Tired of not owning the authentication flow of your SaaS?
openkey is a ~200 lines backed in Redis for control authentication flow. You can deploy to any cloud provider, no vendor lock-in, being cheap at any scale & focused in performance to authenticate requests.
First, Install openkey from your preferred node package manager:
pnpm install openkey
openkey is also available as CLI when you install it globally in your system:
npm install -g openkey
After that, you can access to any command from your terminal:
❯ openkey
openkey> help
version exit keys plans usage stats
openkey> version
1.0.0
openkey> exit
After installation, initialize openkey:
const Redis = require('ioredis')
const redis = new Redis()
const openkey = require('openkey')({ redis })
you can prepend all the keys by passing prefix
:
const openkey = require('openkey')({ redis, prefix: 'http:' })
These will allow you to acess to openkey core concepts: plans, keys, usage, and stats.
It represents the quota, rate limit, and throttle information specified as a plan.
It creates a new plan:
const plan = await openkey.plans.create({
name: 'free tier',
metadata: { tier: 'free' },
limit: 3000,
period: '1d'
})
The options accepted are:
id
string: The id of the plan, it cannot contain whitespaces.period
string: The time window which the limit applies. It accepts ms syntax.limit
number: The target maximum number of requests that can be made in a given time period.metadata
object: A flat object containing additional information. Passnull
or''
to remove all the metadata fields.
Any other field provided will be omitted.
Returns: an object with the options specified, plus:
createdAt
number: The timestamp when the object was created.updatedAt
number: The last timestamp when the object was modified.
It retrieves all the plans:
const plans = await openkey.plans.list()
It retrieves a plan by id:
const { createdAt, updatedAt, ...plan } = await openkey.plans.retrieve('free_tier')
Returns: the plan
object, or null
if it is not found.
It updates a plan by id:
const { updatedAt, ...plan } = await openkey.plans.update('free_tier', {
limit: 1000
})
You can't update the id
. Also, in the same way than .create, any other field that is not a supported option will be omitted.
Returns: the updated plan
object. If the plan is not found, this method will throw an error.
It deletes a plan by id:
await openkey.plans.del('free_tier')
It will throw an error if:
- The plan has key associated. In that case, first keys needs to be deleted.
- The plan id doesn't exist.
Returns A boolean confirming the plan has been deleted.
It represents the credentials used for authenticating a plan.
It creates a new key:
/*
* A random 16 length base58 key is created by default.
*/
const key = await openkey.key.create()
console.log(key.value) // => 'oKLJkVqqG2zExUYD'
/**
* You can provide a value to use.
*/
const key = await openkey.key.create({ value: 'oKLJkVqqG2zExUYD' })
/**
* The key can be associated with a plan when it's created.
*/
const key = await openkey.key.create({ value: 'oKLJkVqqG2zExUYD', plan: plan.id })
The options accepted are:
value
string: The value of the key, being a base58 16 length key generated by default.enabled
string: It determines if the key is active, beingtrue
by default.metadata
object: A flat object containing additional information. Passnull
or''
to remove all the metadata fields.
Any other field provided will be omitted.
Returns: an object with the options specified, plus:
createdAt
number: The timestamp when the object was created.updatedAt
number: The last timestamp when the object was modified.
It retrieves a key by id:
const { createdAt, updatedAt, ...key } = await openkey.key.retrieve('AN4fJ')
Returns: the key
object, or null
if it is not found.
It updates a key by id:
const { updatedAt, ...key } = await openkey.key.update(value, {
enabled: false
})
In the same way than .create, any other field that is not a supported option will be omitted.
Returns: the updated key
object. If the key is not found, this method will throw an error.
It deletes a key by value:
await openkey.plans.del(value)
It will throw an error if the key value doesn't exist.
Returns A boolean confirming the plan has been deleted.
It returns the current usage of a key that is associated with a plan:
const usage = await openkey.usage(key.value)
console.log(usage)
// {
// limit: 3,
// remaining: 3,
// reset: 1714571966087,
// pending: Promise { [] }
// }
Similar to the previous method, but increments the usage by one before returning:
const usage = await openkey.usage.increment(key.value)
// {
// limit: 3,
// remaining: 2,
// reset: 1714571966087,
// pending: Promise { [] }
// }
Additionally you can increment specifying the quantity
:
const usage = await openkey.usage.increment(key.value, { quantity: 3 })
// {
// limit: 3,
// remaining: 0,
// reset: 1714571966087,
// pending: Promise { [] }
// }
It returns the count per every day for a given API key:
const stats = await openkey.stats(key.value)
console.log(stats)
// [
// { date: '2024-05-01', count: 1 },
// { date: '2024-05-02', count: 10 },
// { date: '2024-05-03', count: 5 }
// ]
By default, openkey uses JSON serialization without compression for two reasons:
- The payload isn't large enough to take advantage of compression.
- Storing compressed data makes the content unreadable without first decompressing it.
You can customize serialize
and deserialize
when openkey is instantiated to define how you want your data to be handled.
For example, you can combine openkey with compress-brotli to store compressed data painlessly:
const compressBrotli = require('compress-brotli')
const redis = new Redis()
const openkey = require('openkey')({
redis,
serialize: async data => brotli.serialize(await brotli.compress(data)),
deserialize: data => brotli.decompress(brotli.deserialize(data))
})
openkey has been designed to play well according to RateLimit header fields for HTTP:
module.exports = (req, res) => {
const apiKey = req.headers['x-api-key']
if (!apiKey) send(res, 401)
const { pending, ...usage } = await openkey.usage.increment(apiKey)
const statusCode = usage.remaining > 0 ? 200 : 429
res.setHeader('X-Rate-Limit-Limit', usage.limit)
res.setHeader('X-Rate-Limit-Remaining', usage.remaining)
res.setHeader('X-Rate-Limit-Reset', usage.reset)
return send(res, statusCode, usage)
}
openkey is about making pricing a part of your product development.
It's an excellent idea to combine it with Stripe:
// https://docs.stripe.com/billing/subscriptions/usage-based/implementation-guide
// Set your secret key. Remember to switch to your live secret key in production.
// See your keys here: https://dashboard.stripe.com/apikeys
const stripe = require('stripe')('sk_test_VZqeYMqkpa1bMxXyikghdPCu');
const count = await openkey.usage.get('{{CUSTOMER_KEY}}')
const meterEvent = await stripe.billing.meterEvents.create({
event_name: 'alpaca_ai_tokens',
payload: {
value: count,
stripe_customer_id: '{{CUSTOMER_ID}}',
},
});
Read more about Usage-based billing at Stripe.
Every possible error thrown by openkey has the name OpenKeyError
unique code
associated with it.
if (error.name === 'OpenKeyError') {
return send(res, 400, { code: error.code, message: error.message })
} else {
return send(res, 500)
}
This makes it easier to apply any kind of handling in your application logic.
You can find the list errors in the source code.
Mainly because it's a cheap in-memory database at scale, and mature enough to prevent vendor lock-in.
We considered other alternatives such as SQLite, but according to these requeriments Redis is a no brain choice.
This library is intended to be used millions of times every day. We wanted to have granular control as much as possible, and adding a TypeScript transpilation layer isn't ideal from a performance and maintenance perspective.
Originally this library was implemented using hashes, but then since values are stored as string, it's necessary to cast value (for example, from string to number).
Since we need to do that all the time, we prefer to use key/value. Also this approach allow to customize serializer/deserializer, which is JSON by default.
No, writes operatoins are not atomic because there are very few use cases where that matters. openkey is designed to process a constant stream of requests, where the only thing important to control reaching the limit of each plan.
In case you need it, you can combine openkey with superlock, check the following example.
openkey © microlink.io, released under the MIT License.
Authored and maintained by microlink.io with help from contributors.
microlink.io · GitHub microlink.io · X @microlinkhq