koa-omnibus
Version:
Base middleware for `koa` servers.
86 lines (62 loc) • 3.94 kB
Markdown
# koa-omnibus
[](https://circleci.com/gh/hagemt/node-koa-omnibus/tree/master)
[](https://snyk.io/test/github/hagemt/node-koa-omnibus?targetFile=package.json)
[](https://codeclimate.com/github/hagemt/node-koa-omnibus/maintainability)
[](https://codeclimate.com/github/hagemt/node-koa-omnibus/test_coverage)
I wouldn't use this in production until it's more battle-tested. (at least v1)
It does exactly everything you want an API server to do, and nothing more.
Put this at the front of your `koa@^2` middleware chain and get hacking.
## What do you mean? I want to try before I buy.
Clone, then: `npm install && npm install koa && npm run demo` or:
Slap this in `node server.js | bunyan` after `npm install koa koa-omnibus`:
```
const omnibus = require('koa-omnibus')
omnibus.createApplication().listen()
// or: application.use(omnibus())
```
## Defaults not good enough? Okay...
All dependencies are included only for defaults; you can override anything you want.
```
omnibus({
// simple constants: (all Numbers are in milliseconds)
headers: { timing:String, tracing:String }, // defaults: 'X-Response-Time' and 'X-Request-ID'
limits: Object { age:Number, max:Number, next:Number, rpm:Number }, // 60000/1000000/60000/1000
namespace: String, // default: 'omnibus' (if set false-y, context.state decorated directly)
// middleware factories (unary):
limitRate: options:Object => middleware:AsyncFunction, // 429 Too Many Requests
limitTime: options:Object => middleware:AsyncFunction, // 408 Request Timeout
// redaction transforms (binary):
redactedError: (options, context) => error, // do whatever transform(s) you want here
redactedRequest: (options, context) => _.omit(context.request, ['header']), // " here
redactedResponse: (options, context) => _.omit(context.response, ['header']), // " here
// object factories (ternary):
targetError: (options, context, error) => error, // not redacted (default: Boom)
targetTracer: (options, context, string) => Object { [string]: "$(uuidgen -t)" },
targetLogger: (options, context, object) => log, // will include tracking headers
targetObject: (options, context, object) => object, // context[namespace || 'state']
})
```
## Oh, yeah. One more thing:
This doesn't include any of these modules, but you can `.use` them, as you wish:
* https://www.npmjs.com/package/koa-bodyparser
* https://www.npmjs.com/package/koa-compress
* https://www.npmjs.com/package/koa-etag
* https://www.npmjs.com/package/koa-favicon
* https://www.npmjs.com/package/koa-helmet
* https://www.npmjs.com/package/koa-logger
* https://www.npmjs.com/package/koa-mount
* https://www.npmjs.com/package/koa-ratelimit
* https://www.npmjs.com/package/koa-rewrite
* https://www.npmjs.com/package/koa-session
* https://www.npmjs.com/package/koa-static
* https://www.npmjs.com/package/@koa/cors
## But seriously, what is this, and how does it work?
You want your server to do a few things for you:
* If the user (tracks $max IPs for $age) has made Too Many Requests ($rpm) throw 429.
* If the middleware chain has been executing for too long ($next) 408 Request Timeout.
* It would be nice if my REST API actually produced error bodies for JSON consumers.
* It would be nice if my logging actually worked how I want it to. (see Bunyan)
* It would be nice if my logs included what I want, and none of what I don't.
* It would be nice if the defaults made sense, and nothing got in my way.
* It would be nice if I could later customize everything how I like it.
Sweet. So, how's it work? See `defaults.js`. Swish.