--- id: STATE-WP-0062 type: workplan title: "Two-dimension service catalog (hosting × development) + Services nav section" domain: custodian repo: state-hub status: finished owner: codex topic_slug: custodian created: "2026-06-19" updated: "2026-06-19" state_hub_workstream_id: "b5c9d93f-9f5e-4d10-bb3b-90322c7419b7" --- # STATE-WP-0062 — Two-dimension service catalog + Services nav section **Origin:** UI request to reorganise the single "Services (TPSC)" page into a **Services** section. During refinement the model was generalised: services vary along **two independent dimensions**, not one list of three, so the catalog and its persistence are restructured to reflect that. ## Concept Two orthogonal dimensions classify every service: - **Hosting** — `self_hosted` (coulomb operates the service) vs `cloud_hosted` (coulomb consumes someone else's running service). - **Development** — `first_party` (coulomb is development-responsible) vs `third_party` (coulomb is not development-responsible). Their product gives four classes: | | third-party | first-party | |---|---|---| | **cloud-hosted** | SaaS/APIs (today's TPSC) | our service deployed to a cloud | | **self-hosted** | OSS we run (Gitea, Postgres…) | our service on coulomb infra | **Design principle:** the "extra data" attaches to a **dimension, not a class**, so extension data composes (a self-hosted first-party service gets the self-hosted *and* the first-party extensions) rather than forcing four bespoke record shapes. ## Persistence model (decided) A single common **`service_catalog`** core table plus per-dimension extension tables (each 1:1, optional, keyed by `service_id`): - **`service_catalog`** (core, all services): `id, slug (unique), name, owner_or_provider, category, description, website_url, status (active|deprecated), hosting_type (self_hosted|cloud_hosted), development_type (first_party|third_party), maturity_level (1..3 | null — the Service DoM Level), created_at, updated_at`. - **`service_third_party`** (development_type = third_party): upstream package references, upstream service/support contacts, source URL, license. - **`service_first_party`** (development_type = first_party): FK to the internal dev repo (`managed_repos.id`), owning domain. - **`service_cloud`** (hosting_type = cloud_hosted): GDPR maturity, DPA available, ToS/privacy URLs, data-processing regions, retention notes — the existing TPSC data-processor block (relevant whenever data leaves coulomb infra, independent of who built the service). - **`service_self_hosted`** (hosting_type = self_hosted): three-helix instance/host, deployment ref, runbook ref, upstream OSS project (when third-party). > **Read-model boundary (ADR-001):** catalog writes are administrative > bootstrap data (like the TPSC catalog today), not derived state. They reuse the > existing TPSC write surface pattern; no new sanctioned read-model write is > introduced. ## Migration & back-compat - Existing `tpsc_catalog` rows migrate into `service_catalog` as `(hosting_type=cloud_hosted, development_type=third_party)`; the GDPR/DPA/ ToS/privacy/region/retention fields move into `service_cloud`. - `tpsc_entries` / `tpsc_snapshots` (per-repo dependency declarations) keep pointing at the catalog via `service_id`; the `/tpsc/*` endpoints remain as a filtered view (`development_type=third_party`) so existing ingestion and the current page keep working during transition. ## Open questions (resolve in T1, do not block proposal) - Whether to generalise the `tpsc_catalog` table in place vs. introduce `service_catalog` and keep `tpsc_catalog` as a view/alias. Lean: new `service_catalog`, repoint TPSC reads. - `maturity_level` source of truth: operator-set vs. derived from a future `check-dom` evaluation. Start operator-set; leave room to derive later. --- ## Tasks ### T1 — Core + extension data model and migration ```task id: STATE-WP-0062-T01 status: done priority: high state_hub_task_id: "f74612f6-b442-4b5f-ac4d-7bc6d1d4883f" ``` - [x] `api/models/service_catalog.py`: `ServiceCatalog` core + `ServiceThirdParty`, `ServiceFirstParty`, `ServiceCloud`, `ServiceSelfHosted` extensions. - [x] Alembic migration; register in `api/models/__init__.py`. - [x] Data migration: copy `tpsc_catalog` → `service_catalog` + `service_cloud` with `(cloud_hosted, third_party)`; backfill `service_id` on TPSC entries. ### T2 — API + MCP ```task id: STATE-WP-0062-T02 status: done priority: high state_hub_task_id: "373cddfa-c85b-47a7-bb8e-c86e9341c237" ``` - [x] `GET /services/catalog` with `hosting_type` / `development_type` / `maturity_level` filters; `GET /services/{slug}` returns core + applicable extensions. - [x] Write path to register/update a service and its extensions (generalise `register_service`; keep a `third_party`-shaped compatibility wrapper). - [x] Keep `/tpsc/*` working as a `development_type=third_party` view. ### T3 — Services nav section + four-quadrant pages ```task id: STATE-WP-0062-T03 status: done priority: high state_hub_task_id: "b14bbbdd-347e-4f62-9ac3-ad42360fa766" ``` - [x] Replace the top-level "Services (TPSC)" entry with a **Services** section. - [x] Pages: **Third Party** (current TPSC view), **First Party** (with internal repo link + **Service Maturity Level** column), **Self Hosted** (three-helix infra view). Each is a filtered view over `service_catalog`. - [x] Surface the two dimensions explicitly (e.g. a quadrant filter) so the four classes are navigable. ### T4 — Terminology: Tier → Level (Service DoM only) ```task id: STATE-WP-0062-T04 status: done priority: medium state_hub_task_id: "1c296d7a-c090-430f-8d0d-27b548d88d9e" ``` - [x] `policies/service-dom.md`: rename "Tier 1/2/3" → "Level 1/2/3"; column header "Service Maturity Level". **Do not** touch the DoI tier subsystem (`repos.md`, `check_doi.py`, `doi_cache` migration) — that is a separate, established concept. ### T5 — Tests, docs, seed ```task id: STATE-WP-0062-T05 status: done priority: medium state_hub_task_id: "2edd68ee-0d9a-431d-9891-73a0f6be6a41" ``` - [x] Tests: model + migration round-trip, dimension filters, TPSC back-compat view, first-party↔repo link. - [x] Docs: `/docs/services` reference; update `/docs/tpsc` to point at the generalised model. - [x] Seed the known classes: TPSC SaaS (cloud/third-party), self-hosted OSS from `tools.md` (Gitea, Postgres — self-hosted/third-party), and State Hub itself (self-hosted/first-party). --- ## Acceptance - Every service carries an explicit `hosting_type` and `development_type`; the four classes are queryable and navigable. - Common fields live once in `service_catalog`; dimension-specific data lives in composable extension tables (third-party upstream contacts, first-party repo link, cloud data-processing/GDPR, self-hosted infra). - The existing TPSC page and ingestion keep working through the back-compat view. - The Services nav section exposes Third Party / First Party / Self Hosted, with a Service Maturity **Level** column on First Party. - The Service DoM policy uses "Level", not "Tier". ## See also - `api/models/tpsc.py`, `api/routers/tpsc.py` — prior art being generalised - `policies/service-dom.md` — Service Definition of Mature (Level source) - `dashboard/src/tools.md` — current static self-hosted listing (seed source) - `policies/repo-doi.md` / `dashboard/src/repos.md` — the DoI **tier** subsystem that is intentionally left unchanged