I think it's the lack of hope that comes and gets you. … See, I believe in hope. I believe in belief. — Ted Lasso

And I believe that bringing people together is the best way to build something great. You don't need millions or the ability to throw rockets into space, you need a purpose and determination to build hope out of thin air.

This release is about building teams. GregoryAI is now a proper multi-tenant server and is already being used to speed up our research around brain regeneration.

Brain-regeneration.com was launched on the 16th of April and brings together 3 research labs.

• CNS, Blood and Peripheral Inflammation Lab — Lisbon, Portugal
• REGENERAR — Coimbra, Portugal
• LPJ Lab — Cambridge, UK

This release includes new API and newsletter features, improvements to newsletter subscriptions and lists, analytics on subscribers, per-organisation credentials, a Postgres and Python upgrade, and new data sources (bioRxiv, medRxiv, ClinicalTrials.gov, PNAS).

If you're running GregoryAI on your projects, please let us know. That way we can keep you up to date with the roadmap and add the project to the showcase on the main site.

Read on for the details.

Gregory AI v24

Range: v23 (2025-06-21) → main (2026-05-30). 113 merged PRs, ~717 commits.

Highlights

• GregoryAI is now multi-tenant: API keys, ORCID credentials, and editorial content are scoped per organization and team.
• New full-text search endpoint across articles, trials, and authors, with CSV and streaming export for large downloads.
• New data sources: bioRxiv, medRxiv, ClinicalTrials.gov, and PNAS.
• Announcements: rich, CKEditor-authored emails with CTA buttons and inline images.
• Platform upgrade to Postgres 17 and Python 3.12.

⚠️ Before you upgrade

Two changes will break a deploy if missed. Read these first, then the full list below.

• Postgres 15 → 17 is a major version bump — it needs a dump/restore or pg_upgrade, not an in-place restart.
• The API is now read-only. External clients that POST or PUT must move to the org-scoped write path.

The full breaking-changes list and step-by-step upgrade instructions are in the Breaking changes and Upgrade sections below.

What's new

Multi-organization / multi-site

• API keys per organization and team; org-scoped POST enforcement.
• Per-organization ORCID credentials (env-var dependency removed).
• Flag to mark an organization's API private, with visibility middleware.
• Multi-site support for subscription lists.

API

• Full-text search endpoint (articles, trials, authors) with performance tuning.
• CSV export and streaming responses for large downloads; full-results search export.
• New /stats/ aggregate endpoint with org filter, reduced joins, and caching.
• Author endpoints: sort by article count; author statistics on /categories/.
• New filters: intersection, has_clinical_trials (detects NCT IDs in abstracts), case-insensitive trial status, ML-score threshold, date filters.
• Per-org takeaways and summaries in serializers; read-only viewsets.

Email and subscriptions

• Announcements: CKEditor-authored emails with CTA buttons and inline images, duplicate action, per-org ownership, image-quality and host-config safeguards.
• Weekly digest: per-date or per-relevancy modes, article limits, threshold filtering.
• Subscriber analytics: list distribution, historical active-subscriber chart.
• Dark-mode and Outlook email rendering fixes; sender-name setting; API-domain unsubscribe links; import, reconcile, and prune subscriber commands.

Data sources and ML

• bioRxiv and medRxiv support; ClinicalTrials.gov API; PNAS RSS.
• New RSS feeds for articles and clinical-trials-by-subject.
• Better ML algorithms and category matching with batch processing.

Admin and ops

• Richer admin: sources, subjects (with content analytics), teams (categories, sources, subjects inlines), per-org editorial inlines, subject deletion.
• Multi-arch Docker build and push; Makefile build commands.
• DB pruning for notifications and history; orphan author and article cleanup commands.
• Postgres 17, Python 3.12, codebase cleanup.

⚠️ Breaking changes

Read before upgrading.

• Postgres 15 → 17. Major version bump; requires a dump/restore or pg_upgrade, not an in-place restart. See DEPLOYMENT_RUNBOOK.md.
• Python 3.12. Rebuild the image; pyproject.toml was removed — deploys install from requirements.txt.
• API is now read-only. All ModelViewSets are locked to read-only (Phase 5). External clients that POST or PUT must move to the org-scoped write path.
• DRF Token auth removed. The /api-token-auth/ endpoint and authtoken tables are gone (Phase 6). Use JWT or per-organization API keys instead.
• Legacy article fields dropped. Articles.takeaways and summary_plain_english now live in ArticleOrgContent (per-org). Confirm data was moved before deploying — see MIGRATION_SAFETY.md.
• Team.organization is now NOT NULL. Defensive null-handling removed; every team must belong to an organization.
• allowed_domains moved from Lists to CustomSetting.
• TeamCredentials model removed — ORCID and credentials are now per organization and team.

Upgrade

Review MIGRATION_SAFETY.md first — there are 74 migrations, including irreversible data migrations and a Postgres major-version upgrade. Then follow DEPLOYMENT_RUNBOOK.md.

1. Back up the database and confirm the dump is restorable.
2. Upgrade Postgres 15 → 17 via dump/restore or pg_upgrade.
3. Rebuild the image on Python 3.12 (installs from requirements.txt).
4. Run migrations.
5. Run python manage.py createcachetable gregory_cache to set up the new DB cache backend.
6. Migrate any external POST/PUT clients to the org-scoped write path and off Token auth.