"""LLM Response - Unified message classes for LLM communication

This module provides unified data classes for message passing throughout the LLM pipeline.
"""
from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field


@dataclass
class ParsedDelta:
    """Streaming response delta
    
    Represents a single unit of streaming response data.
    Used for streaming responses where content is accumulated incrementally.
    
    Attributes:
        thinking: Accumulated thinking/reasoning content
        text: Accumulated text content  
        tool_calls: List of tool call requests
        is_complete: Whether this is the final delta
        usage: Token usage statistics
    """
    thinking: str = ""
    text: str = ""
    tool_calls: List[Dict] = field(default_factory=list)
    is_complete: bool = False
    usage: Dict[str, int] = field(default_factory=dict)
    
    def has_thinking(self) -> bool:
        """Check if there's thinking content"""
        return bool(self.thinking)
    
    def has_text(self) -> bool:
        """Check if there's text content"""
        return bool(self.text)
    
    def has_tool_calls(self) -> bool:
        """Check if there are tool calls"""
        return bool(self.tool_calls)
    
    def has_content(self) -> bool:
        """Check if there's any content"""
        return self.has_thinking() or self.has_text() or self.has_tool_calls()


@dataclass
class LLMResponse:
    """Complete LLM response
    
    Represents a complete non-streaming response.
    
    Attributes:
        content: Final text content
        thinking: Final thinking content (if any)
        tool_calls: List of tool calls (if any)
        usage: Token usage statistics
    """
    content: str = ""
    thinking: str = ""
    tool_calls: List[Dict] = field(default_factory=list)
    usage: Dict[str, int] = field(default=dict)
    
    def has_tool_calls(self) -> bool:
        """Check if there are tool calls"""
        return bool(self.tool_calls)