Prompt Builder

Modules concerning the prompt builders used in Gen AI applications.

AgnosticPromptBuilder(system_template='', user_template='', separator='\n', ignore_extra_keys=False)

Bases: BasePromptBuilder

A prompt builder that is agnostic to specific model types and formats prompts with a customizable separator.

The AgnosticPromptBuilder class constructs a prompt by joining the content of the prompt templates using a specified separator. It is designed to work independently of specific model types.

Attributes:

Name	Type	Description
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}.
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}.
prompt_key_set	set[str]	A set of expected keys that must be present in the prompt templates.
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt.
formatter	AgnosticPromptFormatter	The formatter to be used to format the prompt into a string in the format_as_string method.

Initializes a new instance of the AgnosticPromptBuilder class.

Parameters:

Name	Type	Description	Default
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	''
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	''
separator	str	A string used to separate each prompt template's content. Defaults to "\n".	'\n'
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt. Defaults to False.	False

HuggingFacePromptBuilder(system_template='', user_template='', model_name_or_path='', ignore_extra_keys=False)

Bases: BasePromptBuilder

A prompt builder for Hugging Face models that support chat templating.

The HuggingFacePromptBuilder class is designed to create prompts compatible with models hosted in Hugging Face, given that the model support chat templating. It loads the appropriate tokenizer and formats the prompt by injecting placeholder values.

Attributes:

Name	Type	Description
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}.
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}.
prompt_key_set	set[str]	A set of expected keys that must be present in the prompt templates.
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt.
formatter	HuggingFacePromptFormatter	The formatter to be used to format the prompt into a string in the format_as_string method.

Note

If you're trying to access the prompt builder template of a gated model, you'd need to: 1. Request access to the gated repo using your HuggingFace account. 2. Login to HuggingFace in your system. This can be done as follows: 2.1. Install huggingface-hub: pip install huggingface-hub 2.2. Login to HuggingFace: huggingface-cli login 2.3. Enter your HuggingFace token.

Initializes a new instance of the HuggingFacePromptBuilder class.

Parameters:

Name	Type	Description	Default
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	''
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	''
model_name_or_path	str	The model name or path of the Hugging Face model tokenizer to be loaded. Defaults to an empty string.	''
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt. Defaults to False.	False

LlamaPromptBuilder(system_template, user_template, model_name='Meta-Llama-3.1-8B-Instruct', ignore_extra_keys=False)

Bases: HuggingFacePromptBuilder

A prompt builder for Llama models, ensuring compatibility and formatting of prompts.

The LlamaPromptBuilder class is designed to create prompts compatible with specific Llama models. It loads the appropriate tokenizer and formats the prompt by injecting placeholder values.

Attributes:

Name	Type	Description
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}.
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}.
prompt_key_set	set[str]	A set of expected keys that must be present in the prompt templates.
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt.
formatter	HuggingFacePromptFormatter	The formatter to be used to format the prompt into a string in the format_as_string method.

Note

If you're trying to access the prompt builder template of a gated model, you'd need to: 1. Request access to the gated repo using your HuggingFace account. 2. Login to HuggingFace in your system. This can be done as follows: 2.1. Install huggingface-hub: pip install huggingface-hub 2.2. Login to HuggingFace: huggingface-cli login 2.3. Enter your HuggingFace token.

Initializes a new instance of the LlamaPromptBuilder class.

Parameters:

Name	Type	Description	Default
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	required
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	required
model_name	str	The name of the Llama model tokenizer to be loaded. Defaults to Meta-Llama-3.1-8B-Instruct.	'Meta-Llama-3.1-8B-Instruct'
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt. Defaults to False.	False

MistralPromptBuilder(system_template='', user_template='', model_name='Mistral-7B-Instruct-v0.3', ignore_extra_keys=False)

Bases: HuggingFacePromptBuilder

A prompt builder for Mistral models, ensuring compatibility and formatting of prompts.

The MistralPromptBuilder class is designed to create prompts compatible with specific Mistral models. It loads the appropriate tokenizer and formats the prompt by injecting placeholder values.

Attributes:

Name	Type	Description
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}.
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}.
prompt_key_set	set[str]	A set of expected keys that must be present in the prompt templates.
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt.
formatter	HuggingFacePromptFormatter	The formatter to be used to format the prompt into a string in the format_as_string method.

Note

If you're trying to access the prompt builder template of a gated model, you'd need to: 1. Request access to the gated repo using your HuggingFace account. 2. Login to HuggingFace in your system. This can be done as follows: 2.1. Install huggingface-hub: pip install huggingface-hub 2.2. Login to HuggingFace: huggingface-cli login 2.3. Enter your HuggingFace token.

Initializes a new instance of the MistralPromptBuilder class.

Parameters:

Name	Type	Description	Default
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	''
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	''
model_name	str	The name of the Mistral model tokenizer to be loaded. Defaults to Mistral-7B-Instruct-v0.3.	'Mistral-7B-Instruct-v0.3'
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt. Defaults to False.	False

OpenAIPromptBuilder(system_template='', user_template='', ignore_extra_keys=False)

Bases: BasePromptBuilder

A prompt builder for OpenAI models, ensuring compatibility and formatting of prompts.

The OpenAIPromptBuilder class is designed to create prompts compatible with specific OpenAI models. It formats the prompt templates by prefixing each template with its role (converted to title case) and concatenating them into a single prompt string.

Attributes:

Name	Type	Description
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}.
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}.
prompt_key_set	set[str]	A set of expected keys that must be present in the prompt templates.
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt.
formatter	OpenAIPromptFormatter	The formatter to be used to format the prompt into a string in the format_as_string method.

Initializes a new instance of the OpenAIPromptBuilder class.

Parameters:

Name	Type	Description	Default
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	''
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	''
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt. Defaults to False.	False

PromptBuilder(system_template='', user_template='', ignore_extra_keys=False)

A prompt builder class used in Gen AI applications.

Attributes:

Name	Type	Description
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}.
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}.
prompt_key_set	set[str]	A set of expected keys that must be present in the prompt templates.
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt.

Initializes a new instance of the PromptBuilder class.

Parameters:

Name	Type	Description	Default
system_template	str	The system prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	''
user_template	str	The user prompt template. May contain placeholders enclosed in curly braces {}. Defaults to an empty string.	''
ignore_extra_keys	bool	Whether to ignore extra keys when formatting the prompt. Defaults to False.	False

Raises:

Type	Description
ValueError	If both system_template and user_template are empty.

format(history=None, extra_contents=None, attachments=None, **kwargs)

Formats the prompt templates into a MultimodalPrompt.

This method processes each prompt template, replacing the placeholders in the template content with the corresponding values from kwargs. If any required key is missing from kwargs, it raises a ValueError. It also handles the provided history and extra contents. It formats the prompt as a MultimodalPrompt.

Parameters:

Name	Type	Description	Default
history	MultimodalPrompt | None	The optional history to be included in the prompt. Defaults to None.	None
extra_contents	list[MultimodalContent] | None	The optional extra contents to be included in the user message. Defaults to None.	None
attachments	list[Attachment] | None	Deprecated parameter to handle attachments. Will be removed in v0.5.0. Defaults to None.	None
**kwargs	Any	A dictionary of placeholder values to be injected into the prompt templates. Values must be either a string or an object that can be serialized to a string.	{}

Returns:

Name	Type	Description
MultimodalPrompt	MultimodalPrompt	A multimodal prompt.

Raises:

Type	Description
ValueError	If a required key for the prompt template is missing from kwargs.