models.py

"""Data models for the DocEdit Environment — structured document editing via XML-like operations."""

from typing import Optional
from openenv.core.env_server.types import Action, Observation
from pydantic import Field


class DocEditAction(Action):
    """Agent sends an edit operation on the document."""

    operation: str = Field(
        ...,
        description="Edit operation: 'replace', 'insert', or 'delete'",
    )
    target: str = Field(
        default="",
        description="Text to find in the document (for replace/delete operations)",
    )
    content: str = Field(
        default="",
        description="New content (replacement text for 'replace', new text for 'insert')",
    )
    position: int = Field(
        default=-1,
        description="Paragraph index for 'insert' operation (-1 = append at end)",
    )


class DocEditObservation(Observation):
    """Observation returned after each step — current document state + task context."""

    document: str = Field(default="", description="Current document content (XML-tagged paragraphs)")
    target_description: str = Field(default="", description="Natural language description of the editing goal")
    similarity: float = Field(default=0.0, description="Current similarity to target document (0.0–1.0)")
    task_name: str = Field(default="", description="Current task identifier")
    steps_remaining: int = Field(default=0, description="Steps left in this episode")