Skip to content

AIFACTORY-FEAT-079: Studio work area: Quality & releases

Generated from the architecture model

This page is rendered from model/**/*.yaml by tools/docs_model during mkdocs build. Edit the YAML, not this page; hand-written design lives on the linked service, platform and plane pages.

A design artefact, not a tracker

This backlog is generated from model/backlog/ and validated against the architecture model. No work item exists in any tracker, and no item carries a date, an estimate or an assignee (ADR-0029).

Epic Factory Studio (PF-14) · Increment I3 Studio · State elaborated · Route frontend-developer · Build order 119 of 185 (Factory code)

The 6 screens of the Studio work area Quality & releases on the Blazor shell and microfrontends, bound to the read models and commands the model names for them, with their state families, roles, locales and accessibility contracts. Screens come from model/ux/screens.yaml and the prototype alignment.

Delivers —
Stubs —
Realizes ux work areas: Quality & releases
Requirements —
Depends on —
Sources model/ux/screens.yaml, model/ux/prototype-alignment.yaml

AIFACTORY-US-1405: Quality and releases: I read product releases with their components, publication evidence and provenance, create one from a succeeded pipeline run and withdraw one with a reason

behavior · persona operator As an operator, I want to see every release of a generated product with its components, digests, publication evidence and provenance, and to create or withdraw a release from the Studio, so that only a release whose every component is published can be promoted, and a bad release is stopped before it reaches an environment.

An operator, a workspace admin or an observer opens the screen product-releases at /quality/releases. StudioBff composes it under the scope of the session from ReleaseService, reached through DeliveryHost: ListProductReleases by product, version and state, and GetProductRelease for one release with its components and digests, the target feed or registry and the publication state of each component, the SBOM and attestation references, the pipeline run and change proposals that produced it and the asset releases that generated it. The delivery family of model/ux/states.yaml governs the display: a release is Assembling, Candidate, Released, Superseded or Withdrawn, and a Candidate names the components that still lack publication evidence. An operator or a workspace admin creates a release with CreateProductRelease from a pipeline run that succeeded, sent through ForwardCommand with an idempotency key; a run that did not succeed or a version already used is answered 409 or 422 as ReleaseService answers, and the page says which, and no release is shown as created until ReleaseCreated is consumed. They withdraw a release with WithdrawProductRelease and a reason; the page says that withdrawal blocks promotion and revokes nothing already deployed, and a release already withdrawn is answered 409. An observer sees the commands disabled with the reason. StudioBff consumes ReleaseCreated, ReleaseReady and ReleaseWithdrawn and refreshes the release view, so a Candidate turns Released on the open page without a reload. A release of another tenant is answered 404 like one that does not exist. When DeliveryHost does not answer, the source is shown degraded with its last freshness and the commands are refused with 503, never queued.

Acceptance criteria

AC1 Given an operator on product-releases, when StudioBff/GetScreenModel composes it, then releases come from ReleaseService/ListProductReleases and ReleaseService/GetProductRelease with components, digests, publication state, SBOM and attestation references and provenance, and a Candidate names each component still without publication evidence (FUX-02, model/ux/states.yaml).
AC2 Given an operator or a workspace admin, when they create a release, then StudioBff/ForwardCommand forwards ReleaseService/CreateProductRelease with an idempotency key, a 409 or 422 is explained in words, and the release appears only once ReleaseService/ReleaseCreated is consumed; an observer sees the command disabled with the reason (FUX-11).
AC3 Given a release, when an operator withdraws it with a reason through ReleaseService/WithdrawProductRelease, then the page states that promotion is blocked and nothing deployed is revoked, a repeated withdrawal is answered 409, and ReleaseService/ReleaseWithdrawn or ReleaseService/ReleaseReady refresh the open page without a reload (ADR-0015).
Traces: services StudioBff, ReleaseService
Traces: apis StudioBff/GetScreenModel, StudioBff/ForwardCommand, ReleaseService/ListProductReleases, ReleaseService/GetProductRelease, ReleaseService/CreateProductRelease, ReleaseService/WithdrawProductRelease
Traces: events ReleaseService/ReleaseCreated, ReleaseService/ReleaseReady, ReleaseService/ReleaseWithdrawn
Traces: adrs ADR-0015
Traces: requirements FUX-02, FUX-11
Relates: screens product-releases
Depends on feature:studio-bff, feature:component-blazor-shell, feature:release-service, feature:delivery-host
Sources model/ux/screens.yaml, model/ux/states.yaml, model/services/devops-gitops/ReleaseService.yaml

Tasks

Task Kind Expected output Repository
AIFACTORY-TASK-5663 Product releases screen in the Quality microfrontend ui The Quality microfrontend renders product-releases with the delivery and freshness families, the list by product, version and state, the release detail with components, digests, publication state, SBOM and attestation references and provenance, the create form from a succeeded pipeline run, and the withdraw dialog that asks for a reason and states its effect. ConnectSoft.AI.SoftwareFactory.Studio
AIFACTORY-TASK-5664 Screen model and command routes of the product releases screen in StudioBff contract StudioBff composes product-releases through DeliveryHost from ListProductReleases and GetProductRelease with the freshness of the source, forwards CreateProductRelease and WithdrawProductRelease with an idempotency key and the scope of the session, refuses them with 503 while DeliveryHost does not answer, and refreshes the release view on ReleaseCreated, ReleaseReady and ReleaseWithdrawn. ConnectSoft.AI.SoftwareFactory.StudioBff
AIFACTORY-TASK-5665 Verify the product releases screen with refusals and live readiness verification Run dotnet test -c Release --no-build and npx playwright test --project=chromium with a StudioBff stub serving a Candidate and a Released release, a creation from a pipeline run that did not succeed answered 422, a creation with a version already used answered 409, a 409 on a repeated withdrawal, a pushed ReleaseReady and an observer session; the tests prove AC1 to AC3 of AIFACTORY-US-1405 (ReleaseService/CreateProductRelease, ReleaseService/WithdrawProductRelease). ConnectSoft.AI.SoftwareFactory.Studio
AIFACTORY-TASK-5705 Verify the StudioBff side of the product releases screen verification Run dotnet test -c Release --filter "Category=Acceptance&Feature=studio-quality-product-releases" in StudioBff against stubbed platform gateways, where a filter that matches no test fails the run; the tests prove on the StudioBff side AC1 (composition through DeliveryHost from ListProductReleases and GetProductRelease with freshness), AC2 (CreateProductRelease forwarded once with its idempotency key, its 409 and 422 returned unchanged, refresh on ReleaseCreated) and AC3 (WithdrawProductRelease forwarded with its reason, its 409 returned unchanged, refresh on ReleaseWithdrawn and ReleaseReady, 503 while DeliveryHost does not answer and nothing queued) of AIFACTORY-US-1405 (StudioBff/GetScreenModel, StudioBff/ForwardCommand). ConnectSoft.AI.SoftwareFactory.StudioBff

AIFACTORY-US-133: The Quality & releases work area of the Studio is bound to the model

behavior · persona asset-maintainer As an asset maintainer, I want the Quality & releases work area of the Studio (6 screens) bound to the read models, commands, state families and roles the model names for it, so that I can follow product releases from one place now and see plainly that evaluations, evidence, feedback and learning arrive with their services, with every control reflecting what I may do and how fresh what I see is.

This is the integration PBI of the Quality and releases area. The Quality microfrontend is generated from the microfrontend library template, composed into the Studio shell through its registry and holds the six screens of model/ux/screens.yaml with their routes, read models, commands, state families and roles; it calls StudioBff only. A screen is offered to the roles its list names; for any other role it is absent from the navigation and GetScreenModel refuses it with 403. One screen, product-releases, is delivered in I3 by AIFACTORY-US-1405. Five read services of I4: evaluation-runs and release-evidence read the complete QualityScoreService, whose thin delivery only stores quality events and serves no evaluation, and feedback, datasets and learning-proposals read FeedbackService. In I3 each of these five screens is registered, offered to its roles and shows its sources as unknown in the freshness family, named, never hidden and never filled from a fake; none of its commands (StartEvaluation, WaiveEvaluation, RecordFeedback, SnapshotDataset, DecideLearningProposal) is offered, so taking an asset from Draft to Published with evaluation evidence waits for I4. Every screen subscribes through Realtime.Subscribe of StudioRealtimeGateway; when the channel drops the page marks its data stale and when it returns the page reloads the screen model. A command carries the idempotency key and the scope of the session to the owning service, whose rules decide. Every screen renders in English, Russian and Hebrew right to left and is built to the accessibility contract of docs/planes/experience/studio/localization-accessibility.md; the area evidence that follows this PBI checks it on the finished area.

Acceptance criteria

AC1 Given the shell with the Quality microfrontend, when each of the 6 screens of model/ux/screens.yaml opens for a role its list names, then it renders from StudioBff/GetScreenModel only, and for a role its list does not name the screen is not offered and StudioBff/GetScreenModel answers 403 (ADR-0007, ADR-0015).
AC2 Given evaluation-runs, release-evidence, feedback, datasets and learning-proposals, when they open in I3, then each shows its sources as unknown in the freshness family of model/ux/states.yaml with the source named, shows no verdict, score or record, and offers none of its commands (FUX-04).
AC3 Given an open screen and its subscription through StudioRealtimeGateway/Realtime.Subscribe, when the channel drops, then the page marks the data stale and never presents it as current, and when the channel returns the page reloads StudioBff/GetScreenModel and shows the current state without a browser reload (FUX-11).
AC4 Given a command of the delivered screen, when a role issues it, then it is forwarded once through StudioBff/ForwardCommand with an idempotency key and the scope of the session, a role without the right sees the control disabled with the reason, and the screens render in en, ru and he right to left under docs/planes/experience/studio/localization-accessibility.md (ADR-0026).
Traces: services StudioBff, StudioRealtimeGateway, ReleaseService
Traces: apis StudioBff/GetStudioSession, StudioBff/GetScreenModel, StudioBff/ForwardCommand, StudioRealtimeGateway/Realtime.Subscribe
Traces: adrs ADR-0007, ADR-0015, ADR-0026
Traces: requirements FUX-04, FUX-11
Relates: screens evaluation-runs, release-evidence, feedback, datasets, learning-proposals
Depends on AIFACTORY-US-1405, feature:studio-bff, feature:studio-realtime-gateway, feature:component-blazor-shell
Sources model/ux/screens.yaml, model/ux/states.yaml, model/ux/prototype-alignment.yaml, docs/planes/experience/studio/screens/quality-releases.md

Tasks

Task Kind Expected output Repository
AIFACTORY-TASK-445 Screens of the work area Quality & releases: Evaluation runs, Release evidence, Feedback, Evaluation datasets, Learning proposals, Product releases ui The Quality microfrontend is composed into the shell through the registry with one navigation group and the six routes of model/ux/screens.yaml; each screen renders from StudioBff only, and the five screens of I4 sources render their sources as unknown with the source named and no command. ConnectSoft.AI.SoftwareFactory.Studio
AIFACTORY-TASK-446 Commands of the work area Quality & releases implementation The commands of the delivered screen (CreateProductRelease, WithdrawProductRelease) are forwarded through StudioBff with an idempotency key and the scope of the session, and the five commands of the I4 screens (StartEvaluation, WaiveEvaluation, RecordFeedback, SnapshotDataset, DecideLearningProposal) are not offered in I3. ConnectSoft.AI.SoftwareFactory.Studio
AIFACTORY-TASK-447 States, roles, locales and accessibility of the work area Quality & releases implementation Each screen shows the state families the model names for it, hides or disables what the role may not do with the reason, marks stale data on a dropped channel and reloads on its return, renders in every supported locale including right-to-left, and passes the accessibility checks of the UX model. ConnectSoft.AI.SoftwareFactory.Studio
AIFACTORY-TASK-747 Verify studio-quality-releases-work-area by running the declared command verification Run dotnet build -c Release, dotnet test -c Release --no-build and npx playwright test --project=chromium; they pass and prove AC1 to AC4 of AIFACTORY-US-133 (StudioBff/GetScreenModel, StudioRealtimeGateway/Realtime.Subscribe). ConnectSoft.AI.SoftwareFactory.Studio

AIFACTORY-US-1406: Quality and releases area evidence: every screen passes the accessibility and phone-width checks and an end-to-end walk of the area through StudioBff is recorded

evidence · persona observer As an observer, I want proof that every screen of the Quality and releases area is usable by keyboard, in every locale and at phone width, and that the evaluation and feedback screens say they are not yet delivered, so that the area is accepted on a recorded check and no quality score or evaluation verdict is read from a fake.

This PBI is the proof of the Quality and releases area and comes after its integration PBI AIFACTORY-US-133, which registers the six screens. The accessibility and phone-width check opens each of the six screens of the area for a role its list in model/ux/screens.yaml names and asserts the rules of docs/planes/experience/studio/localization-accessibility.md: WCAG 2.2 AA by the automated checks, keyboard operation, live updates announced politely and not animated under reduced motion, digests and versions readable by a screen reader, refusals explained in words, in English, Russian and Hebrew right to left, light and dark, at 1440 by 900 and at 390 by 844 with no horizontal scroll. The end-to-end walk signs in as an operator, opens the product releases of the fixture product of the first outcome, creates a release from its succeeded pipeline run, watches the Candidate become Released when the publication evidence arrives, and withdraws a second release with a reason; as an observer it sees the same screen without commands. The screens whose sources are delivered in I4 (evaluation runs, release evidence, feedback, evaluation datasets, learning proposals) show their sources as unknown and are listed by name. The record holds per screen and role the expected and observed result and the network log, which shows no address other than StudioBff and StudioRealtimeGateway.

Acceptance criteria

AC1 Given each of the 6 screens of the area in model/ux/screens.yaml, when the accessibility check runs in the three locales, both schemes and the two viewports of docs/planes/experience/studio/localization-accessibility.md, then every screen passes the automated WCAG 2.2 AA checks and keyboard operation and has no horizontal scroll at 390 by 844 (NFR-15, FUX-12).
AC2 Given the walk as an operator and an observer, when it runs on the acceptance estate, then the product releases screen renders from StudioBff/GetScreenModel, the create and withdraw commands go through StudioBff/ForwardCommand, and the five screens of later sources are listed by name with their sources unknown (ADR-0007, FUX-04).
AC3 Given the record and the network log of the walk, when they are read, then the browser called no address other than StudioBff and StudioRealtimeGateway and no screen showed an evaluation verdict, a quality score or a feedback record in I3 (FUX-12).
Traces: services StudioBff
Traces: apis StudioBff/GetScreenModel, StudioBff/ForwardCommand
Traces: adrs ADR-0007
Traces: requirements FUX-04, FUX-12
Relates: screens evaluation-runs, release-evidence, feedback, datasets, learning-proposals, product-releases
Relates: nfrs NFR-15
Depends on AIFACTORY-US-1405, AIFACTORY-US-133, feature:studio-bff, feature:component-blazor-shell
Sources model/ux/screens.yaml, docs/planes/experience/studio/localization-accessibility.md, docs/planes/experience/studio/screens/quality-releases.md

Tasks

Task Kind Expected output Repository
AIFACTORY-TASK-5666 End-to-end walk of the Quality and releases area fixture A browser-driven walk in the Studio repository that signs in as an operator and an observer, creates and withdraws releases of the fixture product, opens the five screens of later sources, and records per screen and role the rendered state, the source markers, the offered commands and the addresses the browser called. ConnectSoft.AI.SoftwareFactory.Studio
AIFACTORY-TASK-5667 Accessibility and phone-width check of the Quality screens verification Run dotnet test -c Release --filter Category=Accessibility, npx playwright test --project=chromium at 1440 by 900 and npx playwright test --project=phone-390x844, the Playwright project of the Studio at 390 by 844, on the six screens in en, ru and he, light and dark; the results prove AC1 of AIFACTORY-US-1406 against docs/planes/experience/studio/localization-accessibility.md. ConnectSoft.AI.SoftwareFactory.Studio
AIFACTORY-TASK-5668 Verify the walk record of the Quality area verification Run the walk of AIFACTORY-TASK-5666 on the acceptance estate; its record and network log prove AC2 and AC3 of AIFACTORY-US-1406 (StudioBff/GetScreenModel, StudioBff/ForwardCommand) and are kept with the evidence of increment I3. ConnectSoft.AI.SoftwareFactory.Studio