jobtrackingapp/docs/architecture/application-workspace.md

Application Workspace (Phase 5)

Phase 5, Milestones 1–2 (2026-07-19). The per-application workspace: what it is, what it deliberately is not, and how it composes existing systems. Companion to cv-builder.md, ai-career-assistant.md, career-profile-model.md.

What it is

A dedicated surface for one JobApplication at /applications/{id}, so an application is a place you work rather than a row you edit in a modal. Job tracking stays the product; the workspace is the application's home.

Core principle: the user should never ask "what do I do next?" — the overview always answers it.

What it is NOT

The workspace owns no data and duplicates none. It is an aggregate read plus a navigation shell:

Section	Backed by (existing system)
Timeline	JobEvent — interpreted, never replaced
Analysis / Match	the advert and the master CareerProfile, read deterministically
Checklist	ApplicationChecklistItem — completion state only, seeded from the readiness signals
CV	Phase 4 CvVariant — a lens over the master CareerProfile
Analysis / Match / Interview	Phase 5 AiWorkspacePanel + AiInteraction history
Documents	Attachment
Communication	Correspondence
Activity / Timeline	JobEvent
Stage semantics	JobPipeline

No career data is copied into the application. Nothing in this feature writes to the master profile, a CV variant, or a cover letter.

Backend

GET /api/jobapplications/{id}/workspace → WorkspaceOverviewDto (ApplicationWorkspaceController + ApplicationWorkspaceService).

One aggregate read instead of the page fanning out: role/company/location/salary, status + pipeline group, applied/deadline/follow-up dates, the attached CV variant (id, name, theme), cover-letter presence, document count, AI interaction count + last run, recent JobEvent activity, and the computed next recommended action.

Read-only and tenant-scoped (OwnerUserId), returning 404 for another user's application.

Next recommended action

Milestone 2 moved this onto the checklist: the first pending checklist item, ordered by category priority (preparation → submission → follow-up → interview → custom) then the user's own ordering. null means nothing is outstanding.

That means the overview cannot recommend something the user has already ticked off, a dismissed item never comes back as a recommendation, and a task the user added themselves can legitimately be the next action. There is no second ruleset to keep in sync.

The checklist

ApplicationChecklistItem — a workflow guidance layer, not a store of truth. It records only "is this step done, and does the user still want it". The CV still lives in CvVariant, documents in Attachment, history in JobEvent, follow-up in JobApplication.FollowUpAt.

System items and auto-completion

Each default item carries a stable SystemKey and usually an AutoSignal — the same signal /readiness already computed. On every read the service re-syncs:

• signal satisfied + item pending → done, IsAutoCompleted = true
• signal no longer satisfied + item was auto-completed → back to pending
• a manual tick clears IsAutoCompleted and therefore sticks, even against the signal

So "readiness says the CV is missing" and "the checklist says Prepare a CV is pending" cannot drift apart — they read the same state. The user always wins over the signal.

System key	Category	Signal
review-job-details	preparation	advert text present
complete-career-profile	preparation	career profile with at least one experience
prepare-cv	preparation	CV variant attached, or tailored CV text
review-cv-match	preparation	manual
create-cover-letter	preparation	cover letter present
attach-portfolio	preparation	HasPortfolio
attach-supporting-documents	preparation	at least one Attachment
save-application-answers	preparation	saved answer draft in Notes
capture-recruiter-contact	preparation	Company.RecruiterEmail
confirm-submitted	submission	applied date set and out of the prospect stage
add-follow-up-reminder	follow-up	FollowUpAt set
set-next-action	follow-up	NextAction written
prepare-interview-notes	interview	prep notes present, or not at an interview stage
research-company	interview	manual

Seeding is idempotent per (JobApplicationId, SystemKey) — enforced by a unique index, so a re-read never duplicates. Custom items have a NULL SystemKey; both SQLite and MariaDB treat NULLs as distinct in a unique index, so a user can add as many as they like.

Deleting a system item dismisses it (a hard delete would be undone by the next seed); deleting a custom item removes the row. Dismissed items leave the progress denominator entirely.

API

/api/jobapplications/{id}/checklist — GET (seeds + syncs + returns items and progress), POST (custom item), PATCH /{itemId}, DELETE /{itemId}, PUT /order (array of ids). Tenant-scoped on OwnerUserId; another user's application is a 404.

Relationship to the other systems

• Not JobEvent — the checklist is forward-looking intent; JobEvent is the append-only history.
• Not follow-ups — FollowUpAt, RulesEngine and the reminder hosted service still own scheduling. The checklist only asks whether a follow-up exists.
• Not Attachment / CvVariant — it reads their presence as a signal and stores nothing of them.

Future AI suggestions

An AI-suggested task is just a checklist row with IsSystemGenerated = false and no AutoSignal, created after the user approves it. Nothing in the AI path may create, complete or delete an item without approval — same rule as everywhere else (ai-career-assistant.md).

Schema provisioning

ApplicationChecklistItems follows the established MariaDB-safe path: the scaffolded migration (20260719085904_AddApplicationChecklistItems) is a no-op, and the table is created by the idempotent reconciler in StartupInitializationExtensions, which has correct DDL per provider. A SQLite-scaffolded migration would emit TEXT datetimes and a PK without AUTO_INCREMENT on MariaDB — the failure that crashed prod startup for the Phase 4 tables.

Verified on MariaDB 11: int AUTO_INCREMENT PK, varchar/datetime(6)/tinyint(1) columns, both indexes inside the 3072-byte key limit, cascade delete from JobApplications, the unique index rejecting a duplicate system key, and NULL system keys not colliding.

Frontend

ApplicationWorkspacePage (/applications/:id) — a left nav plus a content pane, section selected by ?section=, so a section is linkable and survives refresh. Reached from the job dialog's "Open application workspace" button.

The dialog passes an optional onOpenWorkspace callback rather than calling useNavigate itself: JobDetailsDialog must stay renderable without a <Router> (several suites mount it standalone), so router context belongs to the caller.

The Checklist section (ApplicationChecklist) groups items by category, shows a completion bar, and supports tick/untick, add, remove and reorder. System items are labelled "Detected" when a signal completed them, custom items "Yours". Every mutation re-reads, because only the backend's sync knows the real post-mutation state.

Implemented now: Overview, Checklist, Job Details, and the sections that reuse an existing component (Analysis/Match/Interview → AiWorkspacePanel, Documents → Attachments, Communication → Correspondence). Sections owned by later milestones state their milestone instead of faking functionality.

Relationship to /readiness

GET /{id}/readiness keeps its DTO shape (score, level, completed, missing, reminders, workflowSignal) and still backs the dialog's Readiness tab — but as of Milestone 2 it no longer runs its own parallel checklist. It projects the persisted checklist:

• completed / missing — the live (non-dismissed) items by status
• score — the checklist completion percentage
• level — Ready ≥ 80, Needs polish ≥ 60, otherwise Needs work
• reminders / workflowSignal — unchanged; BuildWorkflowSignal remains the health/attention view

So the division is: the checklist is the workflow the user drives, readiness is the calculation and health indicator derived from it. One system, two projections.

Application intelligence (Phase 5.3)

Three read-only reads that answer "how suitable is this job", "how does my experience match", "what am I missing", "what happened previously". All deterministic, all owned by nothing:

Endpoint	Reads	Owns
GET /{id}/timeline	JobEvent	nothing
GET /{id}/analysis	JobApplication.Description	nothing
GET /{id}/match	CareerProfile + the advert	nothing

No new table, no new column. ApplicationTimelineService and ApplicationIntelligenceService write nothing at all.

The AI boundary

The deterministic answer and the AI narrative are separate on purpose.

• The three endpoints above never call the AI. Opening the Analysis or Match section costs nothing and cannot change anything — the page renders a computed answer.
• The narrative lives where it already did: AiWorkspaceService's job-analysis and career-match modules, reached from AiWorkspacePanel, generated only when the user asks.
• Every generation is appended as an AiInteraction — that append-only history is the versioning, and the user restores, compares or deletes from it.
• AI output is a suggestion. Nothing in this phase writes to the CareerProfile, a CvVariant, a cover letter, or the JobApplication. Match_never_writes_to_the_career_profile pins that.

So a user gets a trustworthy number for free, and pays for prose only when they want it.

Timeline

JobEvent stays the source of historical truth; the service is an interpretation layer over it. Each row gains a readable summary (("StatusChanged", "Applied", "Interview") → "Moved from Applied to Interview"), a category (lifecycle, stage, follow-up, communication, ai) and a milestone flag. Events group by day with relative labels.

Milestones are the stages that mean something happened — applied, interview, offer, rejected, accepted, declined — plus creation and replies received. They are returned unfiltered: narrowing the detail below must not hide what actually happened.

An unrecognised future Type degrades to a humanised sentence rather than disappearing.

Job analysis

Structured extraction from the advert, reusing the existing SkillTagger so the vocabulary matches the job importer and the CV match. Facts (role, company, location, employment type, seniority, salary), lists (technologies, skills, responsibilities, requirements, keywords, interview topics), and — deliberately — what the advert does not say, which is usually the more useful half.

With no advert saved it degrades to the fields on the application and says so.

Career matching

Feeds the master CareerProfile into the same JobCvMatchService the CV builder uses, so one application scores identically whichever surface asks. Returns the score and band, matched and missing skills, and — the part that makes it actionable — which experience and project entries are the evidence for each matched keyword, ranked by hit count.

Suggestions describe what the user could change. They never change it.

With no profile it returns score 0 and asks the user to build one, rather than implying a bad match.

Extension points

• New section: add to WORKSPACE_SECTIONS and render it; nav is data-driven.
• New next-action rule: add one clause to ApplicationWorkspaceService.NextStep — ordered, so position is the priority.
• More overview data: extend WorkspaceOverviewDto; the page reads one payload.

Milestones

1. ✅ Workspace foundation — route, nav shell, aggregate overview, next recommended action.
2. ✅ Checklist and progress tracking — persisted items, auto-completion from the readiness signals, custom items, reordering, dismissal; readiness refactored into a projection of it.
3. ✅ Application intelligence — timeline interpretation, structured job analysis, career matching (Phase 5.3, all three deterministic and read-only).
4. CV integration.
5. Cover letter workflow. 8. Documents. 9. Interview preparation. 10. Dashboard improvements.