mirror/insomnia
1
0
Fork 0
mirror of https://github.com/Kong/insomnia.git synced 2026-04-20 06:07:24 -04:00
Code Issues Packages Projects Releases Wiki Activity
Files
lib@2.3.0-beta.4
insomnia/packages/openapi-2-kong
History

README.md

openapi-2-kong

This module generates Kong Declarative Config and Kong for Kubernetes config, from OpenAPI 3.0 specifications.

Library Usage

This module exposes three methods of generating Kong declarative config.

export type ConversionResultType = 'kong-declarative-config' | 'kong-for-kubernetes';

export interface DeclarativeConfigResult {
  type: 'kong-declarative-config';
  label: string;
  documents: DeclarativeConfig[];
  warnings: Warning[];
}

export interface KongForKubernetesResult {
  type: 'kong-for-kubernetes';
  label: string;
  documents: K8sManifest[];
  warnings: Warning[];
}

export type ConversionResult = DeclarativeConfigResult | KongForKubernetesResult;

generateFromSpec(
  spec: OpenApi3Spec,
  type: ConversionResultType,
  tags?: string[],
) => ConversionResult

generateFromString(
  spec: string,
  type: ConversionResultType,
  tags?: string[],
) => Promise<ConversionResult>

generate(
  filePath: string,
  type: ConversionResultType,
  tags?: string[],
) => Promise<ConversionResult>

Usage Example

import {
  generateFromString,
  generateFromSpec,
  generate,
} from 'openapi-2-kong';
import YAML from 'yaml';
import { writeFileSync } from 'fs';

const spec = `
openapi: "3.0.0"
info:
  version: 1.0.0
  title: Swagger Petstore
servers:
  - url: http://petstore.swagger.io/v1
paths:
  /pets:
    get:
      summary: Get all pets
`;

const examples = async () => {
  const tags = [ 'MyTag' ];
  const type: ConversionResultType = 'kong-declarative-config'; // or 'kong-for-kubernetes'

  // Generate a config from YAML string
  const config1 = await generateFromString(spec, type, tags);

  // Generate a config from a JavaScript Object
  const specObject = YAML.parse(spec);
  const config2 = generateFromSpec(specObject, type, tags);

  // Generate a config from a JSON string
  const specJSON = JSON.stringify(specObject);
  const config3 = await generateFromString(specJSON, type, tags);

  // generate a config from a file path
  writeFileSync('/tmp/spec.yaml', spec);
  const config4 = await generate('/tmp/spec.yaml', type, tags);

  console.log('Generated:', { config1, config2, config3, config4 });
};

Kong Declarative Config

The following documents the behavior for generating Kong Declarative Config from OpenAPI v3.

$._format_version

This attribute is hardcoded to 1.1 since that is currently the only version.

$.services

Kong services objects are generated mostly from the root servers property.

servers:
  - url: http://petstore.swagger.io/v1

services:
  - host: Simple_API_overview  # Upstream name (see below)
    port: 80                   # Port inferred from protocol if not specified
    path: "/v1"                # Extracted from URL of the first `server`
    protocol: http             # Extracted from URL of the first `server`, or defaulted to http
    name: Simple_API_overview  # Service name (see below)
    routes: []                 # <documented later>
    tags: []                   # <documented later>

Or, if variables are used, their default values will be substituted. The following example will produce the same result

servers:
  - url: https://{subdomain}.swagger.io/v1
    variables:
      subdomain:
        default: petstore
        enum:
          - petstore
          - shoestore

The service name is set to the following

  1. Root x-kong-name attribute
  2. Generated slug from $.info.title
  3. Default to openapi as a last resort

$.upstreams

Upstreams and targets are generated from the servers root property (note that servers defined on operation or path objects are ignored).

One OpenAPI spec will result in one service and one upstream, and each individual server in the root servers property will become a target in the upstream.

servers:
  - url: http://petstore.swagger.io/v1
  - url: https://swagger.io/v1

upstreams:
  - name: Simple_API_overview           # Same as the service name
    targets:
      - target: petstore.swagger.io:80  # Derived from first server entry
      - target: swagger.io:443          # Derived from second server entry
    tags: []                            # <documented later>

$.services[*].routes

Service routes are generated from the root paths property. One service route per path/method combination will be generated.

paths:
  /pets/{id}:
    x-kong-name: create-pet
    put:
      summary: List all pets
      tags: [ Tag ]
      responses: [ ... ]
    get:
      tags: [ Tag ]
      responses: [ ... ]

services:
  - ...
    routes:
      - name: ApiName-create-pet         # Taken from x-kong-name, summary, or generated
        strip_path: false                # Defaults to `false`
        methods: [ PUT ]                 # Only ever a single-entry array
        paths: [ '\/pets/(?<id>\S+)$' ]  # Kong regex-formatted path with variables
        tags: [ Tag ]                    # <documented later>

Route names are constructed from the template <APIName>-<PathName>-<Method>.

  • APIName: Name taken from the global API object, prefixed to ensure uniqueness across services
  • Method: Route's HTTP method
  • PathName: Pulled from x-kong-name, operationId, or a path slug

$..tags

Each generated entity will get the tags as specified as well as the following tags:

  • OAS3_import
  • OAS3file_<filename>

Tags can also be passed to this tool, which will be appended to the existing tags of all created resources.

Kong for Kubernetes

Output structure

The Kong for Kubernetes config will contain at least one Ingress document per server. Depending on where Kong plugins (x-kong-plugin-<plugin-name>) exist in the specification, several Ingress documents may be created, in addition to KongPlugin and KongIngress documents.

KongPlugin and KongIngress resource documents can be reused, and are applied to the specs in an Ingress document via metadata annotations.

The Ingress document

At least one Ingress document will be generated for each server. How many are required for each server, will be determined by the presence of KongPlugin and KongIngress resources that need to be applied to specific paths.

Example

Source spec

openapi: 3.0.0
info:
  title: Insomnia API
servers:
  - url: http://one.insomnia.rest/v1
  - url: http://two.insomnia.rest/v2
paths:

Generated config

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: insomnia-api-1
  annotations:
    kubernetes.io/ingress.class: "kong"
spec:
  rules:
    - host: one.insomnia.rest
      http:
        paths:
          - path: /v1/.*
            backend:
              serviceName: insomnia-api-service-0
              servicePort: 80
---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: insomnia-api-2
  annotations:
    kubernetes.io/ingress.class: "kong"
spec:
  rules:
    - host: two.insomnia.rest
      http:
        paths:
          - path: /v2/.*
            backend:
              serviceName: insomnia-api-service-1
              servicePort: 80

$.metadata.name

The Ingress document metadata.name is derived from sections in the source specification. If several are present, then the highest priority is selected. If none are present, then the fallback name is openapi. The name is also converted to a lowercase slug.

Each of the following specifications generate an Ingress document with the following name:

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: insomnia-api
...

In priority order, these sections are:

  • $.info.x-kubernetes-ingress-metadata.name

    openapi: 3.0.0
info:
  x-kubernetes-ingress-metadata:
    name: Insomnia API
...

  • $.x-kong-name

    openapi: 3.0.0
x-kong-name: Insomnia API
...

  • $.info.title

    openapi: 3.0.0
info:
  title: Insomnia API
...

$.metadata.annotations

Annotations can be used to configure an Ingress document. This configuration applies to that entire Ingress document.

  1. Any annotations that exist under $.info.x-kubernetes-ingress-metadata.annotations in the source specification are added automatically
  2. KongPlugin and KongIngress resources may be generated. These are added to the Ingress document as:
    • konghq.com/plugins references multiple KongPlugin resources via a comma separated string containing resource names
    • konghq.com/override references single KongIngress resource via the resource name

The KongPlugin and KongIngress resources

If plugins are found in the OpenAPI spec (see Plugins), then they are converted to KongPlugin resources by the Kong for Kubernetes config generator, before being applied to the Ingress document via metadata annotations.

If a plugin is found on an operation object in a path in the OpenAPI spec, then an additional KongIngress resource is generated, that refers to the operation (GET, POST, etc).

Example

This example combines all of the information in the previous sections. The input specification contains:

  • x-kubernetes-ingress-metadata to provide a document name and default annotations
  • a global plugin at the OpenAPI spec level
  • two servers, with s1 containing a plugin
  • two paths with one operation
    • no plugins on the first path and operation
    • no plugin on the second path, but one on the operation
Input OpenAPI specification 
openapi: 3.0.0
info:
  x-kubernetes-ingress-metadata:
    name: insomnia-api                          # Kubernetes Ingress document name
    annotations:                                # Annotations to automatically apply
      'example': 'example'
x-kong-plugin-custom-global:                    # Global plugin to always be applied
  name: custom-global
  enabled: true
  config:
    key_in_body: false
servers:
  - url: http://one.insomnia.rest/v1
  - url: http://two.insomnia.rest/v2
    x-kong-plugin-custom-server:                # Server plugin to be applied to documents for this server
      name: custom-server
      enabled: true
paths:
  '/path':
    post:
      summary: Some functionality               # A path with no plugins on the path or operations in it
  '/another-path':
    get:
      summary: Some functionality
      x-kong-plugin-key-auth:                   # Operation plugin on GET /another-path
        name: key-auth
        enabled: true
        config:
          key_names: [api_key, apikey]
          key_in_body: false
          hide_credentials: true

The resultant spec creates:

  • four unique Ingress documents (two servers, two paths, one operation each),
  • three KongPlugin documents (global, server, operation plugin),
  • one KongIngress document (GET operation containing plugin)
Output Kong for Kubernetes config 
apiVersion: configuration.konghq.com/v1
kind: KongIngress
metadata:
  name: get-method                              # KongIngress due to GET /another-path containing a plugin
route:
  methods:
    - get
---
apiVersion: configuration.konghq.com/v1
kind: KongPlugin
metadata:
  name: add-custom-global-g0                    # KongPlugin due to plugin at OpenAPI object level
plugin: custom-global
config:
  key_in_body: false
---
apiVersion: configuration.konghq.com/v1
kind: KongPlugin
metadata:
  name: add-custom-server-s1                    # KongPlugin due to plugin on the second server
plugin: custom-server
---
apiVersion: configuration.konghq.com/v1
kind: KongPlugin
metadata:
  name: add-key-auth-m2                         # KongPlugin due to plugin on GET /another-path
plugin: key-auth
config:
  key_names:
    - api_key
    - apikey
  key_in_body: false
  hide_credentials: true
---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: insomnia-api-0
  annotations:
    kubernetes.io/ingress.class: "kong"
    example: example                            # annotation from x-kong-ingress-metadata
    konghq.com/plugins: add-custom-global-g0    # only global plugin
spec:
  rules:
    - host: one.insomnia.rest
      http:
        paths:
          - path: /v1/path
            backend:
              serviceName: insomnia-api-service-0
              servicePort: 80
---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: insomnia-api-1
  annotations:
    kubernetes.io/ingress.class: "kong"
    example: example
    konghq.com/plugins: add-custom-global-g0, add-key-auth-m2   # global and operation plugin, no server or path
    konghq.com/override: get-method             # restrict document to be for specified operation (due to plugin)
spec:
  rules:
    - host: one.insomnia.rest
      http:
        paths:
          - path: /v1/another
            backend:
              serviceName: insomnia-api-service-0
              servicePort: 80
---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: insomnia-api-2
  annotations:
    kubernetes.io/ingress.class: "kong"
    example: example
    konghq.com/plugins: add-custom-global-g0, add-custom-server-s1    # global and server 1 plugin, no path or operation
spec:
  rules:
    - host: two.insomnia.rest
      http:
        paths:
          - path: /v2/path
            backend:
              serviceName: insomnia-api-service-1
              servicePort: 80
---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: insomnia-api-3
  annotations:
    kubernetes.io/ingress.class: "kong"
    example: example
    konghq.com/plugins: add-custom-global-g0, add-custom-server-s1, add-key-auth-m2   # global, server 1, and operation plugin, no path
    konghq.com/override: get-method             # restrict document to be for specified operation (due to plugin)
spec:
  rules:
    - host: two.insomnia.rest
      http:
        paths:
          - path: /v2/another
            backend:
              serviceName: insomnia-api-service-1
              servicePort: 80

Defaults

While properties for the generated entities will have properties derived from the OpenAPI spec, you may choose to specify defaults for certain properties. You can specify these defaults using the following keys:

  • x-kong-service-defaults applied to the root object
  • x-kong-upstream-defaults applied to the root object
  • x-kong-route-defaults applied to the root, path or operation object

These defaults are only supported by the Declarative Config generator currently.

Plugins

Security Plugins

The security property can be defined on the top-level openapi object as well as on operation objects. The information contained cannot be directly mapped onto Kong, due to the logical and/or nature of how the specs have been set up.

To overcome this Kong will only accept a single securityScheme from the security property.

The additional properties that Kong supports on its plugins can be configured by using custom extensions. The custom extensions are x-kong-security-<plugin-name>.

Supported types are:

  • oauth2
    • NOT YET IMPLEMENTED!
    • except for the implicit flow
    • implemented using the Kong plugin openid-connect
    • extended by: x-kong-security-openid-connect
  • openIdConnect
    • implemented using the Kong plugin openid-connect
    • extended by: x-kong-security-openid-connect
    • properties set from OpenAPI spec:
      • issuer (from openIdConnectUrl property)
      • scopes_required will get the combined set of scopes from the extension defaults and the scopes from the Security Requirement Object
  • apiKey
    • except for the in property, since the Kong plugin will by default look in header and query already. Cookie is not supported.
    • implemented using the Kong plugin key-auth
    • extended by: x-kong-security-key-auth
    • properties set from OpenAPI spec:
      • key_names will get the defaults from the extension and then the name from the securityScheme object will be added to that list
    • requires to add credentials to Kong, which is not supported through OpenAPI specs.
  • http
    • only Basic scheme is supported
    • implemented using the Kong plugin basic-auth
    • extended by: x-kong-security-basic-auth
    • properties set from OpenAPI spec:
      • none
    • requires to add credentials to Kong, which is not supported through OpenAPI specs.

Generic Plugins

Generic plugins can be added on various objects in the OpenAPI spec. The custom extension to use is x-kong-plugin-<plugin-name>. The name property is not required (since it's already in the extension name). Optional properties not specified will get Kong defaults.

Currently, plugins are supported on the following OpenAPI objects by each generator:

  • Declarative Config: OpenAPI root, path and operation
    • plugins on the root will be configured on a Kong service
    • plugins on the path or operation will be configured on a Kong route
  • Kong for Kubernetes: OpenAPI root, server, path, operation

If the same plugin exists in several sections, then the more specific section will take precedence. These sections are:

  1. operation (most specific)
  2. path
  3. server
  4. OpenAPI object (least specific)

This extension needs to hold an object that contains the entire plugin config.

x-kong-plugin-key-auth:
  name: key-auth
  enabled: true
  config:
    key_names: [api_key, apikey]
    key_in_body: false
    hide_credentials: true

References are also supported, so this is valid as well (provided the reference exists):

x-kong-plugin-key-auth:
  $ref: '#/components/kong/plugins/key_auth_config'

Request Validator Plugin

To enable validation the request-validator plugin can be added to the root, path or operation object of the Spec and whichever is more specific will be used. You can either specify the full configuration, or have the missing properties be auto-generated based on the OpenAPI spec.

The request-validator plugin has three parameters which can be generated. These are:

  • config.body_schema
  • config.parameter_schema
  • config.allowed_content_types

If any of these are not specified in the x-kong-plugin-request-validator.config object in the OpenAPI spec, they will be generated if possible, otherwise will be configured to allow all body/parameter/content types.

paths:
  /:
    post:
      parameters:
        - in: path
          name: query
          required: true
          schema:
            anyOf:
              - type: string
      x-kong-plugin-request-validator:
        enabled: true
Reference in New Issue View Git Blame Copy Permalink