from enum import Enum, auto
from dataclasses import dataclass, field


class EmmaState(Enum):
    IDLE = auto()        # Venter på input
    OBSERVING = auto()   # Analyserer kontekst
    REASONING = auto()   # MCoT resonneringsloop
    ACTING = auto()      # Utfører verktøykall/svar
    REFLECTING = auto()  # Evaluerer eget output
    LEARNING = auto()    # Oppdaterer resonans-minne


class EmmaAction(Enum):
    CLARIFY = auto()       # Be om mer info
    RETRIEVE = auto()      # RAG-oppslag
    TOOLCALL = auto()      # API/shell/fil
    RESPOND = auto()       # Generer svar
    COMPRESS = auto()      # Forenkle resonneringskjede (MCoT)
    STOREPATTERN = auto()  # Lagre vellykket mønster i resonansminne


@dataclass
class EmmaContext:
    userinput: str
    history: list
    toolresults: list = field(default_factory=list)
    thoughtchain: list = field(default_factory=list)
    rewardacc: float = 0.0
    stepcount: int = 0
    complexityscore: float = 0.0  # Flynn-tracker


# Overgangstabell: (state, action) -> neste state
TRANSITIONS: dict = {
    (EmmaState.IDLE,       EmmaAction.RETRIEVE):      EmmaState.OBSERVING,
    (EmmaState.OBSERVING,  EmmaAction.CLARIFY):        EmmaState.IDLE,
    (EmmaState.OBSERVING,  EmmaAction.TOOLCALL):       EmmaState.ACTING,
    (EmmaState.OBSERVING,  EmmaAction.RESPOND):        EmmaState.REASONING,
    (EmmaState.REASONING,  EmmaAction.COMPRESS):       EmmaState.REASONING,
    (EmmaState.REASONING,  EmmaAction.RESPOND):        EmmaState.ACTING,
    (EmmaState.ACTING,     EmmaAction.STOREPATTERN):   EmmaState.REFLECTING,
    (EmmaState.REFLECTING, EmmaAction.RETRIEVE):       EmmaState.LEARNING,
    (EmmaState.LEARNING,   EmmaAction.STOREPATTERN):   EmmaState.IDLE,
}


def reward(state: EmmaState, action: EmmaAction, outcome: dict) -> float:
    """Reward-funksjon: w = sim*ln(1+n)*decay^d*R"""
    r = 0.0
    if outcome.get("task_completed"): r += 1.0
    if outcome.get("user_confirmed"): r += 0.5
    if outcome.get("steps_used", 99) < 5: r += 0.3
    if outcome.get("tool_error"): r -= 0.4
    if outcome.get("hallucination_flag"): r -= 0.8
    return r