README.md

metadata

license: mit

Table of Contents

• __init__
• DemonstrationsAtomicFlow
  • DemonstrationsAtomicFlow
    • instantiate_from_config
    • run
• ChatWithDemonstrationsFlow
  • ChatWithDemonstrationsFlow

__init__

DemonstrationsAtomicFlow

DemonstrationsAtomicFlow Objects

class DemonstrationsAtomicFlow(AtomicFlow)

This class implements a Demonstrations Atomic Flow. It is a flow which is usually used to pass demonstrations (of user assistant interactions)

to the ChatAtomicFlow.

Configuration Parameters:

• name (str): The name of the flow. Default: "DemonstrationsAtomicFlow"
• description (str): A description of the flow. This description is used to generate the help message of the flow. Default: "A flow that passes demonstrations to the ChatFlow"
• data (List[Dict[str, Any]]): The data of the demonstrations. If data is None, the data is loaded from the file specified in the params["data_dir"]. Default: No default value this field must be set.
• params (Dict[str, Any]): The parameters specific to the dataset of the demonstrations. Its default parameters are:
  • data_dir (str): The directory where the demonstrations are stored. If the data is not directly passed to the flow through data then the data is loaded from this directory. Default: No default value this field must be set.
  • demonstrations_id (str): The id of the demonstrations (name of the data file). If the data is not directly passed to the flow through data then the data is loaded from this file. Default: No default value this field must be set.
  • demonstrations_k (int): The number of demonstrations to pass to the ChatFlow. If None, all the demonstrations are passed to the ChatFlow. Default: None
• query_prompt_template (Dict[str, Any]): The prompt template used to generate the query of the demonstrations. By default its of type flows.prompt_template.JinjaPrompt. None of the parameters of the prompt are defined by default and therefore need to be defined if one wants to use the query_prompt_template. Default parameters are defined in flows.prompt_template.jinja2_prompts.JinjaPrompt.
• response_prompt_template (Dict[str, Any]): The prompt template used to generate the response of the demonstrations. By default its of type flows.prompt_template.JinjaPrompt. None of the parameters of the prompt are defined by default and therefore need to be defined if one wants to use the response_prompt_template. Default parameters are defined in flows.prompt_template.jinja2_prompts.JinjaPrompt.

Input Interface:

• The input interface expected by its successor flow (e.g. typically ChatAtomicFlow so the input interface is the one expected by ChatAtomicFlow)

Output Interface:

• The input interface expected by its successor flow (e.g. typically ChatAtomicFlow so the input interface expected by ChatAtomicFlow))
• demonstrations (List[Dict[str, Any]]): A list of demonstrations. Each demonstration is a dictionary with the following keys: - idx (int): The index of the demonstration - query (str): The query of the demonstration - response (str): The response of the demonstration

Arguments:

• params (Dict[str, Any]): The parameters specific to the dataset of the demonstrations. It must sould contain the following keys:
• 'data_dir' (str): The directory where the demonstrations are stored. This field is used if the data is not directly passed to the flow through the 'data' field.
• 'demonstrations_id' (str): The id of the demonstrations (name of the data file). This field is used if the data is not directly passed to the flow through the 'data' field.
• 'demonstrations_k' (int): The number of demonstrations to pass to the ChatFlow. If None, all the demonstrations are passed to the ChatFlow.
• 'ids_to_keep' (Optional[Union[str, List[str]]]): The ids of the demonstrations to keep. If None, all the demonstrations are kept.
• query_prompt_template (JinjaPrompt): The prompt template used to generate the query of the demonstrations.
• response_prompt_template (JinjaPrompt): The prompt template used to generate the response of the demonstrations.
• data (Optional[List[Dict[str, Any]]]): The data of the demonstrations. If None, the data is loaded from the file specified in the params.

instantiate_from_config

@classmethod
def instantiate_from_config(cls, config)

This method instantiates the flow from a config file.

Arguments:

• config (Dict[str, Any]): The configuration of the flow.

Returns:

Flow: The instantiated flow.

run

def run(input_data: Dict[str, Any]) -> Dict[str, Any]

This method runs the flow. It returns the input data of the flow with the demonstrations added to it.

Arguments:

• input_data (Dict[str, Any]): The input data of the flow.

Returns:

Dict[str, Any]: The input data of the flow with the demonstrations added to it.

ChatWithDemonstrationsFlow

ChatWithDemonstrationsFlow Objects

class ChatWithDemonstrationsFlow(SequentialFlow)

A Chat with Demonstrations Flow. It is a flow that consists of multiple sub-flows that are executed sequentially.

It's parent class is SequentialFlow.

It Contains the following subflows:
    - A Demonstration Flow: It is a flow that passes demonstrations to the ChatFlow
    - A Chat Flow: It is a flow that uses the demonstrations to answer queries asked by the user/human.

An illustration of the flow is as follows:

        -------> Demonstration Flow -------> Chat Flow ------->

Configuration Parameters:

• name (str): The name of the flow. Default: "ChatAtomic_Flow_with_Demonstrations"
• description (str): A description of the flow. This description is used to generate the help message of the flow. Default: "A sequential flow that answers questions with demonstrations"
• subflows_config (Dict[str,Any]): A dictionary of subflows configurations of the sequential Flow. Default:
  • Demonstration Flow: The configuration of the Demonstration Flow. By default, it a DemonstrationsAtomicFlow. Its default parmaters are defined in DemonstrationsAtomicFlow).
  • Chat Flow: The configuration of the Chat Flow. By default, its a ChatAtomicFlow. Its default parmaters are defined in ChatAtomicFlowModule (see Flowcard, i.e. README.md, of ChatAtomicFlowModule).
• topology (str): The topology of the flow which is "sequential". By default, the topology is the one shown in the illustration above (the topology is also described in ChatWithDemonstrationsFlow.yaml).

Input Interface:

• query (str): A query asked to the flow (e.g. "What is the capital of France?")

Output Interface:

• answer (str): The answer of the flow to the query

Arguments:

• \**kwargs: Arguments to be passed to the parent class SequentialFlow constructor.