Skip to content

Teacher View Stability Analysis Request

Purpose

This document is an external analysis request for the current instability of the SkillPilot teacher UI.

The goal is not another superficial patch, but a clean technical diagnosis of why the teacher experience is currently unstable and sluggish, plus a concrete stabilization plan.

Problem Summary

The teacher UI is currently unstable enough to be considered a release blocker.

Observed symptoms:

  • The teacher view intermittently jumps between:
  • the class detail view, and
  • the teacher dashboard / empty class overview.
  • The middle learning-context tree sometimes disappears completely although a class is selected.
  • The structure switch (Inhalte / Kompetenzen) has shown oscillating behavior and inconsistent state.
  • The UI feels sluggish and unstable overall.
  • The class creation flow has shown context mismatch behavior:
  • after choosing a high-level curriculum such as Gymnasium (DE), the teacher still saw the full global landscape list instead of the subject list inside that curriculum.
  • Historical state stored in browser local storage has repeatedly interfered with current behavior.
  • PWA/service-worker caching may amplify the perception because outdated bundles remain active after fixes.

In short:

  • the teacher UI currently has state management problems,
  • context resolution problems,
  • and likely redundant or conflicting synchronization paths.

Current User-Facing Severity

Severity: High

Why this is critical:

  • The teacher cannot reliably trust what is shown on screen.
  • The screen state can change unexpectedly.
  • Core workflows such as opening a class, seeing the tree, switching student, or switching structure become unreliable.
  • This undermines the usability of the teacher cockpit as a professional tool.

Main Suspected Problem Class

The problem does not currently look like a pure curriculum-data issue.

The more likely class of issue is:

  • conflicting frontend state sources,
  • remount/hydration effects,
  • URL synchronization side effects,
  • localStorage restoration races,
  • teacher-context migration logic,
  • and PWA stale-bundle behavior.

The likely core architectural smell is:

the same teacher context is influenced by too many channels at once.

Relevant channels include:

  • React local state,
  • URL search params,
  • route params,
  • localStorage,
  • migrated stored teacher class sessions,
  • active landscape selection in app core,
  • active goal selection,
  • active student selection,
  • and tree projection mode.

Relevant Areas in Code

Primary files to inspect:

Secondary context:

Recent Changes That May Be Relevant

The teacher area has recently undergone several changes, including:

  • migration from legacy/compatibility curriculum IDs toward canonical Gymnasium subject landscapes,
  • class-session migration logic based on old saved teacher state,
  • root-goal selection fallback logic,
  • teacher learner selection rewiring,
  • class setup landscape scoping,
  • local removal of the structure-mode toggle from learner view,
  • localization and notification cleanup,
  • partial movement from client-side filtering toward server-side scoped learner graph loading,
  • placement-based structural projections and competency-axis projections.

This means the teacher view is currently sitting on top of a moving architecture.

Reproduction Examples

Typical reproduction paths that have shown unstable behavior:

  1. Open teacher role.
  2. Open an existing class such as Mathe LK.
  3. Observe:
  4. no tree appears in the middle panel, or
  5. the UI jumps back to the teacher dashboard.

Other repros:

  1. Open teacher role.
  2. Switch between Alle and a specific student.
  3. Observe inconsistent content pane / tree behavior.

Another repro:

  1. Open teacher role.
  2. Use class setup.
  3. Expect subject choices inside Gymnasium (DE).
  4. Observe global landscape list instead.

Another repro:

  1. Open teacher class with math.
  2. Toggle Inhalte / Kompetenzen.
  3. Observe oscillation, wrong active state, or empty view.

What Has Already Been Tried

Several local fixes have already been attempted, including:

  • mapping legacy teacher landscape IDs to canonical IDs,
  • migrating stored teacher class sessions,
  • reconstructing missing teacher landscapeId values from class names,
  • moving teacher classes/active class restoration into initial state to reduce visible dashboard/detail flicker,
  • removing the learner-facing structure toggle,
  • removing global URL synchronization of structure mode and keeping it local to teacher view,
  • narrowing class setup landscape choices to the current curriculum context.

These changes improved individual symptoms, but the teacher area is still perceived as too unstable.

What We Need From The External Expert

We need a focused technical analysis with the following outputs.

1. Root Cause Analysis

Please identify the real causes behind:

  • teacher view/dashboard jumping,
  • missing middle tree,
  • structure-mode instability,
  • sluggishness,
  • and context inconsistency.

Please do not stop at symptoms. We need the actual state-flow failure modes.

2. State Ownership Model

Please answer:

  • Which component or hook should be the single source of truth for teacher context?
  • Which state should live:
  • in route params,
  • in search params,
  • in local component state,
  • in app-core state,
  • in localStorage,
  • and nowhere at all?

We need a crisp ownership model, not a best-effort sync mesh.

3. Race / Remount Analysis

Please inspect whether the instability is caused by:

  • teacher remounts,
  • localStorage hydration after initial render,
  • route transitions,
  • context updates triggering selection resets,
  • effect ordering,
  • or React StrictMode / double-render sensitivity.

4. Simplification Recommendations

Please propose the minimal architecture that makes the teacher cockpit stable.

We are explicitly open to removing behavior if it reduces instability.

5. Performance Analysis

Please assess:

  • whether the teacher tree does too much work on mount,
  • whether tree projection/filtering is still too expensive,
  • whether repeated projection or recomputation causes UI lag,
  • and whether large graph state is being recalculated unnecessarily.

6. Concrete Remediation Plan

Please provide:

  • a prioritized fix plan,
  • expected impact of each fix,
  • and which changes are safe vs. risky.

Non-Goals

Please do not spend time primarily on:

  • content quality of the curriculum data,
  • wording of nodes,
  • didactic structure decisions,
  • or broader canonical-graph modeling philosophy.

Those are not the blocker here.

The blocker is teacher UI stability.

Desired Deliverables

Please deliver:

  1. A short diagnosis summary.
  2. A state-flow diagram or equivalent explanation.
  3. A list of concrete root causes.
  4. A prioritized remediation plan.
  5. A recommendation on what should be removed or simplified.
  6. A recommendation for verification:
  7. unit tests,
  8. integration tests,
  9. manual repro checklist,
  10. optional runtime instrumentation.

Suggested Diagnostic Questions

The analysis should explicitly answer these questions:

  • Why can the teacher land in dashboard state although a class is selected?
  • Why can a selected class still end up with no visible tree?
  • Which side effect can overwrite teacher context after the user has already made a selection?
  • Which parts of teacher context are currently duplicated?
  • Which sync path should be deleted instead of repaired?
  • Is the structure-mode toggle fundamentally worth keeping in teacher view in its current form?
  • How much of the current sluggishness is due to rendering cost vs. state churn?

Practical Expectation

We are not looking for “one more patch”.

We want an expert to tell us:

  • what the stable teacher architecture should be,
  • what must be removed,
  • what should remain,
  • and in which order the teacher cockpit should be simplified and stabilized.