fabricatio-question 0.1.7

An extension of fabricatio, which provide the capability to question user to make better planning and etc.

fabricatio-question

MIT

Interactive questioning extension for fabricatio agents. Generates LLM-driven selection prompts and presents them to users via terminal UI, enabling agents to gather structured feedback during planning and decision-making.

Installation

pip install fabricatio[question]
# or
uv pip install fabricatio[question]

For a full installation including all fabricatio components:

pip install fabricatio[full]

API

Capability: Questioning

Extends fabricatio_core.capabilities.propose.Propose. Mix into an agent to add interactive questioning.

Method	Description
async selection(q: str, k: int = 1, **kwargs) -> str | List[str]	Generates a selection question via LLM, presents it interactively, and returns the user's choice(s). k=1 returns a single string; k>1 returns a list of up to k selections.
async selection_string(q: str, k: int = 1, **kwargs) -> str	Like selection(), but returns a formatted string with the question topic, options, and selected indices.

Model: SelectionQuestion

A SketchedAble model representing an interactive selection prompt. Generated by the LLM from a user-provided topic.

Field	Type	Description
q	str	Question text displayed to the user.
option	List[str]	Available choices.

Method	Description
async single() -> str	Presents a single-choice selection via questionary.select.
async multiple(k: int = 0) -> List[str]	Presents a multi-choice checkbox. When k > 0, validates that exactly k options are selected.

Utilities

fabricatio_question.utils

Function	Description
async ask_retain(candidates: List[str]) -> List[str]	Presents a checkbox list with all candidates pre-checked, returning the user's retained subset.
async ask_retain(candidates, value_mapping) -> List[V]	Overload that maps retained choices back through value_mapping.
async ask_edit(text_seq: List[str]) -> List[str]	Prompts the user to edit each string sequentially; only non-empty edits are returned.

Configuration

fabricatio_question.config.question_config — a QuestionConfig dataclass with template names used during question generation:

Field	Default	Description
selection_template	"built-in/selection"	Template for generating the SelectionQuestion prompt.
selection_display_template	"built-in/selection_display"	Template for rendering selection_string output.

Usage

from fabricatio_question.capabilities.questioning import Questioning

class MyAgent(Questioning):
    ...

async def example():
    agent = MyAgent()

    # Single choice — returns a str
    choice: str = await agent.selection("Pick a database backend")

    # Multiple choices — returns list[str]
    picks: list[str] = await agent.selection("Select features to enable", k=3)

    # Get a formatted response string
    result: str = await agent.selection_string("Choose a deployment target")

from fabricatio_question.utils import ask_retain, ask_edit

# Let the user prune a candidate list
kept: list[str] = await ask_retain(["auth", "caching", "logging", "metrics"])

# Let the user edit a list of strings in-place
edited: list[str] = await ask_edit(["step 1", "step 2", "step 3"])

Dependencies

• fabricatio-core — core interfaces and capabilities
• questionary>=2.1.0 — interactive terminal prompts

License

MIT — see LICENSE