Skip to main content

System Requirements

Functional Requirements​

Transaction Ingestion​

  • Upload CSV files
  • Validate and normalize input
  • Idempotent ingestion (no duplicates on re-upload)

Designed-for (not implemented)​

  • Bank integrations via adapters
  • Async processing for large files

Acceptance criteria​

  • 10k transactions ingested in <5 seconds
  • Partial failures are reported clearly

Transaction Management​

  • Users must view a scrollable list of all transactions.
  • Users must be able to manually edit the "Category" of any transaction.
  • Changing a category must update the spending report/charts without a page reload.

Smart Categorization (AI)​

  • System must attempt to categorize imported transactions automatically.
  • Immediate categorization based on keywords (e.g., "Uber" -> "Transport").
  • Ability to mark complex/unknown transactions for "AI Analysis" (mocked latency).

Reporting & Analytics​

  • Spending Breakdown: Visual pie/bar chart showing total spend per category.
  • Charts must reflect the current state of transactions instantly (including optimistic updates).

Non-Functional Requirements​

Performance & Responsiveness​

  • UI interaction latency: <100–200ms
  • 10k Transaction Support: The main dashboard and list must render and scroll smoothly (60fps) with 10,000 items loaded.
  • Latency: Category updates must be reflected in the UI within 100ms (Optimistic UI).

Usability & Experience​

  • Design must use modern best practices.
  • Micro-animations, dark mode and accessibility considerations.
  • System must handle simulated "slow network" states without blocking the UI, displaying in-progress indicators as needed and allowing continued interaction.
  • Globalization support.
  • UI is responsiveness to screen size changes, but the mobile experience is best served by a dedicated native app.

Privacy & Security​

  • Data is encrypyted at rest and in transit.
  • No sensitive data is logged.
  • Mock backend does not persist data beyond session.

Reliability​

  • Use of Error Boundaries to catch React rendering errors.
  • Typo-tolerance in CSV parsing (basic validation).

Observability​

  • Metrics for:
    • Ingestion latency
    • Categorization accuracy
    • Override frequency
  • Logs and traces across services
  • Feedback loop for AI-native categorization evaluation and training

Cost​

  • AI inference is selectively applied.
  • Tiered service levels for heavy users

Scalability​

  • Horizontal scaling of microservices
  • Load balancing
  • Caching of predictions