Template Task

The Template Task generates text from a template and task inputs. Templates can be used to prepare data for a number of tasks including generating large language model (LLM) prompts.

Example

The following shows a simple example using this task as part of a workflow.

from txtai.workflow import TemplateTask, Workflow

workflow = Workflow([TemplateTask(template="This is a {text} task")])
workflow([{"text": "template"}])

Configuration-driven example

This task can also be created with workflow configuration.

workflow:
  tasks:
    - task: template
      template: This is a {text} task

Methods

Python documentation for the task.

__init__(action=None, select=None, unpack=True, column=None, merge='hstack', initialize=None, finalize=None, concurrency=None, onetomany=True, **kwargs)

Creates a new task. A task defines two methods, type of data it accepts and the action to execute for each data element. Action is a callable function or list of callable functions.

Parameters:

Name	Type	Description	Default
action		action(s) to execute on each data element	None
select		filter(s) used to select data to process	None
unpack		if data elements should be unpacked or unwrapped from (id, data, tag) tuples	True
column		column index to select if element is a tuple, defaults to all	None
merge		merge mode for joining multi-action outputs, defaults to hstack	'hstack'
initialize		action to execute before processing	None
finalize		action to execute after processing	None
concurrency		sets concurrency method when execute instance available valid values: "thread" for thread-based concurrency, "process" for process-based concurrency	None
onetomany		if one-to-many data transformations should be enabled, defaults to True	True
kwargs		additional keyword arguments	{}

Source code in txtai/workflow/task/base.py

def __init__(
    self,
    action=None,
    select=None,
    unpack=True,
    column=None,
    merge="hstack",
    initialize=None,
    finalize=None,
    concurrency=None,
    onetomany=True,
    **kwargs,
):
    """
    Creates a new task. A task defines two methods, type of data it accepts and the action to execute
    for each data element. Action is a callable function or list of callable functions.

    Args:
        action: action(s) to execute on each data element
        select: filter(s) used to select data to process
        unpack: if data elements should be unpacked or unwrapped from (id, data, tag) tuples
        column: column index to select if element is a tuple, defaults to all
        merge: merge mode for joining multi-action outputs, defaults to hstack
        initialize: action to execute before processing
        finalize: action to execute after processing
        concurrency: sets concurrency method when execute instance available
                     valid values: "thread" for thread-based concurrency, "process" for process-based concurrency
        onetomany: if one-to-many data transformations should be enabled, defaults to True
        kwargs: additional keyword arguments
    """

    # Standardize into list of actions
    if not action:
        action = []
    elif not isinstance(action, list):
        action = [action]

    self.action = action
    self.select = select
    self.unpack = unpack
    self.column = column
    self.merge = merge
    self.initialize = initialize
    self.finalize = finalize
    self.concurrency = concurrency
    self.onetomany = onetomany

    # Check for custom registration. Adds additional instance members and validates required dependencies available.
    if hasattr(self, "register"):
        self.register(**kwargs)
    elif kwargs:
        # Raise error if additional keyword arguments passed in without register method
        kwargs = ", ".join(f"'{kw}'" for kw in kwargs)
        raise TypeError(f"__init__() got unexpected keyword arguments: {kwargs}")