# Integrated Algebra platform audit and implementation plan

Checked November 4, 2026. The existing implementation provides 185 canonical curriculum lessons, 740 MCQs, 185 written prompts, 56 mapped TEKS, account-owned attempts, and server-side grading. It does not yet provide a separate STAAR bank, persisted examinations, a content editor, teacher mark overrides, time-on-task, or adaptive study plans.

Reuse Next 16 App Router, mysql2 pool, JWT account cookies, `users`, existing parent/student ownership checks, admin access rules, Algebra dashboards and brand components. Keep scheduled classroom records intact. Extend public self-study through normalized curriculum, question-bank, session-response, activity, evaluation-review, and study-plan tables. Source content seeds these tables idempotently; drafts never overwrite published snapshots. Examination responses snapshot their original questions so later edits cannot change an active exam's key.

## Verified assessment evidence

- Current TEA mathematics index: https://tea.texas.gov/data-reports/staar/staar-mathematics-resources
- Linked blueprint: https://tea.texas.gov/data-reports/staar/2022-staar-algebra-i-blueprint-0.pdf — 50 items, 59 points, 41 one-point and nine two-point items; five category ranges. Counts describe the scored blueprint, not extra field-test items.
- Latest released form examined: https://tea.texas.gov/data-reports/staar/released-test-questions/2026-staar-algebra-1-answer-key.pdf — includes multiple choice, multiselect, graphing, drag-and-drop, inline choice, and match-table-grid. Format frequencies in a single release are not a mandated distribution for future tests.
- https://teastudentassessments.zendesk.com/hc/en-us/articles/360040794571-Is-there-a-time-limit-for-STAAR-assessments — schools schedule three to four hours, with additional time within the school day up to seven hours. Practice timer defaults are internal settings, not an assertion that every student receives an identical fixed official duration.
- Algebra I does not have a verified prose constructed-response item in the examined release. AI review of working is supplemental learning feedback and never replaces deterministic item scoring or changes a mock score.

## Implementation phases

1. Normalize and seed existing curriculum/TEKS and a separate original STAAR bank; retain current URLs and answer privacy.
2. Add topic, category, mixed, weak-area and configurable blueprint simulations. Persist autosaves, server deadlines, flags and results with transactions and revision checks.
3. Extend structured AI evaluation with per-step evidence, configurable rubrics, uncertainty flags and immutable original evaluations plus teacher review records.
4. Add activity-based lesson completion, separate performance summaries, revision recommendations, adjustable 7/14/30-day schedules and a unified learning dashboard accessible from the existing dashboards.
5. Add existing-admin-authorized editing, draft/publish validation, import/export, exam settings and review queues. Test numerical grading, partial credit, timer expiry, ownership, stale saves, idempotency, content validation and legacy regression.

Live OpenAI remains subject to a valid configured server key; the previous request returned `invalid_api_key`. No production deployment or browser verification is implied by server tests.