With Go installed, you should be able to run:
make build
This will build the Integra executable and put it at ./local/integra. You can
put it into your PATH or run it from there. To match examples below, you should
move it to a PATH directory like /usr/local/bin.
This project is still experimental, piloting with these services:
- github
- digitalocean
- spotify
- google-calendar
There is also google-keep, but it can only be described/inspected. Using this API is
only available to Google enterprise users so it is not yet supported.
The integra call and integra fetch commands will need authentication. This is
still a mostly manual process for each service that involves getting an access token
and setting it as an environment variable before using these commands.
Create a personal access token with all scopes
and set it in your environment as GITHUB_TOKEN.
Create a personal access token with all scopes
and set it in your environment as DIGITALOCEAN_TOKEN.
If you don't have OAuth client credentials...
you need to make an app
in the Spotify Developer Dashboard. It should be set up for "Web API" and "Web Playback API". It should
also have a Redirect URI of http://localhost:4532/auth/callback. You want the Client ID and Client Secret
from the app settings once created.
With Spotify OAuth client credentials, set them in your environment as SPOTIFY_CLIENT_ID
and SPOTIFY_CLIENT_SECRET. Now run:
integra auth spotify
It should open your browser to login and authorize, then redirect to a page you can close.
The output of the command should contain an access token valid for 1 hour that you can
set in your environment as SPOTIFY_TOKEN.
If you don't have a Google client credentials JSON file...
you need to create a project on the Google API Console. Enable the "Google Calendar API" for the project by
searching for it in the API Library
making sure the new project selected in the top bar. Click the result and then the "Enable" button. Now
create an OAuth client ID
with Application Type of "Web application" and an Authorized redirect URI of
http://localhost:4532/auth/callback. Download the client_secret.json file from the API Console.
There are expanded instructions here.
With a credentials JSON file, open it with a text editor and copy the single line contents into the clipboard. Then
set it in your environment as GOOGLE_CLIENT_JSON using single quotes, like this:
export GOOGLE_CLIENT_JSON='<json data here>'
Now you can run:
integra auth google-calendar
It should open your browser to login and authorize, then redirect to a page you can close.
The output of the command should contain an access token valid for 1 hour that you can
set in your environment as GOOGLE_CALENDAR_TOKEN.
Integra commands often take a selector in this format: <service>.<resource>.<operation>.
The resource and operation parts are both optional, so a selector could just be a
service name. To see available services run integra describe without a selector.
The integra describe <selector> subcommand can take a selector and outputs information about the
selected service, resource, or operation. When describing a service, this includes
the available resource names (often grouped into categories). When describing a
resource, this includes the available operation names.
The integra call <selector> [data...] subcommand will perform an operation by selector. After the selector
you can optionally provide parameters and input data using CLON syntax.
In the simple case this is just key=value arguments.
This command requires access tokens to be present in the environment for the selected service.
The integra fetch <service> <directory> subcommand will attempt a one-way sync of data from the
service API to the specified directory. This command is a work in progress, so it
likely won't sync everything, and the organizational structure may change. However,
it's important to know it will only fetch data for subjective endpoints,
that is, endpoints that return data specific to the authenticated user. Some APIs like
digitalocean, this is every endpoint.
This command requires access tokens to be present in the environment for the selected service.
Outside normal API modeling concepts, we introduce "content orientation",
which is whether content in a response is relative or absolute.
Relative content is specific to the authenticated user. Absolute content is content
that is always the same. A collection of endpoints, such as endpoints associated with a
resource or the API as whole, can also be mixed.
This is used to identify endpoints returning content owned by or associated with the authenticated user. An API can have a content orientation to help classify it as an API for account-specific content (like DigitalOcean, or Google Calendar), or content that is the same for everybody (like an RSS feed for a blog), or is both (like GitHub, or Spotify).
Integra supports OpenAPI Descriptions or Google Discovery Documents. If the API and provider are the same, for example digitalocean, you can make a directory under services. If a provider has multiple APIs, make a directory for the provider, like google, and a subdirectory for the API, calendar, which would make a service named google-caledar. In either case, the service directory needs a meta.yaml file and a directory for specific versions of the API description. This directory is named by the major version number of the API, so an API with version 1.0 would be 1. Integra expects either an openapi.yaml file, openapi.json file, or a googleapi.json file in this directory.
This would end up looking something like this:
services
├── digitalocean
│ ├── 2
│ │ └── openapi.yaml
│ └── meta.yaml
├── google
│ └── calendar
│ ├── 3
│ │ └── googleapi.json
│ └── meta.yaml
└── etc...
The minimal content of meta.yaml is a latest key with the major version directory
name as a string value. This file will be used to add extra metadata to services. Here
is the current data of meta.yaml:
| Key | Type | Description |
|---|---|---|
| latest | string | Required. The latest/default version directory to use. |
| contentOrientation | string | Either "mixed", "relative", or "absolute". Default: "mixed" |
| relativeContentPaths | list of regexp strings | Matched paths are marked "relative" |
Once all this is set up, the service should be available to integra describe after rebuilding. Here is what you can
run to make sure everything looks right:
integra describe --info <service>
Ideally all fields are non-empty.
integra describe --resources <service>
For OpenAPI, Integra tries to infer resources from paths. This is where most massaging happens.
Before the listing output, it should log any issues preventing it from inferring resources
from paths. I dump these into a NOTES file for the service where we can work out
the best way to improve our inference system. Even if we end up using service specific
techniques, we're trying to make sure any API added will have reasonable resources by default.
Outside of logged issues, the resources need to be inspected to make sure they look right. Resource names are singular, but some words need to be marked as invariant or as an acronym. Acronyms should be uppercased and not singularized. Any other weird looking resource names should have a GitHub issue filed.
integra describe <service>.<resource>
This may output the same errors as before, but any unique errors should be reported. Otherwise it should should non-empty fields under info, and a list of operations.
integra describe <service>.<resource>.<operation>
Same as before, you may see familiar errors, and new errors should be reported. Info fields should look correct. After the Info section the sections will be different depending on the operation, but all involve summarizing the top-level schema of the data used in the operation. First, there may be a "parameters" section for query and URL parameters. Then, there may be an "input" section showing the top-level schema for data sent as request body. Lastly, there should either be an "output" section or a "response" then "output" section. Output focuses on the principal data type expected to be returned. If response is shown, it is the schema for the envelope of the output. For example, pagination data in list operations.
Descriptions in the schemas aren't required, but type information is. If the type is blank for a field, double check the service document has this information, or if Integra is not parsing properly. For now, only the latter should be an issue.