MOAI is a platform for aggregating content from different sources and publishing it through the Open Archive Initiative Protocol for Metadata Harvesting (OAI-PMH). It can harvest data from various sources — OAI feeds, SQL databases, XML files, Fedora Commons, EPrints, DSpace — and serve multiple OAI feeds from a single server, each with independent configuration.
Support graciously provided by

This is a maintained fork of MOAI by Infrae, adding Python 3 support, modern packaging (pyproject.toml, uv), and GitHub Actions CI. Changes were offered upstream via PR #5.
Note: Other than modernizing the tooling, there are no major functional changes. Some parts of the documentation below may be outdated. Patches welcome.
| Python | 3.9 | 3.10 | 3.11 | 3.12 | 3.13 |
|---|---|---|---|---|---|
| ✓ | ✓ | ✓ | ✓ | ✓ |
We recommend using uv for dependency management. Instructions below are for Unix, but MOAI should also work on Windows.
cd moai
uv syncpip install MOAI-iplwebuv sync --extra test
uv run pytestThe development server should never be used in production. It is convenient for testing and development.
cd moai
uv run paster serve settings.iniThis will print something like:
Starting server in PID 7306.
Starting HTTP server on http://127.0.0.1:8080
You can now visit localhost:8080/oai to view the MOAI OAI-PMH feed.
Configuration is done in the settings.ini file. The default settings file uses the Paste#urlmap application to map WSGI applications to a URL.
In the [composite:main] section there is a line:
/oai = moai_example
Which maps the /oai URL to a MOAI instance. This makes it easy to run many MOAI instances in one server, each with its own configuration.
The [app:moai_example] configuration lets you specify the following options:
| Option | Description |
|---|---|
name |
The name of the OAI feed (returned in Identify verb) |
url |
The URL of the OAI feed (returned in OAI-PMH XML output) |
admin_email |
The email address of the admin (returned in Identify verb) |
formats |
Available metadata formats |
disallow_sets |
List of setspecs that are not allowed in the output of this feed |
allow_sets |
If used, only sets listed here will be returned |
database |
SQLAlchemy URI to identify the database used for storage |
provider |
Provider identifier where MOAI retrieves content from |
content |
Class that maps metadata from provider format to MOAI format |
The MOAI system is designed to periodically fetch content from a provider, and convert it to MOAI's internal format, which can then be translated to the different metadata formats for the OAI-PMH feed.
MOAI comes with an example that shows this principle:
In the moai/moai directory there are two XML files. Let's pretend these files are from a remote system, and we want to publish them with MOAI.
In the settings.ini file, the following option is specified:
provider = file://moai/example-*.xml
This tells MOAI that we want to use a file provider, with some files located in moai/example-*.xml.
The following option points to the class that we want to use for converting the example content XML data to MOAI's internal format:
content = moai_example
The last option tells MOAI where to store its data, this is usually a SQLite database:
database = sqlite:///moai-example.db
Now let's try to add these two XML files. First visit the OAI-PMH feed to make sure nothing is already being served:
http://localhost:8080/oai?verb=ListRecords&metadataPrefix=oai_dc
This should return a noRecordsMatch error.
To add the content, run the update_moai script with the section name from the settings.ini as argument:
uv run update_moai moai_exampleThis will produce the following output:
/ Updating content provider: example-2345.xml
Content provider returned 2 new/modified objects
100.0%[====================================================================>] 2
Updating database with 2 objects took 0 seconds
Now when you visit the OAI-PMH feed again you should see the two records:
http://localhost:8080/oai?verb=ListRecords&metadataPrefix=oai_dc
When you run the update_moai script again, it will create a new database with all the records. It is also possible to specify a date with the --date switch. When a date is specified, only records that were modified after this date will be added. The update_moai script can be run from a daily or hourly cron job to update the database.
It's possible — and most of the time, needed — to extend MOAI for your use-cases. The Provider and Content classes from the example might be a good starting point. All your customizations should be registered with MOAI through entry_points. Have a look at MOAI's pyproject.toml for more information.
The best approach would be to create your own Python package with pyproject.toml and install it in the same environment as MOAI. This will let MOAI find your customizations. Note that when you change something in your package metadata, you have to reinstall the package for MOAI to pick up the changes.
The moai.interfaces file contains documentation about the different classes that you can implement.
Instead of writing your own provider/content classes, you can also register your own custom database. Implementing a replacement for moai.database.SQLDatabase can be more complicated than writing a provider/content class, but it has the advantage that MOAI is always up to date and you don't need a second SQLite database.
Have a look at the pyproject.toml file — it registers several databases. You could use this mechanism to register your own database from your own Python package.
In the settings.ini configuration you can then reference your database (mydb://some+config+variables).
For the database, have a look at the generic database provider in database.py. The only methods that you need to implement are: oai_sets, oai_earliest_datestamp and oai_query.
The oai_query method returns dictionaries with record data. The keys of these dictionaries are defined in the metadata files (for example metadata.py) — have a look at the source.
For oai_dc there are the following names:
title, creator, subject, description, publisher, contributor, type, format, identifier, source, language, date, relation, coverage, rights
So a return value would look like:
{'id': '<oai record id>',
'deleted': '<bool>',
'modified': '<utc datetime>',
'sets': ['<list of setspecs>'],
'metadata': {
'title': ['<list with publication title>'],
'creator': ['<list of creator names>'],
...}
}BSD-3-Clause — see LICENSE.txt for details.