Changelog

All notable changes to Cellarium Cloud will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

1.8.4 - 2026-04-29

Added

  • Added API endpoint for serving precomputed Cell Ontology (CL) resources (#214)

  • Added OntologicalColumnInfo model and API response fields for ontology column metadata (#214)

  • Added script to generate and upload CL ontology resources (#214)

  • Added unified build-and-deploy GitHub Actions workflow for CAS services

  • Increased Uvicorn worker count and Cloud Run concurrency settings

Fixed

  • Fixed VPC egress setting to private-ranges-only in deploy action

  • Fixed image tag handling in build and deploy workflows

Changed

  • Raised the minimum supported client version to 1.8.0

  • Updated dependencies: unpinned numpy (now resolved transitively), bumped torch lower bound to >=2.4.0

  • Temporarily downgraded torch and cellarium-ml for compatibility, then re-pinned to stable versions

1.8.3 - 2026-04-15

Added

  • Added memory_budget configuration option for the TileDB vector index

Fixed

  • Fixed incorrect batch key handling in model inference (#212, #213)

Changed

  • Updated Cloud Run deployment configuration

1.8.2 - 2026-04-07

Added

  • Added TileDB Vector Search integration as the vector search backend (#209)

  • Added support for querying cell metadata directly from the SOMA table

  • Added support for multiple ontologies in cell annotation requests (#211)

Removed

  • Removed Vertex AI Matching Engine client and all related vector search code (#209)

Changed

  • Updated minimum required client version to 1.7.5

  • Refactored vector search layer with a factory pattern and type aliases

1.8.1 - 2026-03-20

Fixed

  • Fixed Read the Docs configuration for Python 3.12 (#207)

1.8.0 - 2026-03-20

Added

  • Added pyproject.toml and Poetry-based dependency management for the project

Changed

  • Restructured the codebase from the legacy casp layout into cellarium.cas_backend with separate apps and core modules

  • Renamed the main API service to CAS Compute and updated related configuration, documentation, and templates

  • Upgraded the runtime and container stack to Python 3.12, simplified Docker base images, and refreshed entrypoints

  • Reworked GitHub Actions and deployment workflows around service-specific build and deploy inputs

  • Replaced the previous formatting and linting toolchain with Ruff and moved away from the older tox-driven flow

Fixed

  • Fixed environment loading, import paths, and SQL template locations after the package restructuring

  • Resolved remaining refactor regressions in tests, Cloud Run deployment wiring, and floating-point assertions

1.7.3 - 2025-03-20

Changed

  • Updated the default Cloud Run deployment configuration

  • Optimized Dockerfile.ns.pytorch to build a faster deployment image

1.7.2 - 2025-03-19

Added

  • Added dedicated unit tests for the reimplemented summary statistics strategy

Changed

  • Reimplemented the summary statistics consensus strategy used by cell operations

  • Refactored API service construction to use dependency injection consistently

  • Refreshed the unit-test lifecycle and SQLite-backed test database setup

Fixed

  • Locked starlette and updated aiohttp and the PyTorch Docker image dependencies to restore compatibility

1.7.1 - 2024-11-13

Added

  • Added more filtering criteria to SQL-template-driven BigQuery extraction workflows

Changed

  • Refactored the bq_ops scripts and refreshed related documentation

  • Updated dependencies to keep the audit workflow passing

Fixed

  • Fixed non-random sampling behavior in extract workflows

  • Fixed a timestamp handling bug affecting data processing

1.7.0 - 2024-10-25

Changed

  • Lowered the initial user quota defaults while allowing quota increases through the REST API

  • New users now default to non-admin accounts when created through the admin flow

  • Updated the release version and minimum supported client version

  • Refreshed the welcome email templates sent to new users

1.6.2 - 2024-10-21

Fixed

  • Pinned the WTForms dependency to restore reliable admin installs

1.6.1 - 2024-10-21

Changed

  • Updated the welcome and new-key email templates

Fixed

  • User activity totals now default to 0 instead of None for processed cells and requests

1.6.0 - 2024-10-09

Added

  • Added lifetime cell quotas, including database support and service logic for tracking them

Changed

  • Raised the minimum supported client version for the lifetime-quota release

1.4.7 - 2024-09-06

Added

  • Added a client compatibility endpoint so clients can check whether their version is supported

  • Added a script for building the cell type ontology resources used by the consensus engine

  • Added pip-audit checks to CI to catch vulnerable dependencies earlier

Changed

  • Merged the model inference service into the API service and cached model assets to reduce repeated downloads

  • Updated onboarding emails and raised the default weekly quota to 100K cells

  • Allowed each deployment to set its own Sentry environment and refreshed release documentation

Fixed

  • Reduced Sentry noise from infrastructure-level failures and pinned dependency versions with breaking changes

  • Added stricter annotate input validation and removed redundant AnnData reads in the request flow

1.4.6 - 2024-07-03

Added

  • Added user feedback capture, including redirect and opt-out support

  • Added support for propagating x-client-session-id through request context and logs

  • Added an integration-test workflow that boots a test server and exercises the API end to end

  • Added description fields for models and indexes in the database schema

Changed

  • Made deployed index IDs optional to support more flexible index configuration

  • Updated the cell-related models and removed token display from the admin UI

  • Updated deployment labels, email defaults, and Cloud SQL connector configuration

Fixed

  • Resolved the multiple-head Alembic migration issue

  • Fixed admin database connection management and user access logging

1.4.5 - 2024-06-04

Added

  • Added opaque API keys with admin flows for creating, emailing, and authenticating user keys

  • Added centralized logging and tracing support across services and downstream model calls

  • Added weekly user cell quotas and quota-aware API behavior

  • Added GitHub Actions support for building and deploying Cloud Run services

  • Added support for connecting to PostgreSQL over VPC in cloud environments

Changed

  • Centralized application configuration and generalized API exception handling

  • Moved the development .env file under src/settings and updated deployment and secrets documentation

  • Reworked Docker builds so secrets are no longer baked into images

Fixed

  • Fixed logging credential loading and Sentry context setup order

  • Fixed email template loading and related authentication edge cases

1.4.4 - 2024-05-02

Added

  • Added REST API support for vector search indexes through a dedicated matching client layer

  • Added request and cells-processed activity tracking in a dedicated user activity table

  • Added the first consensus engine implementation, including ontology-aware strategies

  • Added cell metadata and index-related database models plus supporting admin tooling

Changed

  • Switched index search calls to an async REST client and stopped requesting unnecessary feature vectors

  • Expanded tests and coverage around matching, consensus, and cell operations behavior

Fixed

  • Fixed repeated matching-client initialization during search requests

  • Fixed an ontology-aware strategy failure caused by missing temporary tables

1.4.2 - 2024-02-13

Added

  • Added Documentation in rst format for Sphinx build in docs/ directory

Changed

  • src/casp/services/README.md is now casp/services/README.rst

  • src/casp/services/admin/README.md is now casp/services/admin/README.rst

  • src/casp/services/api/README.md is now casp/services/api/README.rst

  • src/casp/services/model_inference/README.md is now casp/services/model_inference/README.rst

  • src/casp/services/db/README.md is now casp/services/db/README.rst

  • src/casp/services/wdl_workflows/README.md is now casp/services/wdl_workflows/README.rst

  • docs/VERTEX_AI_MATCHING_ENGINE.md is now docs/vertex_ai_matching_engine.rst

  • Requests to matching engine are now made in batches of 5 to avoid overloading Vertex API

  • Requests to matching engine are now made with retry logic to account for errors caused by rapid increases in traffic