Cost Tracking

HyperSaaS tracks every LLM API call and calculates costs based on per-model token pricing.

Models

AIUsage

Logs every individual LLM interaction:

class AIUsage(BaseModel):
    user = models.ForeignKey(User, null=True, blank=True)
    workspace = models.ForeignKey(Workspace, on_delete=models.CASCADE)
    chat_session = models.ForeignKey(ChatSession, null=True, blank=True)
    model_used = models.CharField(max_length=100)
    input_tokens = models.PositiveIntegerField()
    output_tokens = models.PositiveIntegerField()
    input_cost_per_token = models.DecimalField(max_digits=15, decimal_places=10)
    output_cost_per_token = models.DecimalField(max_digits=15, decimal_places=10)
    total_cost = models.DecimalField(max_digits=15, decimal_places=8)  # Auto-calculated

UserCost

Aggregated cost per user across all workspaces:

class UserCost(BaseModel):
    user = models.OneToOneField(User, on_delete=models.CASCADE)
    total_cost = models.DecimalField(max_digits=14, decimal_places=6)
    overage_cost = models.DecimalField(max_digits=12, decimal_places=2)
    last_reset_at = models.DateTimeField(null=True)

Pricing Configuration

Token pricing is defined per model in chat/pricing_config.py:

PRICING_CONFIG = {
    "gpt-4o": {
        "input_cost_per_1k_tokens": Decimal("0.0025"),
        "output_cost_per_1k_tokens": Decimal("0.01"),
    },
    "claude-sonnet-4-5-20250514": {
        "input_cost_per_1k_tokens": Decimal("0.003"),
        "output_cost_per_1k_tokens": Decimal("0.015"),
    },
    # ... all supported models
}

Credit Limits

Subscription plans map to credit limits via PLAN_CREDIT_MAPPING:

PLAN_CREDIT_MAPPING = {
    "price_xxx_monthly": Decimal("9.99"),
    "price_xxx_yearly": Decimal("99.90"),
    # ...
}

Credit Usage API

GET /api/users/credit-usage/

Returns:

{
  "total_credit_limit": "39.99",
  "total_cost": "12.45",
  "percentage_used": 31.1
}

Before creating a chat session, the system checks has_sufficient_buffer(user, workspace) to ensure the user hasn't exceeded their credit limit.