Fintech Development

Currency Conversion Service Prompt

Build a currency conversion service that handles exchange rate staleness, source selection, and rounding correctly across currency pairs.

Anselm Fowel / Public collection27 Sept 2026

Your prompt

The prompt

You are a fintech engineer who has built currency conversion for a cross-border payments product and has seen the specific ways a converted amount can be wrong even when the exchange rate itself was correct at some point in time.

I'm building a conversion service in [LANGUAGE/FRAMEWORK] that converts between [CURRENCY PAIRS, e.g., USD/EUR/GBP/JPY] for [USE CASE: CROSS-BORDER PAYMENTS / MULTI-CURRENCY WALLET / MARKETPLACE PAYOUTS], sourcing rates from [RATE PROVIDER: A THIRD-PARTY FX API / A BANKING PARTNER].

First, design the rate-fetching and caching strategy: how frequently rates are refreshed from [RATE PROVIDER], how long a cached rate is considered valid before it's treated as stale, and what the service does when a conversion request comes in and the cached rate has expired but the provider is slow or down, fail the request, or fall back to a stale rate within a defined tolerance, and why.

Next, design how the rate actually used for a specific conversion is recorded and locked in: whether you quote a rate that's guaranteed for some time window before the customer confirms, how you prevent a rate from changing between when it's shown to a user and when the conversion is executed, and what happens if it's expired by execution time.

Then address rounding and precision specifically for cross-currency math: converting an integer-minor-unit amount in one currency to another using a rate with many decimal places, choosing a consistent rounding rule, and handling currencies with different minor-unit conventions (JPY has none, most have 2, some like BHD have 3). Also design how the service handles markup or spread on top of the raw market rate, and how that's made transparent and auditable rather than baked invisibly into the displayed rate.

Finally, output the core conversion function in [LANGUAGE/FRAMEWORK] with comments covering rate lookup, staleness check, and rounding, followed by a table of edge cases (stale rate at execution time, zero-decimal currency, provider outage, rate change mid-transaction) mapped to expected behavior.

2,072 characters

Your prompt

Recent articles