Note

💡 This document was written with assistance from GPT-4o.

Introduction

A naive polygon segment selector using OpenAI Agents SDK function calling for prompt-based segment selection. The repository explores possibilities for an AI-assisted geometric task.

https://github.com/PARKCHEOLHEE-lab/segment-selector

Key Components

1. SegmentSelectorOutput

Represents the structured output of the segment selection using Pydantic:

class SegmentSelectorOutput(BaseModel):
    """`final_output` of the segment selector agent"""
		
		# output
    selected_indices: list[int]
		
		# parameters for debugging
    directions: list[str]
    selection_count: Union[float, int]
    similarity_threshold: float

• JSON output schema, and JSON-related instruction are not needed.
• selected_indices: List of selected segment indices.
• directions: List of direction keywords.
• selection_count: Total count (or ratio) of segments to select.
• similarity_threshold: Cosine similarity threshold for direction alignment.