task-refiner-llm 2025.9.9

Refine unstructured briefs into implementation-ready JSON via LLM7 and llmatch.

refine_task_with_llm

refine_task_with_llm is a Python package that provides a function to convert loose user briefs into structured, implementation-ready task descriptions formatted in strict JSON. It leverages language models with robust extraction techniques to ensure precise, usable output for software engineering tasks.

Installation

Install via pip:

pip install refine_task_with_llm  

Usage

The main function, refine_task_with_llm, accepts an LLM instance and a raw user brief to return a structured JSON object describing the refined task. If no LLM is provided, it initializes a deterministic ChatLLM7. The function employs llmatch for reliable extraction of JSON from LLM output, raising errors if extraction or parsing fails.

Example

from langchain_core.language_models import BaseChatModel  
from refine_task_with_llm import refine_task_with_llm  
  
# Optional: create your own LLM or pass None to use default  
result = refine_task_with_llm(  
    llm=None,  
    custom_text="Create a script to analyze sales data and generate a report.",  
    project_name="SalesAnalysis",  
    audience="junior developer",  
    include_examples=True  
)  
  
print(result)  

This call refines an unstructured brief into a detailed, JSON-formatted task description suitable for implementation.

Author

Author: Eugene Evstafev [e-mail]
Repository: https://github.com/chigwell/refine_task_with_llm