Plan caching for a slow endpoint

Prompt that plans caching for a slow endpoint: what may be cached, where the cache sits and how stale entries are removed. Paste it into an AI tool of your choice, which then plans caching.

Topic: Apps & prototypes
Type: Prompt
Level: Intermediate
Time: about 8 min

Results depend on the tool, the model and what you enter.

Result preview

Example: written by our editors, not real AI output.

Root cause first: the query has no index on "category"; that alone will probably save a lot of time. Cache: application, key product_list:{category}:{filter}:{page}, valid for 5 minutes Never cached: prices with customer discounts, basket Invalidation: when a product is saved, delete all keys for its category Concurrent requests: only one recomputes, the others briefly get the old entry Measure: hit rate, response time at peak times Assumption: stock levels may be up to 5 minutes old.

Fill in

What you type stays in this browser window. We store and send nothing.

EndpointWhich endpoint, what does it return?

ChangesHow often does the data change?

ProblemWhat is slow today?

Your prompt

Plan caching for the following slow endpoint.
Endpoint and what it returns: [Endpoint]
How often the data changes: [Changes]
What is slow today: [Problem]
Approach: 1. First check whether caching fits here at all or whether the cause lies elsewhere, such as a missing database index. 2. Decide what is cached and what never is, for example account-specific data without a matching key. 3. Choose the location, such as browser, CDN, application or close to the database, and justify the choice. 4. Define cache keys, lifetime and how entries are removed when data changes. 5. Describe what happens when many requests hit an expired entry at the same moment, and how you prevent it. 6. Name the metrics you will use to measure success, such as hit rate and response time. Do not invent measurements, and mark assumptions with "Assumption:".

Take particular care to check the real cause first and never cache account-specific data without a matching key.

Before you answer, check: Are all instructions met? If key information is missing or a placeholder in square brackets is still in the text, ask instead of guessing. Mark assumptions as assumptions.

Plan caching for the following slow endpoint.
Endpoint and what it returns: [Endpoint]
How often the data changes: [Changes]
What is slow today: [Problem]
Approach: 1. First check whether caching fits here at all or whether the cause lies elsewhere, such as a missing database index. 2. Decide what is cached and what never is, for example account-specific data without a matching key. 3. Choose the location, such as browser, CDN, application or close to the database, and justify the choice. 4. Define cache keys, lifetime and how entries are removed when data changes. 5. Describe what happens when many requests hit an expired entry at the same moment, and how you prevent it. 6. Name the metrics you will use to measure success, such as hit rate and response time. Do not invent measurements, and mark assumptions with "Assumption:".

Take particular care to check the real cause first and never cache account-specific data without a matching key.

Before you answer, check: Are all instructions met? If key information is missing or a placeholder in square brackets is still in the text, ask instead of guessing. Mark assumptions as assumptions.

This is an instruction for an AI, not the result itself.

Where this prompt works

These AI tools are a good fit for this prompt:

How to use it

Fill in the fields or tap a suggestion. Leave out names, addresses and other personal data.
Copy the finished prompt into an AI tool of your choice.
Introduce caching for a single endpoint first, measure hit rate and response time for a week and only then extend it.

Why does this work?

The question of what must never be cached comes before the technology, so that nobody receives another user's data. A clear way to remove entries prevents a cache that is faster but wrong.

Tips

Even a cache of a few seconds often absorbs load peaks well and keeps the risk of stale data small.

You may use our prompts freely, including at work. Please do not republish them as your own collection. We give no warranty for AI results; check them yourself. Terms of use

Plan caching for a slow endpoint

Result preview

Fill in

Your prompt

Where this prompt works

How to use it

Why does this work?

Tips

Also a good fit

Choose pagination for an API list New

Write an API reference people can use right away New

Plan a throttled data backfill New