Scrum4Me/scripts/README.md
Janpeter Visser eaabec8471
feat(PBI-66): wekelijkse sync van model_prices via Anthropic /v1/models (#167)
Nieuw script `npm run db:sync-model-prices` haalt de actuele Claude 4.x
modellijst op bij de Anthropic API en upsert prijzen in `model_prices`.
Anthropic biedt geen prijs-API, dus prijzen blijven onderhouden in een
PRICE_TABLE constante in het script. Cache-tier-prijzen worden afgeleid
via vaste multipliers (read 0.1x, write 1.25x). Nieuwe Claude 4.x modellen
worden gedetecteerd en gelogd als warning zodat duidelijk is wanneer de
tabel handmatig moet worden bijgewerkt.

- scripts/sync-model-prices.ts: idempotent upsert, --dry-run, retry op 5xx
- ANTHROPIC_API_KEY als optional env-var (.env.example, lib/env.ts)
- scripts/README.md: gebruiksinstructies + edge cases
- docs/plans/sync-model-prices.md: ontwerpdocument

Verificatie: `npm run lint`, `vitest` (563/563), TypeScript clean.
Echt gedraaid tegen DB: 3 created (eerste run) -> 3 unchanged (tweede run).

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-08 09:38:33 +02:00

5.2 KiB
Raw Blame History

Scripts

sync-model-prices.ts

Wekelijks handmatig draaibaar script dat de tabel model_prices synchroniseert. Haalt de actuele Claude 4.x modellijst op via GET /v1/models (Anthropic API) en upsert de prijzen vanuit een hardcoded PRICE_TABLE in het script. Anthropic biedt geen prijs-API; bij elke prijswijziging update je de tabel in scripts/sync-model-prices.ts.

Prerequisites

What How
ANTHROPIC_API_KEY in .env.local Genereer op console.anthropic.com → API Keys. Free Evaluation tier is voldoende — /v1/models is een gratis metadata-call.
DATABASE_URL in .env.local Standaard Scrum4Me-setup.

Gebruik

# Eerst droog draaien — toont wat er zou gebeuren, schrijft niets
npm run db:sync-model-prices -- --dry-run

# Echt synchroniseren
npm run db:sync-model-prices

Output

Fetching /v1/models from Anthropic API...
  → 12 models received, 4 match Claude 4.x filter

Syncing prices:
  ✓ claude-opus-4-7                   (unchanged)
  ✓ claude-sonnet-4-6                 (unchanged)
  ✓ claude-haiku-4-5-20251001         (unchanged)
  ⚠ claude-sonnet-4-9                 (geen prijs in PRICE_TABLE — ...)

Result: 0 created, 0 updated, 3 unchanged, 1 skipped

Bij een nieuw model (⚠ skipped)

  1. Open de Anthropic pricing-pagina.
  2. Voeg het model toe aan PRICE_TABLE in scripts/sync-model-prices.ts:
    'claude-sonnet-4-9': { input: 3.0, output: 15.0 },
    
  3. Draai het script opnieuw.

Edge cases

  • API geeft 401: controleer ANTHROPIC_API_KEY.
  • API geeft 5xx: script doet 1× retry met 2s delay, daarna falen.
  • Model in DB maar niet meer in API: wordt niet verwijderd — alleen gelogd, zodat oude claude_jobs rijen kostberekening blijven hebben.

test-api.sh

Runs curl-based tests against all 7 Scrum4Me API endpoints. Covers happy paths, auth (401), demo-block (403), and input validation (400).

Prerequisites

What How
Database seeded npx prisma db seed
Dev server running npm run dev
curl installed curl --version

Step 1 — Get a token for lars

  1. Open http://localhost:3000 in your browser
  2. Log in as lars / scrum4me123
  3. Go to Settings → API Tokens
  4. Click New token, give it any label
  5. Copy the token — it is shown only once

Open scripts/test-api.sh and paste it into TOKEN="".

Step 2 — Find your IDs

You need four IDs. The easiest way is to use the API itself:

# 1. Get PRODUCT_ID — run this after setting TOKEN
curl -s -H "Authorization: Bearer <TOKEN>" http://localhost:3000/api/products | grep -o '"id":"[^"]*"' | head -1

# 2. SPRINT_ID — look in the database or the UI (Sprint → copy ID from the URL)
# 3. STORY_ID  — open a story in the Sprint Board, copy the ID from the URL or the activity log
# 4. TASK_ID   — open a task, copy its ID

Or query the database directly:

# With psql / Neon CLI:
SELECT id FROM "Sprint" WHERE status = 'ACTIVE' LIMIT 1;
SELECT id FROM "Story"  WHERE status = 'IN_SPRINT' LIMIT 1;
SELECT id FROM "Task"   WHERE story_id = '<story-id>' LIMIT 1;

Paste the four values into the variables at the top of test-api.sh.

Step 3 — (Optional) Get a demo token for 403 tests

The 403 demo-block tests are skipped unless you set DEMO_TOKEN.

  1. Log out, log in as demo / demo1234
  2. Go to Settings → API Tokens, create a token
  3. Paste it into DEMO_TOKEN="" in test-api.sh

Step 4 — Run

bash scripts/test-api.sh

Expected output when all IDs are set and a demo token is provided:

============================================================
  Scrum4Me API Test Suite
  Base URL : http://localhost:3000
  Token    : scrum4me... (lars)
  Demo     : demo1234... (403 tests active)
============================================================

── GET /api/products ──────────────────────────────────────────────
  PASS  TC-P-09  happy path (lars)
  PASS  TC-P-01  no token → 401
  PASS  TC-P-02  invalid token → 401

── GET /api/products/:id/next-story ───────────────────────────────
  PASS  TC-NS-08  happy path (lars, 200 or 404 both valid)
  PASS  TC-NS-01  no token → 401
  ...

============================================================
  Results: 30 passed, 0 failed
============================================================

Notes

  • product_id is required for POST /api/todos. The Zod schema enforces z.string().min(1). Sending a todo without a product_id returns 400.
  • TC-NS-08 accepts both 200 and 404. After a fresh seed lars has no active sprint, so 404 is the expected response unless you create a sprint and add stories to it.
  • Reorder test (TC-RO-10) uses a single task ID. A reorder with one element is valid (sort_order is updated to 1).
  • The todos happy-path test appends $(date +%s) to the title to avoid duplicate-title issues on repeated runs.