Types

Prompts Types

class AnthropicPrompt(messages, kwargs)

Bases: _FormattedPrompt

Represents a formatted prompt for Anthropic message models.

messages

A sequence of message parameters.

Type:

Sequence[MessageParam]

kwargs

Keyword arguments specific to Anthropic message model invocation.

Type:

AnthropicMessageModelKwargs

kwargs

messages

class GoogleGenerativeaiPrompt(messages: 'Sequence[protos.Content]', kwargs: 'GoogleModelKwargs')

Bases: _FormattedPrompt

kwargs

messages

class OpenAIPrompt(messages, kwargs)

Bases: _FormattedPrompt

Represents a formatted prompt for OpenAI chat completion models.

messages

A sequence of chat completion message parameters.

Type:

Sequence[ChatCompletionMessageParam]

kwargs

Keyword arguments specific to OpenAI chat completion model invocation.

Type:

OpenAIChatCompletionModelKwargs

kwargs

messages

class PromptVersion(prompt, /, *, model_name, description=None, model_provider='OPENAI', template_format='MUSTACHE')

Bases: object

Represents a version of a prompt for different model providers.

format(*, variables=MappingProxyType({}), formatter=None, sdk=None)

Formats the prompt for a specific SDK.

Parameters:

• variables (Mapping[str, str]) – A mapping of variable names to values to use in the prompt. Defaults to an empty mapping.

• formatter (Optional[TemplateFormatter]) – A custom template formatter to use for the prompt. Defaults to None.

• sdk (Optional[SDK]) – The SDK to format the prompt for. Defaults to None.

Returns:

The formatted prompt.

classmethod from_anthropic(obj, /, *, template_format='MUSTACHE', description=None, model_provider='ANTHROPIC')

Creates a prompt version from an Anthropic message model.

Parameters:

• obj (MessageCreateParamsBase) – The message create parameters.

• template_format (Literal["F_STRING", "MUSTACHE", "NONE"]) – The format of the template to use for the prompt. Defaults to “MUSTACHE”.

• description (Optional[str]) – A description of the prompt. Defaults to None.

• model_provider (Literal["ANTHROPIC"]) – The provider of the model to use for the prompt. Defaults to “ANTHROPIC”.

Returns:

The prompt version.

Return type:

PromptVersion

classmethod from_aws(obj, /, *, template_format='MUSTACHE', description=None, model_provider='AWS')

classmethod from_google_generativeai(obj, /, *, template_format='MUSTACHE', description=None, model_provider='GOOGLE')

classmethod from_openai(obj, /, *, template_format='MUSTACHE', description=None, model_provider='OPENAI')

Creates a prompt version from an OpenAI chat completion model.

Parameters:

• obj (CompletionCreateParamsBase) – The completion create parameters.

• template_format (Literal["F_STRING", "MUSTACHE", "NONE"]) – The format of the template to use for the prompt. Defaults to “MUSTACHE”.

• description (Optional[str]) – A description of the prompt. Defaults to None.

• model_provider (Literal["OPENAI", "AZURE_OPENAI", "DEEPSEEK", "XAI", "OLLAMA"]) – The provider of the model to use for the prompt. Defaults to “OPENAI”.

Returns:

The prompt version.

Return type:

PromptVersion

property id

Prompt Version ID if stored in the Phoenix backend

Spans Types

class Concatenation(key='', kwargs=<factory>, separator='\n\n')

Bases: object

Represents a concatenation operation in a span query.

key = ''

kwargs

separator = '\n\n'

to_dict()

class Explosion(key='', kwargs=<factory>, primary_index_key='context.span_id')

Bases: object

Represents an explosion operation in a span query.

key = ''

kwargs

primary_index_key = 'context.span_id'

to_dict()

class Projection(key='')

Bases: object

Represents a projection in a span query.

key = ''

to_dict()

Convert to dictionary format.

class SpanFilter(condition='', valid_eval_names=None)

Bases: object

Represents a filter condition in a span query.

condition = ''

to_dict()

valid_eval_names = None

class SpanQuery(_select=None, _filter=None, _explode=None, _concat=None, _rename=None, _index=None, _index_has_been_set=False)

Bases: object

Represents a query for spans using the query DSL.

concat(key, **kwargs)

Concatenate a field from the spans.

explode(key, **kwargs)

rename(**kwargs)

Rename fields in the result.

select(*fields)

to_dict()

where(condition)

Filter spans based on a condition.

with_index(key)