Design: Tokens lists every light, dark and product token with its hex, Shell describes the rail, the panel, the topbar, the breakpoints and the two cookies, Console describes the three regions, the approval card, streaming and the gates that check them, and Screenshots embeds the reference frames captured from the web app.
• Decisions: the decision log lists the thirty-four ADRs: twelve of 2026-09-13, eight phase-2 records of 2026-09-14, six phase-3 records of 2026-09-15 and eight phase-4 records of 2026-09-16 (workbench roles, the evidence panel, skills, harnesses, memory tiers, inline interaction), each with the five fixed headings.
• Roadmap: what exists today, what R3 built and the gates it must pass, and what R4 (the workbench), R5 and R6 hold, without dates.
• Runbooks: First release and the deployment record cover the first release; Role split, Partner onboarding, Tenant offboarding and Backup and restore cover phase 2.

Every reference and specification page ends with a Source: line naming the repository files that were read while writing it; counts are recounted on publish day. Vendor names appear in bodies only, never in a heading; agent output is “agent-generated” or “proposed”; the pilot is described as “non-GxP decision support with a light validation package” and no page makes a compliance claim.