Project Budgets

Control LLM spending with monthly budgets and real-time alerts

Overview

Dakora’s budget feature helps you maintain control over LLM costs. Set spending limits at the project level, receive proactive alerts, and enforce budget caps to prevent overspending.

Key Capabilities

Monthly Budgets

Set hard or soft spending limits per project

Real-time Alerts

Get notified before reaching budget thresholds

Flexible Enforcement

Choose between alert-only or hard budget blocks

Multi-Environment

Separate budgets for dev, staging, and production

Setting Up Budgets

Configure in Studio

1. Navigate to Settings > Budget
2. Set your monthly budget limit in USD
3. Configure alert thresholds (e.g., notify at 80%)
4. Choose enforcement mode: alert-only or hard limit

Budget Configuration Options

Setting	Description
Monthly Budget	Maximum USD spend per month
Alert Threshold	Percentage at which to send alerts (e.g., 80%)
Enforcement Mode	alert (notify only) or block (stop executions)

Budget Alerts & Enforcement

Alert System

When spending approaches your configured threshold:

1. Email notification sent to project owners
2. Dashboard warning displayed in Studio
3. API response includes budget status in metadata

Enforcement Modes

Alert-only mode: Continue executions but send notifications

Block mode: Reject new execution requests once budget is exceeded

{
  "error": "Budget exceeded",
  "details": {
    "budget_limit": 100.00,
    "current_spend": 102.45,
    "enforcement_mode": "block"
  }
}

Best Practices

Start with Alert Mode

Begin with alert-only enforcement to understand your spending patterns before enabling hard limits that could disrupt service.

Set Multiple Thresholds

Configure alerts at 50% and 80% to give yourself adequate time to react before hitting the budget limit.

Use Environment-Specific Projects

Create separate projects for development, staging, and production with appropriate budget limits for each environment.

Next Steps

Observability

Track execution logs and traces

Templates

Optimize your prompts for efficiency