Next roadmap after production hardening

[Farukhsb]

Summary

The main production-hardening sprint has been completed and merged. This issue tracks what is left so the project has a clear next roadmap rather than starting random work.

Completed baseline

GradeAI now has:

• AI grading response validation
• grade breakdown validation
• explanation/tutoring response validation
• integrity/plagiarism response validation
• rate limiting for high-cost Edge Functions
• environment variable validation
• structured logging with sensitive-context sanitisation
• critical workflow type-safety cleanup
• error boundary and network failure coverage
• student visibility E2E coverage
• updated README hardening summary
• updated trust model documentation
• production hardening summary document
• green GitHub Actions after merge

What is left

1. Evidence packaging

Prepare the hardening work as a clear portfolio/GTV evidence item.

Evidence to collect:

• green GitHub Actions screenshot
• merged PR screenshots
• README.md recent hardening section
• docs/TRUST_MODEL.md
• docs/GRADEAI_PRODUCTION_HARDENING_SUMMARY.md
• screenshots of key workflows if needed

Suggested output:

• one-page evidence summary
• short technical narrative
• before/after hardening table

2. More E2E trust workflows

Add browser-level coverage for remaining high-value workflows.

Priority flows:

• external examiner export governance
• moderation gating before release
• lecturer review -> approve -> release status transitions
• integrity review decision persistence

3. Gradual TypeScript strictness

Do not enable full strict mode in one jump.

Next steps:

• continue reducing remaining low-risk any
• enable selected TypeScript/ESLint rules as warnings first
• focus on academic workflow files before UI-only files
• use Zod-inferred types where possible

4. RLS and live permission verification

Add deeper checks for hosted environment permissions.

Focus areas:

• student cannot access other student data
• moderator sees assigned moderation records only
• admin read-only oversight behaves as intended
• lecturer visibility remains scoped to their assignments/students

5. Migration history hygiene

The README already documents legacy migration metadata. Before a wider rollout, create a clear reconciliation plan.

Tasks:

• document current migration ledger state
• decide whether reconciliation is needed
• avoid renaming historical migration IDs without a deliberate plan

6. Operational runbook

Create a small deployment/incident runbook.

Include:

• deploy frontend
• deploy Edge Functions
• required environment variables
• Supabase migration checks
• what to check when grading/plagiarism/explain-grade fails
• how to interpret rate limit responses
• how to check Sentry/logging without exposing student data

Recommended next task

Start with evidence packaging before adding more code.

Suggested next document:

docs/GTV_EVIDENCE_SUMMARY_GRADEAI_HARDENING.md

Purpose:

• explain what was built
• explain the risks addressed
• explain the engineering decisions
• list proof points and files
• make the work understandable to non-technical reviewers

Suggested order

1. Evidence packaging
2. External examiner export E2E
3. Moderation gating E2E
4. Gradual TypeScript cleanup
5. RLS/live permission verification
6. Operational runbook

[Farukhsb]

Roadmap status update

Items 4, 5, and 6 are now complete for the current baseline.

• RLS / permission verification has been completed for students, moderators, admins, and lecturers.
• Migration history hygiene has been resolved for the current baseline because a new database has been created. The reconciliation plan remains documented in docs/SUPABASE_MIGRATION_HISTORY_RECONCILIATION_PLAN.md for audit/history.
• Operational runbook has been added in docs/OPERATIONAL_RUNBOOK.md.

The runbook covers deployment, Edge Functions, environment variables/secrets, Supabase migration checks, grading/integrity/explain-grade troubleshooting, rate limits, logging/Sentry safety, permission re-checks, and incident handling.

Remaining before this roadmap issue can be closed:

1. More E2E trust workflows, unless split into separate follow-up issues.
2. Gradual TypeScript strictness, unless split into a separate follow-up issue.