Roadmap¶
Status: Living document. Version: 0.1 Date: 2026-08-11
Kovent is at design stage. This page states honestly what exists, so nobody reading the rest of this site mistakes a specification for an implementation.
What exists today¶
| Artefact | State |
|---|---|
| Design system (KADLS) | Built. Foundations, primitives, and domain signatures in both themes |
| Supervision prototype | Built. Returns triage register and full return detail screen |
| Architecture doctrine | Written. Ten laws with enforcement mechanisms |
| Engine map | Written. Ten engines traced to the requirement's modules |
| R01 and R02 specs | Written. Fields, rules, checklists, worked examples |
| This documentation site | Built |
| Backend runtime | Nothing. No Quarkus service exists yet |
Phase A — Prove the premise¶
The goal is to falsify the central claim as fast as possible: that a return can be defined as configuration and rendered, validated, and reviewed without code.
If this fails, everything downstream changes. So it goes first.
| # | Deliverable | Proves |
|---|---|---|
| A1 | Quarkus service skeleton, PostgreSQL, Redis, outbox, OpenTelemetry | The stack starts and traces |
| A2 | Return Studio template model, R02 v2.4 as seed configuration | A template is data |
| A3 | Strata submission with jsonb conformance check against the template version |
Immutability and template-as-schema |
| A4 | Assay — R02's twelve rules as Kogito DMN, worked examples as fixtures | Rules are configuration, not code |
| A5 | Seal hash-chained entries in the same transaction | Auditability by construction |
| A6 | Generic return renderer in web components, driven only by the template | The premise. No R02-specific UI code |
| A7 | R01 added as configuration alone, with no code change | The premise, independently confirmed |
A7 is the gate. If adding R01 requires a code change, the architecture is wrong and we stop and rethink rather than proceeding.
Phase B — Close the loop¶
A supervisor can receive, review, and decide a real return.
| # | Deliverable |
|---|---|
| B1 | Registry firm master, licence categories, user mandates, OIDC integration |
| B2 | Cadence obligation materialisation and due-date arithmetic with the holiday calendar |
| B3 | Bench triage, assignment, checklist, approval gate |
| B4 | Query loop with amendment lineage |
| B5 | Maker-checker enforced at the API, with the bypass test in the suite |
| B6 | Firm portal: draft, pre-submission validation, certify, file, receipt |
| B7 | Prism scoring with driver decomposition |
At the end of Phase B, R01 and R02 are live end to end for a pilot group of firms.
Phase C — Scale the catalogue¶
| # | Deliverable |
|---|---|
| C1 | Specs for R03–R10 through design workshops |
| C2 | Cross-return consistency — the deferred PENDING rule model |
| C3 | Event-driven obligation kind for R08, R15, R29 |
| C4 | Excel import mapped to template fields, rejecting rather than coercing |
| C5 | Projections and the operational dashboards |
| C6 | Attachment-heavy returns — R03 audited financial statements |
Phase D — Enforcement and analytics¶
| # | Deliverable |
|---|---|
| D1 | Docket case lifecycle and evidence bundles |
| D2 | Penalty recommendation computation |
| D3 | Directive tracking and compliance monitoring |
| D4 | Sector heatmaps, trend analytics, export packs |
| D5 | Chain-head publication for Seal |
| D6 | Multi-tenant hardening for a second supervisor deployment |
Load testing gates¶
Two numbers must be measured before we commit to the data design, because both are assumptions today:
| Question | Why it blocks |
|---|---|
jsonb query performance across a realistic catalogue at quarter-end volume |
Determines whether projections are an optimisation or a necessity |
| Concurrent submission throughput when every firm files on the deadline | Determines whether the write path holds at the only moment it matters |
Neither is a Phase A blocker, but both must be answered before Phase C adds twenty returns on top of an unproven shape.
Dependencies on the supervisor¶
Kovent cannot proceed past specification on its own. These are needed from SECZim:
- Confirmed legal names and instrument references for each return
- Minimum capital figures by licence category, and the instrument setting them
- The early warning floor, and whether the proposed 1.20 advisory band is adopted
- Due-date notice periods and the public holiday calendar
- Late-filing and penalty parameters
- Licence category taxonomy and the return applicability matrix
- A pilot group of firms for Phase B
The spec workshop inputs list what each return needs individually.