Proaixy

A simple OAI proxy adding sets to the records, to enable a "search" by author name for instance.

Proaixy serves the metadata it harvests from its sources on a single endpoint, and adds sets to these records. These "virtual" sets are created by extractors using the metadata itself.

Proaixy is written as a Django website, using Celery as task manager.

Harvesting method

Proaixy was designed to harvest from unstable sources i.e. that do not exactly comply with the OAI-PMH standard, or return a dirty XML output, or return various kinds of errors.

The goal is to minimize the amount of records we have to download again after these failures. Hence proaixy harvests records by batches of records contained in a small timeframe (one week by default), starting with the earliest datestamp declared by the interface.

Installation

You need the following dependencies (on Debian): postgresql-server-dev-9.4 python-dev build-essential libxml2-dev libxslt1-dev

I recommend installing the following dependencies in a virtualenv: Python dependencies can be installed using pip install -r requirements.txt (you can do it in a virtualenv).

1. Create a postgresql database and put the access details in proaixy/settings.py
2. Run python manage.py syncdb. You will be prompted to create an admin account on the interface.
3. Run celery --app=proaixy.celery:app worker -B -l INFO
4. In parallel, run python manage.py runserver

You can access the interface at http://localhost:8000/

The OAI endpoint can be found at http://localhost:8000/oai

Configuration

Log in to http://localhost:8000/. You will find a form to add a new OAI-PMH source. Two fields are required:

• The URL of the endpoint
• A short identifier for the source, preferably without spaces and special characters (something you can write in an OAI set) Proaixy will query the endpoint to get more details about it, using the Identify verb.

Once it is added, you can harvest it by clicking the appropriate link. Harvesting can be stopped by "revoking" the appropriate task. Only the current batch of records will have to be downloaded again in a later harvest.