FEAT: Add HumanizerTool for Production-Ready Output Control and Style Transfer #3990
Description
Feature Area
Core functionality
Is your feature request related to a an existing bug? Please link it here.
Goal: To introduce a new HumanizerTool that allows agents to perform advanced style and tone transfer on their final output, solving the "last-mile" problem of robotic AI text.
Describe the solution you'd like
The tool will encapsulate a specialized multi-step agentic process (derived from commercial experience) to ensure the output is not just grammatically correct, but matches a specified human persona and style.
Proposed Tool Interface:
Python
class HumanizerTool(BaseTool):
name: str = "Humanizer Tool"
description: str = "Rewrites raw AI text to match a specified human style, tone, and mode for production-ready communication."
def _run(self, text: str, style: str, mode: str) -> str:
# Core logic for style transfer and humanization
...
Key Parameters (based on #https://writenaturallyai.com/ ):
text: The raw, AI-generated text output from a previous agent step.
style: The desired human style (e.g., Casual, Professional, Neutral, Academic).
mode: The intensity of the rewrite (e.g., Standard, EXTREME).
This allows a final agent in a CrewAI workflow to execute: final_report = HumanizerTool.run(raw_text, style='Professional', mode='EXTREME').
Describe alternatives you've considered
The primary alternative is to rely on in-line prompting within the agent's final task description (e.g., "Rewrite this in a professional tone").
Why the Tool is Superior:
Consistency: A dedicated tool ensures a standardized, multi-step, and tested process for humanization, guaranteeing consistent quality across all agents and crews. In-line prompting is highly susceptible to LLM drift and inconsistent results.
Complexity Abstraction: The tool abstracts away the complex, multi-step prompting required for true style transfer (which often involves multiple LLM calls and self-correction loops), keeping the agent's main task clean and focused.
Production-Proven: This feature is based on the core logic of a successful commercial product, Writenaturallyai.com, which has been tested against real-world AI detectors and achieved a 98% user satisfaction rate on output quality.
Additional context
This tool directly addresses a critical need for commercial and production-grade agent deployments. Agents are increasingly used for client communication, marketing copy, and academic writing, where the "AI sound" is a major liability.
My experience as the founder of #https://writenaturallyai.com/ gives me a unique understanding of the linguistic patterns that AI detectors flag (burstiness, perplexity, unnatural sentence structure). I can contribute a tool that is specifically engineered to counteract these patterns.
Willingness to Contribute
Yes, I'd be happy to submit a pull request