Wildbook 11.0

Wildbook 11.0 expands automated data submission and identification, adds organization-owned API credentials, and improves everyday submission, match review, and catalog management. This release also strengthens background processing, search consistency, and access controls.

These notes cover changes since 10.14.

Highlights

• Automated sighting submissions — opt-in pilot. A new submissions API supports creating drafts, uploading images, validating records, committing imports, and tracking results. New submissions default to detection and individual matching, with an explicit import-only option. Field-specific validation errors, shared sighting IDs, and API source labels in the imports list make integrations easier to operate. Includes an OpenAPI contract and a new submit-sightings agent guide. (#1776)
• Organization service keys. Site administrators and organization administrators can create and revoke scoped service keys through API Access. Integrations exchange these keys for short-lived tokens without relying on a person's login. Key management includes expiration, creator information, and last-use tracking. (#1824, #1826)
• Multiple identification pipelines per species. Wildbook can run multiple configured detection pipelines for a species, including pipelines on different ML service endpoints. Extraction, matching, and PairX inspection use the appropriate model for each annotation class, supporting workflows such as separate body and face identification. (#1802, #1805)
• Expanded MiewID model support. OpenSearch now supports 1,536-dimensional MiewID-DINOv3 embeddings alongside existing 2,152-dimensional embeddings. Nearest-neighbor matching filters by model during candidate selection, preventing other models' vectors from displacing relevant candidates. (#1813, #1821)
• Clearer match review and better mobile layouts. Match images now show encounter dates, filenames, and location IDs where permitted. Submission and match review layouts work better on phones, and manual annotation shows a loading indicator while its image loads. (#1832, #1775, #1816)
• Individual taxonomy editing. Authorized curators can set an individual's taxonomy or derive it from agreeing encounters. The page distinguishes an explicitly recorded taxonomy from one inherited from encounters. (#1739)

Fixes and reliability improvements

• More dependable identification: persist identification requests before dispatch, improve WBIA annotation registration, run all requested bulk-import identification options, and select complete result runs. Exhausted ML service retries now reach a terminal status instead of remaining pending. Explicitly empty identification configurations are respected.
• More resilient queues: consumers recover from transient failures, consumer supervision improves recovery after unexpected exits, and pending delayed retries are flushed during application shutdown. An optional dedicated queue lets interactive matches proceed independently of a bulk detection backlog.
• More consistent search and permissions: indexing is dispatched after database commits, repeated requests for records already being indexed are retained, and full reindexes preserve encounter visibility. A scheduled permissions audit detects and repairs access-control drift across encounter, annotation, and individual indexes.
• Correct sighting imports: submissions preserve shared sighting IDs, and bulk imports derive sighting coordinates after all rows have been populated.
• Catalog and interface fixes: resolve merge-page hangs and loading failures for legacy individual IDs; show researcher comments to users who can view the individual; fix annotation overlay rendering and zoom control labels.
• Stronger account and upload controls: fix organization membership collisions; limit organization-admin account edits to authorized users; preserve site-admin-managed roles during account edits; restrict user consolidation to site administrators; and harden upload path handling and staging cleanup.
• Better integration diagnostics: correctly handle explicit data:read token requests and ML service HTTP errors. Prometheus metrics now expose queue health and Modified Groth/I3S scan activity. API documentation expands guidance on vector search and encounter change polling.

Upgrade notes

• Existing databases require an organization membership migration. Stop Wildbook, run archive/sql/organization_members_pk.sql against the installation database, then start 11.0. The migration replaces positional membership rows with unique organization/user pairs. The script also documents rollback requirements.
• Check the annotation vector index before upgrading. Model-filtered nearest-neighbor queries require a filter-capable Faiss or Lucene index. Legacy NMSLIB indexes must be replaced with a compatible index; the previous post-filter fallback has been removed.
• Deploy the updated servlet and security mappings. This build excludes WEB-INF/web.xml from the packaged WAR. Installations that manage the descriptor separately must deploy the updated mappings alongside the application.
• Submissions remain disabled by default and are limited to context0. Enabling the pilot requires explicit api-submission enrollment, private staging storage, schema setup, and admission/worker configuration. Follow the submissions pilot runbook.
• Review background processing settings. The dedicated interactive match queue is opt-in through iaFastlaneQueueEnabled. Deployments with multiple application instances sharing a database must enable the permissions audit runner on only one instance.
• Legacy features removed: the COCO export route, Microsoft Access importer, legacy WebImport spreadsheet importer, and unused ImportIA, ImportSRGD, UpdateStandard, and FixStandardImport endpoints. Use the supported bulk-import workflow or the enabled submissions API for ingestion, and API Access for programmatic data retrieval.

Contributors and reviewers

Thank you to the open source contributors and code reviewers who helped make this release possible:

• Shreyash Neeraj (@shreyashsri79) — individual taxonomy editing, researcher comment visibility, and ML service HTTP error handling.
• @marekl11 — annotation overlay rendering and match review control labels.
• @UrbanFercec — code reviews, implementation feedback, and pull request approvals across identification, permissions, account management, and interface changes.
• @naknomum — code reviews and pull request approvals throughout the release, including deployment and database compatibility feedback.
• The Wild Me team — additional development and maintenance recorded under the shared Wild Me commit identity.

Thanks also to the researchers, users, and community members who report issues, test workflows, and help improve Wildbook.

Full changelog: 10.14 → 11.0