Method increment_token_count

config/global_state.py:162–205

Increments the token count for the current profile and checks if it exceeds the limit. [GR-01] Implements Graceful Rotation logic.

(cls, count: int, model_id: str = "default")

    @classmethod
    def increment_token_count(cls, count: int, model_id: str = "default"):
        """
        Increments the token count for the current profile and checks if it exceeds the limit.
        [GR-01] Implements Graceful Rotation logic.
        """
        if count <= 0:
            return

        # Ensure model_id is a valid string for key usage
        safe_model_id = model_id if model_id else "default"
        model_key = safe_model_id.lower()

        cls.current_profile_model_usage[model_key] += count
        current_usage = cls.current_profile_model_usage[model_key]

        # Retrieve limit (fallback to global hard limit)
        limit = MODEL_QUOTA_LIMITS.get(model_key, QUOTA_HARD_LIMIT)

        # Check Hard Limit (Emergency Kill / Model Exhaustion)
        if current_usage >= limit:
            logger.critical(
                f"⛔ HARD LIMIT REACHED ({model_key}): {current_usage} >= {limit}. Marking model as exhausted."
            )
            cls.current_profile_exhausted_models.add(model_key)
            # Trigger global rotation signal
            cls.set_quota_exceeded(message=f"Quota exceeded for model {model_key}")
            # Raise exception to propagate up to request processor
            raise QuotaExceededError(
                f"Quota exceeded for model {model_key} ({current_usage} >= {limit})"
            )

        # Check Soft Limit (Graceful Signal)
        # Note: Using global soft limit as baseline for rotation signal
        if current_usage >= QUOTA_SOFT_LIMIT and not cls.NEEDS_ROTATION:
            logger.warning(
                f"🔄 SOFT LIMIT REACHED ({model_key}): {current_usage} >= {QUOTA_SOFT_LIMIT}. Setting NEEDS_ROTATION flag."
            )
            cls.NEEDS_ROTATION = True

        # Log status
        limit_str = f"{QUOTA_SOFT_LIMIT}(Soft)/{limit}(Hard)"
        logger.info(
            f"📊 Token usage updated ({model_key}): +{count} => {current_usage} (Limits: {limit_str}) | Rotation Pending: {cls.NEEDS_ROTATION}"
        )