Skip to content

ASSESSMENT_HANDOFF.md

Latest commit

 

History

History
53 lines (40 loc) · 1.82 KB

ASSESSMENT_HANDOFF.md

File metadata and controls

53 lines (40 loc) · 1.82 KB
Raw

Assessment Handoff

Quick reviewer path for the TP Malaysia AI Intern assessment.

What To Review First

  1. Open the live app: https://tp-intern-malaysia.vercel.app/.
  2. Upload one to five receipt images.
  3. Select the files to process and click Extract.
  4. Open the preview modal to inspect the uploaded image.
  5. Expand a completed or needs-review row.
  6. For Needs review, compare the inline receipt preview with the editable fields and fix missing values.
  7. Confirm complete rows show a download icon.
  8. Download one result or selected complete results as JSON.

Commands

npm install
npm run dev
npm run ci
npm run test:e2e
npm run test:e2e:live

Environment

GEMINI_API_KEY=your_gemini_api_key_here
GEMINI_MODEL=gemini-2.5-flash

GEMINI_API_KEY is required for real extraction. GEMINI_MODEL is optional.

Assessment Signals

  • Uses a server-side API route so the Gemini key is not exposed to the browser.
  • Handles file type and file size validation on the client and API boundary.
  • Supports batch queue behavior without adding unnecessary backend complexity.
  • Treats AI extraction as a draft and keeps human review in the loop.
  • Distinguishes failed extraction from partial extraction.
  • Requires manual correction for incomplete AI fields before download.
  • Shows inline receipt preview during review to reduce user friction.
  • Validates required receipt fields before a result becomes downloadable.
  • Separates deterministic CI tests from live AI validation.
  • Documents deployment, testing, and known MVP limits.

Known Scope

The app intentionally keeps queue state in the browser and uses a single extraction API route. Durable persistence, authentication, PDF support, real Google Drive upload, URL upload, and accounting-system integrations are outside this assessment version.