description-harvester 0.10.0

A tool for working with archival description for public access.

description_harvester

A tool for working with archival description for public access. description_harvester reads archival description into a minimalist data model for public-facing archival description and then converts it to the ArcLight data model and POSTs it into an ArcLight Solr index using PySolr.

description_harvester is designed to be extensible and harvest archival description from a number of sources. Currently the only available sources harvests data from the ArchivesSpace API using ArchivesSnake or EAD 2002 XML files. Its also possible to add additional output modules to serialize description to EAD or other formats in addition to or in replace of sending description to an ArcLight Solr instance. This potential opens up new possibilities of managing description using low-barrier formats and tools.

description_harvester is designed to be a drop-in replacement for the ArcLight Traject indexer. It also includes a plugin that attempts to recognized IIIF manifests included as file versions and uses manifests to fully index digital objects from digital repositories and other sources, including item-level metadata fields, embedded text, OCR text, and transcriptions.

Tested on ASpace up to v3.5.1 but needs some better error handling. Validation is also very minimal, but there is potential to add detailed validation with jsonschema .

Installation

pip install description_harvester

First, you need to configure ArchivesSnake by creating a ~/.archivessnake.ymlfile with your API credentials as detailed by the ArchivesSnake configuration docs.

Next, you also need a ~/.description_harvester/config.yml file that lists your Solr URL and the core you want to index to. These can also be overridden with args. description_harvester reads your config.yml as utf-8, so if you're creating this file in a Windows environment you should ensure its utf-8.

solr_url: http://127.0.0.1:8983/solr
solr_core: blacklight-core
last_query: 0
cache_expiration: 3600
component_id_separator: "_"
online_content_label: "Online access"

The component_id_separator allows for a customizable separator between the collection and component IDs in ArcLight URLs. This can be set to component_id_separator: "" for pre-ArcLight v1.1.0 defaults which had no separator. This will default to _ if this setting isn't set in config.yml as ArcLight now does.

The online_content_label setting allows you to customize the label displayed for items with online content in ArcLight. The default is "Online access".

Adding custom digital object metadata

You can also add custom digital object metadata fields by adding them to your config.yml under the Solr suffix you would like them to be indexed as. These fields must match metadata fields in your IIIF manifests.

metadata:
- ssi:
  - date_uploaded
- ssm:
  - date_digitized
  - extent
- ssim:
  - legacy_id
  - resource_type
  - coverage
  - preservation_package
  - creator
  - contributor
  - preservation_format
  - source
- tesm:
  - processing_activity
- tesim:
  - description

Repositories

By default, when reading from ArchivesSpace, description harvester will use the repository name stored there.

To enable the --repo argument, place a copy of your ArcLight repositories.yml file as ~/.description_harvester/repositories.yml. You can then use harvest --id mss001 --repo slug to index using the slug from repositories.yml. This will overrite the ArchivesSpace repository name.

There is also the option do customize this with a plugin.

Encoding note: While ArcLight does not explicitly read repositories.yml as utf-8, its Rails stack means that you're likely reading it in a utf-8 (non-Windows) environment. Since description_harvester enables you to index from a Windows machine, it expects your ~/.description_harvester/repositories.yml file to be utf-8.

Indexing from ArchivesSpace API to ArcLight

Once description_harvester is set up, you can index from the ASpace API to ArcLight using the to-ArcLight command.

Index by id_0

You can provide one or more IDs to index using a resource's id_0` field

harvest --id ua807

harvest --id mss123 apap106

Index by URI

You can also use integers from ASpace URIs for resource, such as 263 for https://my.aspace.edu/resources/263

harvest --uri 435

harvest --uri 1 755

Indexing by modified time

Index collections modified in the past hour: harvest --hour

Index collections modified in the past day: harvest --today

Index collections modified since las run: harvest --updated

Index collections not already in the index: harvest --new

Indexing from EAD 2002

harvest --ead path/to/ead.xml

You can also give it a directory and it will harvest all *.xml

harvest --ead path/to/ead_files

Verbose output

harvest --id ger071 -v

Caching

description_harvester will cache collections from the ArchivesSpace API, storing them by default to ~/.description_harvester/cache after they are converted to the description model. Cache time is set in seconds as cache_expiration in ~/.description_harvester/config.yml. Thus, cache_expiration: 3600 will use the cached data instead of the ArchivesSpace API for data less than 1 hour old.

You can override the cache path in config or turn caching off gobally with cache_dir: false.

cache_dir: "~/path/to/my_cache"
cache_dir: "C:/Users/username/my_cache"
cache_dir: false

Deleting collections

You can delete one or more collections using the --delete argument. This uses the Solr document ID, such as apap106 for https://my.ArcLight.edu/catalog/apap106.

harvest --delete apap101 apap301

Plugins

Plugins let you add institution-specific customization without modifying the core package. Common use cases might be:

• Customizing repository names based on collection identifiers
• Enriching digital objects with data from local systems (e.g., IIIF manifests, preservation systems)

UAlbany's local plugins may be a helpful example.

Creating a Plugin

1. Copy the template: Copy default.py to ~/.description_harvester/ (or use DESCRIPTION_HARVESTER_PLUGIN_DIR environment variable). You can rename the file.

2. Rename the class: The class name is optional and only for your readability. You could use any class name, but the plugin_name class variable must be unique to your implementation.:

   class MyInstitutionPlugin(Plugin):
       plugin_name = "my_institution_plugin"
   

3. Implement methods: Override any combination of customization hooks:

   • custom_repository(self, resource): Customize repository names

     • Input: ArchivesSpace resource API object
     • Output: Repository name string or None for default behavior

   • update_record_id(self, record_id, record): Customize record IDs

     • Input: Generated record ID string and Component object
     • Output: Custom ID string or None for default behavior

   • update_dao(self, dao): Enrich digital objects

     • Input: DigitalObject with identifier, label, metadata, etc.
     • Output: Modified DigitalObject with additional metadata

Plugin Discovery

Plugins are automatically loaded from (in order):

1. Built-in plugins in the package (e.g., default.py)
2. ~/.description_harvester/ directory
3. Custom directory set via DESCRIPTION_HARVESTER_PLUGIN_DIR environment variable

Note: The .py filename doesn't matter - plugin identification is based on the unique plugin_name class variable. You can have multiple plugin classes in a single .py file, or split them across multiple files.

Example Plugin

from description_harvester.plugins import Plugin
from description_harvester.iiif_utils import enrich_dao_from_manifest

class MyInstitutionPlugin(Plugin):
    plugin_name = "my_institution"
    
    def custom_repository(self, resource):
        # Use custom names for special collections
        if resource['id_0'].startswith('sc'):
            return "Special Collections & Archives"
        return None  # Use default for others
    
    def update_record_id(self, record_id, record):
        # Add repository prefix to collection-level IDs
        if record.level and record.level.lower() == "collection":
            return f"sc_{record_id}"
        return None  # Use default for component IDs
    
    def update_dao(self, dao):
        # Enrich digital objects with IIIF manifest data
        if 'manifest.json' in dao.identifier:
            enrich_dao_from_manifest(dao, manifest_url=dao.identifier)
            # Add custom logic
            dao.metadata['institution_id'] = 'my_institution'
        return dao

IIIF Utilities

For plugins working with IIIF manifests, description_harvester.iiif_utils provides helper functions:

from description_harvester.iiif_utils import (
    fetch_manifest,              # Fetch and parse manifest from URL
    extract_text_from_manifest,  # Extract OCR/transcription text
    get_thumbnail_url,           # Get thumbnail image URL
    get_rights_statement,        # Get rights/license info
    extract_metadata_fields,     # Get all metadata as dict
    enrich_dao_from_manifest,    # All-in-one enrichment
)

def update_dao(self, dao):
    if 'manifest.json' in dao.identifier:
        # Option 1: Use convenience function
        enrich_dao_from_manifest(dao, manifest_url=dao.identifier)
        
        # Option 2: Fine-grained control
        manifest = fetch_manifest(dao.identifier)
        if manifest:
            dao.text_content = extract_text_from_manifest(manifest)
            dao.thumbnail_href = get_thumbnail_url(manifest)
            dao.rights_statement = get_rights_statement(manifest)
            # Optionally add metadata fields from manifests
            dao.metadata.update(extract_metadata_fields(manifest))

            # Set metadata fields with whatever local logic
            dao.metadata['custom_field'] = 'custom_value'
    
    return dao

See the iiif_utils module documentation for all available functions.

SSL Bypass

You may encounter SSL certificate verification errors when fetching IIIF manifests, indicating that the server is not sending the complete certificate chain.

The recommended solution is to work with the server administrator to fix their certificate chain configuration. But if you do need to bypass SSL verification, you can do this by setting the DESCRIPTION_HARVESTER_VERIFY_SSL environment variable to "false".

Use as a library

You can also use description harvester in a script

from description_harvester import harvest

harvest(["--id", "myid001"])