cesnimda
4.7 KiB
FlowCV research and Jobjakt document-workflow plan
Completed 21 October 2026. Research and planning only. No Jobjakt code, database, migration, dependency or deployment changes. All artifacts are outside the repository.
Read the deliverables
- Detailed FlowCV UX audit — final coverage matrix, manual sequences, network limits and 37 screenshot + DOM/accessibility checkpoints.
- Component and interaction inventory — controls, states, affordances, accessibility observations and evidence IDs.
- Jobjakt source inventory and gap analysis — existing routes, editors, models, renderers, design settings, localisation, tests and constraints, with source locations.
- Original Jobjakt design proposal — application context, provenance, reversible design choice and a complete letter workflow using current systems.
- Desktop/mobile wireframe specifications — annotated layouts, responsive behaviour, focus, save/error states and interaction rules.
- Phased technical implementation plan — files, proposed contracts, dependencies, six reviewable tasks and rollback/release boundaries.
- Acceptance criteria and test plan — 15 criteria, existing tests to extend, failure/abuse/resilience matrix and a deliberate Break Jobjakt phase.
Decision summary
Verified — FlowCV: split content/preview editing, template gallery → detail → explicit apply, one successful template undo/redo, mouse section reorder, responsive edit/preview, explicit resume contact/design copy into a letter, and completed PDF response paths. AI generation and successful duplication were blocked by plan gates. Network metadata supports automatic saving before Done; the exact persistence mechanism remains Inferred.
Verified — Jobjakt source: the existing CV editor already has autosave, optimistic concurrency, recovery, settings history, undo/redo, responsive physical-page preview, custom content, design controls, PDF and reviewed AI proposals. It should be extended, not replaced. Runtime behaviour was not exercised.
Inferred priority: improve rendered template candidates and ownership/history wording first. Complete application cover-letter preview, compatible design matching and PDF as a separate project. Before persistent letter design or autosave, unify revision/history handling across every writer; several source paths currently bypass the dedicated letter version service. This is a source-derived risk, not a reproduced race.
Proposed architecture: existing React/MUI/theme/routes and .NET/EF services; no new dependency initially. Keep letters application-scoped and explicitly saved. Contact copy and design matching are reviewed snapshots; optional identity fields are excluded by default. Keep frozen application records distinct from live CV settings history.
Evidence and privacy
Findings use Verified, Inferred or Not Tested throughout. Verified source is not a runtime pass. The browser was operated manually by the user; REA was passive. Only curated synthetic-document checkpoints are included. The initial pre-synthetic observation and raw session ledger are excluded. No passwords, cookies, auth tokens, raw request/response bodies or original personal CV content are in this package.
Each checkpoint has a PNG plus *-evidence.json containing selected DOM/AX, findings, manual action and original REA evidence identifiers. Network metadata is limited to selected method/path/status/type; no identifiers/query values are retained. Proprietary sample-template screenshots are reference evidence only, not reusable design assets. File integrity manifest records SHA-256 hashes for the deliverables and curated evidence files; it is a packaging check, not an independent attestation of original capture time.
Source baseline: b2a9fd68b081a8cc1701b1c58eaea7e1070274af, branch codex/ui-redesign-phase4, repository E:/Storage/Projects/Active/Job tracker. Working tree clean at initial and final verification. The gap report uses repository-relative path:line evidence for portability.
Limits and next decision
No Jobjakt tests, build, database, runtime UI or production checks were run. FlowCV paid features, initial resume creation, live cross-document synchronisation, version history, manual/pinch zoom, real phone use and downloaded PDF fidelity remain untested. The proposal is ready for review; implementation requires your explicit approval. No production deployment is included in the plan.