SSU is a scripted web site navigator & scraper. It was originally designed and conceived as part of Wesabe's infrastructure and has since been open-sourced. Its original design goal was to extract OFX data given bank usernames and passwords for use on wesabe.com.
The system it uses to get this data is XulRunner, a project from Mozilla that provides a customizable (and scriptable) browser. SSU has scripts for each financial institution it supports that describes how to log in and download data from that institution's web site.
It was originally written in JavaScript but is currently mostly CoffeeScript. You can write bank scripts in either language, though CoffeeScript will be the preferred one going forward.
If you're trying to aggregate transaction data from multiple financial institutions, possibly for a large number of people, then this project might be useful to you.
First, clone the SSU repo:
$ git clone https://github.com/wesabe/ssu
The easiest way to try this is on your laptop/desktop computer running
Linux or Mac OS X. Windows isn't supported. SSU comes with a bunch of
scripts for financial institutions that it already supports. Your
initial experience trying out SSU is going to be much easier if you have
an account at one of these institutions. To check, go to the
fi-scripts folder and start looking for your bank. Let's
say your bank is Chase, whose site is chase.com. We store the scripts
for financial institutions in a reverse DNS folder structure, so you
need to look in the com
directory for the chase.coffee
script.
If your financial institution is supported, then great! Next you'll need
to install XulRunner. If you're on Linux, you'll want to use your
package manager (e.g. apt-get
). If you're on OS X you can use the
bundled setup script:
ssu$ ./bootstrap
That'll install it if it's not installed and tell you the version you have installed if it is already. Now go ahead and start the app itself in a terminal window:
ssu$ bin/server
You'll see some logging output along with some startup messages and a blank browser window titled "Wesabe DesktopUploader". As long as you don't see any errors you should be good to go. Next you can generate a credentials file to test with. Again, let's assume you have an account at Chase. In another terminal window, run this:
ssu$ bundle
ssu$ bundle exec script/generate credential com.chase chase
That'll create a file at credentials/chase
that looks like this:
{"creds": {"username": "FIUSERNAME", "password": "FIPASSWORD"}, "fid": "com.chase"}
Just change FIUSERNAME
and FIPASSWORD
to your username and password
for Chase and save the file.
Now fire up the test client and start a job:
ssu$ script/console
>> job = Job.create chase
Your first terminal window and the blank browser should now be doing something -- ideally logging into your financial institution site and getting your recent transaction data. If it succeeds it'll store the downloaded statement's in the app's profile directory. You can get a list like so from the console:
>> list = Statement.all
=> [#<Statement:0x10f014018 @id="1D87E4D6-DCCD-0001-FFFF-1FFF1FFF1FFF">]
>> list.first.read
=> "OFXHEADER:100\r\n..."
Congrats, you've successfully gotten data out of your financial institution's website!
The original application that used SSU is the one SSU was written for at Wesabe: pfc. You can see how to manage SSU in this file that controls the SSU process and this file that talks to it.
Basically, SSU sets up a tiny HTTP server (at port 5000 by default) for commands. Here's a request to list all the statements that have been downloaded:
POST /_legacy HTTP/1.0
{"action": "statement.list"}
Here's one that starts a job with credentials:
POST /_legacy HTTP/1.0
Content-Type: application/json
Content-Length: 76
{"action": "job.start", "body": {"fid":"com.ingdirect", "creds":{"username":"joesmith","password":"iamgod"}}}
Responses will be in JSON regardless of what Accept header you pass:
# a successful response to the /statements request
HTTP/1.0 200 OK
Content-Type: application/json
{"status": "ok", "statements": ["1D8787AA-6D2D-0001-DFF3-9EB052301CD4"]}
# an example error response
{"status": "error", "error": "ReferenceError: foo is not defined"}
NOTE: The /_legacy
route is a temporary compatibility measure with the
old pure-socket way of communicating with SSU. Eventually it'll be replaced by
REST-based routes (e.g. GET /statements
, POST /jobs
, etc).
You can use any programming language you like that supports spawning processes and HTTP to manage an SSU instance. This project ships with development tools that also serve as basic examples in the server (spawning) and console (communication & managing via api.rb) scripts.
Yep, there's a generator for that which will build a skeleton script for your financial insitution:
ssu$ bundle
ssu$ bundle exec script/generate player com.ally "Ally Bank" https://www.ally.com/
Generating with player generator:
[ADDED] application/chrome/content/wesabe/fi-scripts/com/ally.coffee
[ADDED] application/chrome/content/wesabe/fi-scripts/com/ally/login.coffee
[ADDED] application/chrome/content/wesabe/fi-scripts/com/ally/accounts.coffee
You can probably leave the base script (ally.coffee
in this example) alone
and start filling in login.coffee
with the info required to navigate the
site. Once you've added something and created a matching credential file, go
ahead and try it out:
ssu$ script/console
>> job.start ally
There are lots of examples in the fi-scripts
directory for you to reference
as you build your own script. Once you're satisfied just commit your files and
send a pull request so we can add your financial institution for others to use.
Using XulRunner means that SSU can access any bank site that Firefox can, so you don't have to use mechanize or some other tool that doesn't fully emulate the browser environment. This matters because, by its nature, navigating any website in a scripted way is brittle and anything we can do to reduce the breakage is good. Websites are intended to be viewed in web browsers and their authors worked hard to make that function properly -- that is work you don't have to do when you use a browser as your scraper.