Citrine

**What's new:**

- A new interface for interacting with design workflows (502). Design workflows can now be created and inspected through the `DesignWorkflowCollection`, and executions are contained in a `DesignExecutionCollection`. Design workflow executions have a `.candidates` [method](https://github.com/CitrineInformatics/citrine-python/blob/master/src/citrine/resources/workflow_executions.py#L82), which returns structured `DesignCandidate` objects instead of a JSON blob. Design candidates are paginated for easy access. The old method of retrieving design candidates, `.results`, still functions and is now paginated as well.
- Added unique name support for datasets (512). A dataset can optionally be given a globally unique name, which makes retrieving that dataset faster.


**Improvements:**

- Added exact range bounds to csv columns (503). This information is filled in automatically by the server when processing a CSV file.
- Reorg and fix of documentation, especially documentation on how to get started with citrine-python (510, 507, 508, 506).


**Fixes:**

- Fix type hint for categorical constraint (515). This allows categorical constraints to be used in design workflows.


**Deprecated**

- Descriptors without units have been deprecated (411). Please include units when defining all RealDescriptors. If the value is dimensionless, use an empty string as the units.
- The unused `threshold` parameter has been removed from ChemicalFormulaDescriptor (411).

This release includes two new features: the ability to check for updates to predictors and the ability to create more expressive product design spaces.

**What's new:**

- Added the ability to check for [predictor updates](https://github.com/CitrineInformatics/citrine-python/pull/499/files). This can be used to detect the existence of new training data, which can then be used to update the predictor.
- A new feature of [product design spaces](https://github.com/CitrineInformatics/citrine-python/pull/497): subspaces! A product design space can now be defined as the product of _other_ design spaces as well as univariate dimensions. This makes the product design space more expressive and capable of solving a wide variety of problems. For example, a typical mixing process can now be represented as a product of a formulation design space, to capture the ingredients that can be mixed together, and univariate dimensions to describe the parameters of the mixing process.

**Improvements:**

- Improved [documentation](https://github.com/CitrineInformatics/citrine-python/pull/493) of formulation design spaces and constraints
- Client and documentation [version](https://github.com/CitrineInformatics/citrine-python/pull/495) are now synchronized in a single `__version__.py` file.
- Several bug fixes (492 491 501)

This release includes client support for estimating `Bounds` for columns of data in hosted CSV files, migration utilities for up-converting `EnumeratedDesignSpace`s into `DataSourceDesignSpaces`, and improvements to documentation and argument validation.

**What's new:**

- The `dataset.files.process` method now includes information about what `Bound` might be appropriate for the data in each column (https://github.com/CitrineInformatics/citrine-python/pull/489)
- Two migration methods that up-convert `EnumeratedDesignSpace`s to `DataSourceDesignSpace`, either locally or on a platform (https://github.com/CitrineInformatics/citrine-python/pull/487)

**Improvements:**

- The AI Engine documentation makes better use of utility methods from our SDK. (https://github.com/CitrineInformatics/citrine-python/pull/483)
- The `IngredientQuantityByProcessAndName` class validates that the present of a `unit` is consistent with the type of quantity (https://github.com/CitrineInformatics/citrine-python/pull/490)

This release adds a formulation design space and associated ingredient and label constraints.

**What's new:**

- `FormulationDesignSpace` defines a formulation domain from a set of ingredient names, ingredient labels and formulation constraints.
- There are 3 new formulation constraints:
1. `IngredientFractionConstraint` limits the total ingredient fraction in a formulation. It can be used in a `FormulationDesignSpace` to restrict formulations or with a `Score` to penalize candidates that do not satisfy the constraint.
2. `LabelFractionConstraint` limits the total fraction of ingredients with a given label in a formulation.
3. `IngredientCountConstraint` limits the total number of ingredients in a single formulation. If a label is specified, only ingredients with the label will count towards the total.

This is a maintenance release that fixes a bug in the json support module that hindered feature development.

Improvements:
* Default values of any type can be set when defining json properties (https://github.com/CitrineInformatics/citrine-python/pull/485)

This release includes initial Alpha-level support for deep asynchronous validations on data object updates.

**What’s New**

* Alpha-level support for deep asynchronous validations on data object updates. This allows us to support editing of templates, specifically the narrowing of the template bounds and the allowed names and labels on a template. The relevant data will be checked to ensure the update is valid relative to existing data. This also includes a dry run option to check if the update would succeed.