CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

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

unreleased

0.14.5 - 2025-02-14

Fixed

• Random 410 errors caused (presumably) by using a BackgroundTask in the /compute endpoint to set the task_id in celery's backend. I suspect that when users query /compute/output/{task_id} right after submitting a big batch it's possible the BackgroundTask hasn't set the task_id in celery's backend yet and so the spurious 410 is returned due to a race condition.

0.14.4 - 2025-02-08

Fixed

• Removed type aliasing of ProgramOutput generic types that was required for earlier versions of pydantic but was now causing SinglePointResults objects (or ProgramInput objects) to get consumed as Files (or FileInput) objects since that type was listed first and could consume the SinglePointResults data.

0.14.3 - 2025-02-08

Changed

• Removed black and isort in favor of ruff.
• Upgraded to Python 3.13 for docker container.
• Moved code quality checks from CircleCI to GitHub actions.
• Fixed a few httpx calls to use the new API (content= vs data= for text/bytes data).

0.14.2 - 2024-09-12

Changed

• Upgraded BigChem 0.9.0 -> 0.10.0.
• Upgraded poetry.lock package dependencies.

Added

• crest to SupportedPrograms.

0.14.1 - 2024-08-07

Changed

• Updated to qcio ^0.11.7 which removes NoResult and moves .files from ProgramOutput to ProgramOutput.results.

0.14.0 - 2024-07-12

Changed

• Updated qcio (0.10.1 -> 0.10.2). Structure.ids -> Structure.identifiers
• Updated qcop (0.7.1 -> 0.7.3)
• Updated bigchem (0.8.0 -> 0.8.1)

0.13.0 - 2024-07-10

Changed

• 🚨 Updated to BigChem 0.8.0 which uses new qcio Structure in place of Molecule.

0.12.0 - 2024-06-14

Changed

• Updated to BigChem 0.7.2 which uses latest Generics structures from qcio. Not a breaking change since the json representation of these objects is all the same as before; however, will bump major version since this feels like a shift to core data structures.
• 🚨 Renamed top-level argument to /compute collect_wavefunction to collect_wfn to match qcop nomenclature.
• Updated a number of dependencies and subpackages to latest versions. (poetry lock)
• Modernized typing syntax by removing Dict, List, Union declaration in favor of dict, list, |.
• Renamed TaskState -> TaskStatus
• Renamed Output to ProgramOutputWrapper.
  • 🚨 Renamed attributes from state -> status and result -> program_output.

0.11.6 - 2024-03-16

Changed

• Updated bigchem to 0.6.5 which incorporates qcop 0.5.0 which raises exceptions by default and therefore changed the compute signature in BigChem by removing the raise_exc=True default argument.
• Updated black from `23.x.x -> 24.x.x'.

0.11.5 - 2024-01-12

Fixed

• Updated bigchem to 0.6.4 which incorporates qcop update that properly captures geometric exceptions.

0.11.4 - 2023-09-22

Added

• Added test for propagate_wfn passed to program adapters that do not support it. An updated in qcop and bigchem resolved this issue resulting in a ProgramFailure object being correctly added to the exception so it can be returned in the /compute/output/{task_id} endpoint.

Changed

• Updated dashboard styling to be more modern.

0.11.3 - 2023-09-20

Changed

• Updated BigChem from 0.5.x -> 0.6.x to account for renaming of DualProgramArgs to SubProgramArgs in qcio package.

0.11.2 - 2023-09-19

Fixed

• Fixed bug with /v1/ -> /v2/ change in API prefix and Auth0 callbacks.

Changed

• Changed exception handing in /compute/output/ endpoint to use celery result.traceback rather than traceback.format_exc() to capture the full traceback of the error.

0.11.1 - 2023-09-16

Fixed

• Fixed /compute/output/{task_id} endpoint to properly return *Output or ProgramFailure objects for each calculation in a list of inputs.

Removed

• Removed distinction between Result and ResultGroup and created a single output object for the /compute/output/{task_id} endpoint, Output.

0.11.0 - 2023-09-08

Added

• Typos to pre-commit.
• pydantic-settings packages as part of pydantic v1 -> v2 upgrade.
• Added x-max_batch_inputs to OpenAPI schema so clients can query the max number of inputs they can send in a list.

Changed

• Removed QCElemental models in favor of qcio.
• Upgraded from pydantic v1 -> v2.
• Updated BigChem to latest version 0.5.x and running jobs on the premise that exceptions will be raised by failed tasks.
• Api prefix updated from /api/v1/ -> /api/v2/
• /compute/results/ updated to /compute/output/ to more closely match qcop nomenclature of "inputs" and "outputs."
• Now returning the celery state directly as the result.status value. Need to decide how to handle status and result returning in the future (I think it can be significantly simplified), but for now this setup works.
• Always install the latest version of poetry in the docker image.
• Build BigChem in /opt.

Removed

• mypy.ini and setup.cfg and moved all configuration to pyproject.toml.
• flake8 in favor of ruff.
• compute-procedure endpoint as it's not longer needed with qcio.
• Dropped git, python3-dev, and gcc from main docker image since psutil is no longer required (was required by qcengine)

0.10.1 - 2023-02-20

Changed

• Running BigChem containers from mtzgroup rather than coltonbh

0.10.0 - 2023-02-20

Changed

• Changed pipenv for poetry
• Updated python 3.9 -> 3.11
• Updated tests to only depend upon httpx and removed requests library
• Updated BigChem to 0.4.0 and installing from PyPi now

0.9.1 - 2022-12-27

Changed

• Updated BigChem to 0.3.0 released on PyPi instead of building from GitHub.
• Updated a bunch of packages in Pipfile.lock
• Updated dependencies for isort and added explicit Optional[] typing for mypy

0.9.0 - 2022-07-19

Changed

• Changed projects name from QCCloud to ChemCloud
• Updated bigqc to bigchem

0.8.1 - 2022-07-15

Changed

• Changed distributed algorithms from being called qcc algorithms to bigqc algorithms to correctly identify where they are coming from.

0.8.0 - 2022-07-14

Changed

• Updated project name from TeraChem Cloud to Quantum Chemistry Cloud

0.7.2 - 2022-07-11

Changed

• Pass only strings for default values to compute_tcc function that executes distributed algorithms. Passing Enum values caused celery to try and deserialize objects containing references to the terachem_cloud package where the Enums were defined.

0.7.1 - 2022-07-12

Changed

• Update to BigQCv0.1.3

0.7.0 - 2022-06-14

Changed

• Refactored all backend workers and middleware to a separate out the BigQC project. This project now comprises just a web layer on top of the BigQC engine. This change resulted in the need to change a few environment variables to reflect the change to BigQC. CELERY_BROKER_CONNECTION_STRING -> BIGQC_BROKER_URL and CELERY_BACKEND_CONNECTION_STRING -> BIGQC_BACKEND_URL.
• Changed dependencies to reflect the refactor.

0.6.2 - 2022-06-02

Changed

• Removed web server memory constraints at docker level. Large results were crashing server because of 500MB limit. When (300MB+) results were pulled into memory server would require >500MB memory and would be terminated by process manager.
• Increased gunicorn worker timeout from 30s -> 60s to accommodate larger results.
• Set results to expire in 3 days instead of default 1 day in backend store.

0.6.1 - 2022-04-25

Added

• Patched geomeTRIC to enable transition=True for json inputs used in QCEngine

0.6.0 - 2022-04-02

Changed

• /compute/result changed to GET request with signature /compute/result/{task_id}
• Removed much unnecessary internal code around managing Task state due to saving task structure to the DB so that clients only have to supply a single task ID regardless if the task is a batch or single calculation.

0.5.1 - 2022-03-27

Added

• Base64 encode bytes input values submitted to terachem_fe by client

0.5.0 - 2022-03-26

Added

• 🚨BREAKING CHANGE🚨: Support for native_files protocol that enables end users to retrieves files generated by a QC package during computation. E.g., can retrieve c0 files from a TeraChem computation. Only a breaking change if end user has an old version of qcelemental that does not support the native_files protocol and output field.
• Users can upload c0 files to seed a TeraChem computation with a wave function.
• TeraChem Frontend containers to XStream deployment to access files created by TeraChem job.

Changed

• 🚨BREAKING CHANGE🚨: Supported Engine for TeraChem changed from terachem_pbs -> terachem_fe
• 🚨BREAKING CHANGE🚨: tcc_kwargs -> tcc:keywords
• Pegged versions of psi4, rdkit, and xtb-python. Updated qcelemental to work with psi4v1.5. Updated python packages without pegged versions.
• Updated app to run python3.9.
• Updated CircleCI python3.7->3.9.
• Updated XStream TeraChem workers to new image.

0.4.2 - 2021-06-11

Added

• "tcc" compute engine
• Parallel hessian and parallel frequency analysis methods to tcc engine

Changed

• Changed celery serializer from json to pickle. This allows chained methods to receive python data types directly, rather than serialized representations. Since users do not have direct access to the celery layer and can only submit json serialized (and validated) objects via the API I do not feel there are security risks.

0.4.1 - 2021-06-07

Added

• Private queues for compute

0.4.0 - 2021-06-04

Added

• Batch compute capabilities to /compute and /compute-procedure

Changed

• 🚨BREAKING CHANGE🚨: Updated /compute* endpoints to return Task objects including GroupTask objects that may contain subtasks instead of returning a single str of the task_id
• 🚨BREAKING CHANGE🚨: Updated /result endpoint to accept Task objects instead of just a str of task_id. This included updating the endpoint to be a POST instead of GET endpoint.

0.3.5 - 2021-05-26

Added

• xtb to compute engines
• Added cached-property package since kombu didn't install it by default as a dependency into the linux docker container.

Changed

• Updated tcpb>=0.8.0
• Updated celery version

0.3.4 - 2021-05-21

Added

• geomeTRIC optimizer
• rdkit engine (adds force fields which can be used for single point computations or in optimizers)

Changed

• Run compute tests on hydrogen instead of water for faster results

0.3.3 - 2021-05-19

Added

• compute_procedure task for doing geometry optimizations using pyberny
• VSCode settings file added to repo.

Changed

• Combined two commands from the CI/CD pipeline used against XStream to get fewer auth errors
• Updated mypy definitions on Settings object to pass mypy checks while still allowing tests to run on CircleCi without needing Auth0 configuration. I.e., the web application can be run without needing Auth0 config (for testing purposes).
• pre-commit mypy now loads from GitHub repo instead of local mypy install.

0.3.2 - 2021-04-06

Added

• Test assertions to check that results are deleted from celery backend after retrieval by client.

Changed

• Added flags --without-heartbeat --without-mingle --without-gossip to celery workers to reduce network overhead chattiness.
• Delete celery results from backend upon retrieval by client. This will prevent Redis from becoming a memory hog at times of high computational load.

0.3.1 - 2021-03-10

Added

• Changelog
• Updated tcpb package to >=0.7.1 to enable molden file creation using imd_orbital_type="whole_c" keyword.

0.3.0 - 2021-01-26

Added

• Added TeraChem Protocol Buffer server to available compute engines.
• XStream deployment for both dev and prod now require GPUs to power Terachem. Deployment config for dev/prod was split into docker-compose.xstream.dev.yaml and docker-compose.xstream.prod.yaml.
• TeraChem license.key docker secret to swarm on XStream
• architecture.md to /docs/development to document core architectural decisions.
• Links on user dashboard to qccloud python client, /logout, and a brief description of how to change password.
• Links to users/dashboard on the main documentation page.
• Forgotten __init__.py to terachem_cloud package. Added __version__ to file. This causes mypy checks to fail as they were previously not inspecting this package fully due to missing __init__.py.
• TaskStatus enum to hold task status values.

Changed

• /compute/result/{task_id} can return either AtomicResult or FailedOperation objects.
• CircleCi build pipelines to only include a single build step for web and workers instead of a split pipeline for dev/prod
• Can approve CircleCi build and deploy steps upfront without having to wait for tests to pass. Build/deploy will still only occur if tests pass.
• XStream stacks for dev/prod no longer derive from the same template file.
• Changed CeleryAtomicResult to TaskResult and change atomic_result attribute to just result to note that result may now be an AtomicResult or a FailedOperation (a result could either of these data types).

0.2.1 - 2021-01-21

Removed

• Removed docker_layer_caching from CircleCi build process

0.2.0 - 2021-01-21

Added

• refresh_flow to /oauth/token endpoint

0.1.0 - 2020-12-18

Added

• Three endpoints all available on /api/v1:
  1. /oauth/token to get JWTs,
  2. /compute to submit AtomicInput compute request,
  3. /compute/result/{task_id} to request a result delivered as an AtomicResult object.
• Auth provided by Auth0.