Seed Server

A seed server hosts available seeds for you to search and install. It will also be able to automatically respond to Git post-commit hooks to automatically deploy new versions.

Current Status

The server is not implemented yet. The following describes the intended behavior

Usage

Start a new server instance with:

seed server start

By default, server cache is stored in ~/.seeds/server. You can name an alternate root with the --cache option. It will also start on the port 4050. You can change this with the --port option.

HTTP API

The seed server is a RESTful HTTP service

Authentication

Authentication is handled via an API token which is assigned to a user account on the system. The token must be passed as a query parameter to any mutation URL as well as to any package that is marked private.

Obtaining a New API token

To obtain an new token you must have an existing, valid token. This basically means someone else who already has access must give you access as well. Get a new token with the URL:

POST /seed/users/cjolley/token

The username is the name that will be used to assign permissions. This will return a 201 Created with a Location header identifying where you can fetch the new token. In general the new token is the last component of the URL. Probably something like:

/seed/tokens/1234

In this case 1234 would be the token.

Withdrawing a Token

If you think a token is compromised or want to otherwise cancel the token just delete it:

DELETE /seed/tokens/1234

Setting User Permissions

To give a particular user permission to modify a package, you must create an ACL entry for the user:

PUT /seed/access/jake/cjolley
{ "access": "commit" }

This will make cjolley a committer for the jake package. You can also leave the body blank - this will automatically grant commit access.

You can withdraw permission by deleting the same resource:

DELETE /seed/access/jake/cjolley

Valid access modes include "commit" and "retrieve".

Requests made with no API token are given the "anonymous" username. You can make a package private by deleting the anonymous access.

Finding Packages

You can search for package information by getting the /seed/packages URL. You can specify the following query options:

• name=NAME: the name of the package info to retrieve
• version=VERSION: searches for the exact version in the repo
• compatible=VERSION: returns any comptaible versions
• dependencies=true|false: return also package info for dependencies (this defaults to false if not specified)
• q=KEYWORDS: returns any packages matching the comma or space separated keywords

Example URL:

GET /seed/packages?name=sproutcore&dependencies=true&compatible=1.0.1000

The return value from this request is a JSON document describing any matching packages. The format is:

{
	"packages": [
		{
			"name": "sproutcore",
			"version": "1.0.1000",
			"dependencies": {
				"seed": "1.1",
				"tiki": "2.1.12"
			},
			"description": "SproutCore HTML5 Application Framework",
			"link-self": "/seeds/packages/jake/0.1.2/package.json",
			"link-asset": "/seeds/packages/jake/0.1.2/asset.zip"
		}
	]
}

The most important parts are the name, version, and location fields.
The name and version fields should be self explanatory. The location field describes the URL you should use to download the asset. If the field does not begin with a full URL (i.e. "http://hostname/...") then attach the same hostname used to make this request.

If you pass dependencies=true as a query parameter, then the returned results will include both the packages field and a dependencies field.
The dependencies field is an array of package descriptions that do not match the search query but appear in the dependency chain of those that do.

If a package matches the search query and is a dependency of another package then it will appear in the packages field only, not in both.

Retrieving Individual Package Information

In general, you should only retrieve individual package information by following the URLs provided in a general /seed/packages search. This way seed server can change its other URL structures without breaking your code.

You can, however, retrieve the information about a single package by calling GET from the package URL. The package URL contains the package name and version.

GET /seed/packages/jake/0.1.2/package.json

Downloading a Package

You will almost always download a package after discovering the package URL via a search. You should never try to compute directly the URL used to download the package.

In general, once you obtain an URL, you simply GET the URL to download the package.

GET /seed/packages/jake/0.1.2/asset.zip

Adding a Package

Post a ZIP or tarball'd version of your package content to:

POST /seed/packages

The return value will be "201 Created" with a Location header containing the URL where you can fetch the new package information. Make sure your Content-Type header is set properly or else the package will not be properly installed.

Withdrawing a Package

Generally once a package is published it should not be removed. It is possible to withdraw a package by sending a DELETE to the package URL. However this should be exercised with caution.

DELETE /seed/packages/jake?version=0.1.2&token=TOKEN

Git Post Commit Hooks

Seed server can automatically update a particular package. POST'ing to this URL will cause the server to clone the git repository and try to update the named package or packages. It will look at the package in the repository. If the version named in the package.json is not a currently published version, it will be added to the repository automatically.

A post commit hook should look like this:

POST /seed/ping?from=GIT_URL&paths=path1,path2&token=TOKEN

The query parameters are very important for this to work. Note that the body of the post is ignored. Your queries should be:

• from: URL to use to clone the git repository
• paths: optional one or more paths within the repository where packages can be found
• token: API token. the token must match a user with permission to update any discovered packages or they will be ignored

Event Feeds

Seed server maintains event feeds for each package and for the services as a whole. You can use these event feeds, for example, to build a UI around the service.

To get a feed of all the events that have occurred on a particular package use the form:

GET /seed/packages/jake/events

Or to get a feed of all the events that have occured on packages a particular user has access to use:

GET /seed/users/cjolley/events

To get the public feed (essentialy the events feed for the anonymous user), use:

GET /seed/users/anonymous/events

All event feeds take an optional page param which can be used to select a page of events. Event feeds are JSON with the following format:

{
	"events": [
		{
			"kind": "kind of event",
			"description": "human readable description",
			"created": "time event occurred"
		}
	]
}

Other metadata may be included in event feeds as well.