This is the Commons Machinery metadata catalog. It stores Media Annotations (metadata) for creative works and supports lookup by URI or by perceptual hashes (for images).
- Node.js
- MongoDB
- ZeroMQ
On Ubuntu 14.04:
sudo apt-get install build-essential nodejs npm libzmq3-dev libkyotocabinet-dev kyotocabinet-utils
You need to have a MongoDB for the Catalog to strore its data. It's recommended
run MongoDB in a Docker image during
development. In production all parts should be run in Docker (or on a
PaaS). For details, see doc/docker.html.
If you're just testing things out, you can simply install MongoDB in your regular system. Please do note that MongoDB is limited to 2GB databases on 32-bit systems. A 64-bit system is highly recommended! To install MongoDB on Ubuntu 13.10 and 14.04:
sudo apt-get install mongodb-server
The searches for perceptual hashes uses hmsearch, which is available from http://github.com/commonsmachinery/hmsearch and the Catalog expects an initialised database in its root folder with the name hashes.kch.
Initialise the hash database by running hm_initdb from the
hmsearch library:
/path/to/hm_initdb hashes.kch 256 10 10000
The second argument must be 256. The third is the maximum hamming distance that should be allowed. The last argument is an indication of the expected number of hashes, and is used to tune the database index.
See doc/config.md for documentation. Default settings currently work
for local/development setups.
Run ./setup_devenv.sh in the top dir to install all dependencies.
Run make to build the CSS files necessary for the web interface.
There are a number of entry points to different parts of the system,
see doc/codestructure.md.
Current development is focused on the frontend index. It can be started with suitable env vars for development like this:
BLUEBIRD_DEBUG=1 DEBUG='catalog:*' NODE_ENV=development nodejs frontend/index/main.js
To run the full (very much work-in-progress) catalog frontend and all
backend tasks in a single process, run main.js in the top directory.
BLUEBIRD_DEBUG=1 DEBUG='catalog:*' NODE_ENV=development nodejs main.js
doc/example-works.txt contain a list of 100 works from Wikimedia Commons
that can be used as sample data. The list of works is in the Data Package
specification, which can be found in doc/datapackage.md. To load the sample
data into the database, you need to import the data package, and then
populate the search catalog.
To import the data package, you must specify to which user the imported works should be assigned. If you're running the Catalog in development mode, you can also create a fake test user account. You can get the identifier of your account by running the following command. This will create a test user if it doesn't already exist:
curl -u test: -X GET http://localhost:8004/users/current
This command will return something like this:
Moved Temporarily. Redirecting to http://localhost:8004/users/542af1de876096426387c9a1
Where the hash at the end of the string represents the identifier of the
user. This is what you'll now use to load the sample works, calling on
modules/core/scripts/load.js to do the job. Replace the user identifier
below with the identifier from your own installation:
nodejs scripts/load/load-db.js --user 542af1de876096426387c9a1 --verbose true doc/example-works.txt
The hash database is populated by a separate script. If the database
isn't initialised yet, run the hm_initdb command above. Populate
the hash database from the data package:
nodejs scripts/load/load-hash.js --verbose true doc/example-works.txt
The hash database cannot be updated if a catalog process is already running. Stop it to run the populate script. As an alternative, you can specify a different hash file than the default by providing a configuration env var. E.g.:
cp hashes.kch new-hashes.kch
CATALOG_SEARCH_HASH_DB=new-hashes.kch nodejs scripts/load/load-hash.js --verbose true doc/example-works.txt
Then stop the catalog, replace hashes.kch with new-hashes.kch, and
restart.
The primary login mechanism on the web pages is Mozilla Persona: https://login.persona.org/
For testing the mocked IDs from https://mockmyid.com/ can be used:
simply login with [email protected], if you don't want to use a
real email adress.
When the frontend is run in development mode there are also simple
test accounts that doesn't require any password at all. Either login
on the web page or pass --user test: to curl (choosing whichever
username you need). For these test accounts, a faked email address
user@test is created.
TODO: OAuth access to the REST API.
The API is documented here: http://docs.cmcatalog.apiary.io/
Click on the double downward arrows to expand each call showing example usage as plain JSON or in various programming languages.
All PUT, POST and DELETE require a valid user session (see above
about development accounts). GET will return publically visible
information without any session.
The web and REST endpoints are overloaded on the same paths, so to get
a REST response to a GET request you must set the Accept header to
application/json.
Here are some useful curl commands to poke the API:
Find out current user:
curl -k -v -u test: -X GET http://localhost:8004/users/current
Get user profile:
curl -k -v -u test: -H 'Accept: application/json' -X GET http://localhost:8004/users/53a80969b22cfae451ec8ed4
Update user profile:
curl -k -v -u test: -d '{"alias":"new alias"}' -H 'Content-Type: application/json' -X PUT http://localhost:8004/users/53a80969b22cfae451ec8ed4
List works:
curl -H 'Accept: application/json' http://localhost:8004/works
Get a work:
curl --user test: -H 'Accept: application/json' http://localhost:8004/works/542af1de876096426387c9a1