citrine.informatics.catalyst.assistant module

class citrine.informatics.catalyst.assistant.AssistantRequest(*, question: str, predictor: GraphPredictor, temperature: float | None = 0.0, language_model: LanguageModelChoice | None = LanguageModelChoice.GPT_4)

Bases: Serializable[AssistantRequest]

A query to the model assistant.

classmethod build(data: dict) → Self: Build an instance of this object from given data.

dump() → dict: Dump this instance.

language_model = 'gpt-4'

predictor = None

question = None

temperature = 0.0

class citrine.informatics.catalyst.assistant.AssistantResponse

Bases: PolymorphicSerializable[AssistantResponse]

The parent type for all Model Assistant responses.

classmethod build(data: dict) → SelfType: Build the underlying type.

classmethod get_type(data) → Type[AssistantResponse]: Return the subtype.

class citrine.informatics.catalyst.assistant.AssistantResponseConfig

Bases: Serializable[AssistantResponseConfig], AssistantResponse

A successful model assistant invocation, whose response includes a modified predictor.

classmethod build(data: dict) → Self: Build an instance of this object from given data.

dump() → dict: Dump this instance.

classmethod get_type(data) → Type[AssistantResponse]: Return the subtype.

predictor = None

class citrine.informatics.catalyst.assistant.AssistantResponseExecError

Bases: Serializable[AssistantResponseExecError], AssistantResponse

A failed model assistant invocation, due to some internal issue.

This most likely indicates the assistant got some unexpected output when asking its query. It indicates an internal error, not a problem with the client invocation.

classmethod build(data: dict) → Self: Build an instance of this object from given data.

dump() → dict: Dump this instance.

classmethod get_type(data) → Type[AssistantResponse]: Return the subtype.

error = None

class citrine.informatics.catalyst.assistant.AssistantResponseInputError

Bases: Serializable[AssistantResponseInputError], AssistantResponse

A single input failure.

Contains the error message, and the field it applies to.

classmethod build(data: dict) → Self: Build an instance of this object from given data.

dump() → dict: Dump this instance.

classmethod get_type(data) → Type[AssistantResponse]: Return the subtype.

error = None

field = None

class citrine.informatics.catalyst.assistant.AssistantResponseInputErrors

Bases: Serializable[AssistantResponseInputErrors], AssistantResponse

A failed model assistant invocation, due to malformed input.

This should only happen if there’s some field omitted by the client, or one of its values is outside acceptable ranges.

classmethod build(data: dict) → Self: Build an instance of this object from given data.

dump() → dict: Dump this instance.

classmethod get_type(data) → Type[AssistantResponse]: Return the subtype.

errors = None

class citrine.informatics.catalyst.assistant.AssistantResponseMessage

Bases: Serializable[AssistantResponseMessage], AssistantResponse

A successful model assistant invocation, whose response is only text.

classmethod build(data: dict) → Self: Build an instance of this object from given data.

dump() → dict: Dump this instance.

classmethod get_type(data) → Type[AssistantResponse]: Return the subtype.

message = None

class citrine.informatics.catalyst.assistant.AssistantResponseUnsupported

Bases: Serializable[AssistantResponseUnsupported], AssistantResponse

A successful model assistant invocation, but for an unsupported query.

This will cover any user query which the model assistant could not map to a functionality it supports. For example, asking a generic question, or something unrelated to materials.

classmethod build(data: dict) → Self: Build an instance of this object from given data.

dump() → dict: Dump this instance.

classmethod get_type(data) → Type[AssistantResponse]: Return the subtype.

message = None