Skip to content

PRD.md

Latest commit

 

History

History
435 lines (284 loc) · 5.36 KB

PRD.md

File metadata and controls

435 lines (284 loc) · 5.36 KB

Tutor Intelligence Dashboard

Version 2.0 Multi-User SaaS-Scoped Go + Next.js + Supabase

1. Product Scope

Core objective:

Structured, time-stamped student notes AI-generated summaries Grade-based filtering Fast recall Zero unnecessary features

Non-goals remain strict: No payments No attendance automation No messaging No predictive scoring

2. System Architecture

Frontend

  • Next.js (App Router)
  • TypeScript
  • React
  • Server components for read-heavy pages
  • Client components for note input

Backend

  • Go API (Gin or Fiber)
  • Stateless REST endpoints
  • JWT validation via Supabase Auth
  • Deployed on Fly.io or similar

Database

  • Supabase Postgres
  • Row-level security enabled
  • Multi-tenant via teacher_id

AI

  • LLM API integration from backend only
  • Deterministic prompts
  • Low temperature
  • Strict output formatting

Deployment

  • Frontend on Vercel
  • Edge caching for read routes
  • Backend region aligned with Supabase

3. Multi-Tenant Model

Every domain entity includes:

teacher_id (required)

Row Level Security enforced in Supabase.

Backend verifies:

JWT teacher_id from token All queries scoped

No cross-tenant reads possible.

4. Core Domain Entities

Teacher

  • id
  • email
  • created_at

Student

  • id
  • teacher_id
  • full_name
  • current_grade
  • academic_year
  • batch_name (optional)
  • created_at

StudentNote

  • id
  • student_id
  • teacher_id
  • content
  • tag
  • created_at

WeeklySummary

  • id
  • student_id
  • teacher_id
  • week_start
  • summary_text
  • generated_at

MonthlyReport

  • id
  • student_id
  • teacher_id
  • month
  • report_text
  • generated_at

ScheduleSlot

  • id
  • teacher_id
  • day_of_week
  • start_time
  • end_time
  • batch_or_student_label

Indexes required:

student_id teacher_id created_at academic_year current_grade

5. Authentication

Supabase Auth handles:

Signup Login Password reset JWT issuance

Backend validates JWT via Supabase public key.

No custom password logic.

6. Student Import

Requirement:

CSV and Excel upload supported.

Process:

  • Upload file

  • Parse in backend

  • Validate required columns:

    • full_name
    • current_grade
    • academic_year
    • optional batch

  • Deduplicate by:

    • teacher_id + full_name + academic_year

  • Bulk insert transactionally

Error reporting:

  • Row-level validation feedback
  • Reject invalid rows
  • Allow partial success

7. Dashboard Requirements

Student Dashboard must support:

Filters:

  • Academic Year
  • Grade
  • Batch
  • Inactivity (no notes in X days)
  • Search by name

Sorting:

  • Alphabetical
  • Last note date
  • Note frequency

Displayed columns:

Name Grade Academic Year Batch Last Note Date Weekly Summary Preview

Performance requirement:

<500ms load time for 200 students

Server-rendered initial load. Edge cached if possible.

8. Student Detail Page

Must include:

Header: Name Grade Academic Year Batch

Primary action: Add note

Notes: Reverse chronological Optimistic UI Inline edit allowed within short window

Right panel:

  • Notes per week
  • Tag distribution
  • Inactivity indicator

Pagination required after 100 notes.

9. AI Requirements

Weekly Summary

Trigger: Scheduled job (weekly) Manual refresh allowed

Input: Last 7 days of notes

Constraints: 2–4 sentences Fact-based No speculation No personality labeling No invented claims

Stored deterministically by: student_id + week_start

Monthly Report

Trigger: Manual or scheduled

Structured output: Overview Strengths Areas to monitor Focus suggestions

Editable before export.

AI failure must not block UI.

10. Analytics Requirements

Derived only from notes.

Metrics: Notes per week Rolling 4-week delta Tag distribution Inactivity detection

Displayed minimally. No composite scores. No predictive flags.

11. Grade-Based Filtering

Grade is structural, not metadata.

Must support:

Viewing all Grade 6 students within academic year Viewing Grade 7 next year without data conflict

Academic year required for correct filtering.

Grade change requires update per academic year.

12. Timetable

Visual weekly grid.

No automation beyond optional local notifications.

Stored per teacher.

No attendance linkage.

13. Performance

Target:

Student page <500ms Dashboard <500ms Mobile load <1s

Strategies:

Server components for read-heavy pages Edge caching for dashboards Optimistic UI for writes Precomputed summaries Database indexes on filter fields

No live AI calls during page load.

14. State Management

Minimal global state.

Per-page data fetching.

Optimistic note append.

No Redux unless future complexity demands it.

15. Security

Row-level security enforced.

All queries require teacher_id.

Rate limit login attempts handled by Supabase.

AI requests contain only teacher-owned data.

Full export capability required.

16. Testing Requirements

Unit tests: Domain logic AI input validation Grade filter correctness

Integration tests: Note → summary pipeline CSV import edge cases

E2E: Signup Import Add note Generate summary Filter by grade

CI runs on every PR.

17. Deployment & Optimization

Frontend: Vercel Edge caching for GET routes

Backend: Low-latency region near Supabase

Database: Connection pooling Prepared statements

Zero unnecessary round-trips.

18. Explicit Constraints

System must remain:

Note-first AI-secondary Operationally simple Low visual noise

If a feature increases cognitive load without improving recall speed, reject it.