Cost Controls
Track AI spend by model, user, department, and routing rule. See savings from caching, routing, and compression in real time.
Cost Overview
The Cost Controls tab in the dashboard provides a comprehensive view of your AI spending:
- Total spend for the selected time period
- Period-over-period comparison showing spend trends
- Savings breakdown from caching, routing, and compression
- Per-model breakdown showing which models cost the most
- Per-rule breakdown showing which routing rules drive spend
- Per-user breakdown showing individual user costs
Savings Breakdown
Xilos saves you money through three mechanisms:
| Mechanism | How It Saves | Typical Savings |
|---|---|---|
| Semantic Caching | Reuses responses for similar queries — no LLM call | 30-70% of repeat queries |
| Smart Routing | Routes to cheaper models when possible | 20-40% vs. always using premium models |
| Context Compression | Reduces token count before LLM call | 50-90% token reduction |
The savings breakdown shows the dollar value of each mechanism for the selected period.
Per-Model Breakdown
See which models are consuming the most budget:
- Total cost per model
- Token count (input + output)
- Number of queries routed to each model
- Average cost per query
- Cost trend over time
Use this to identify opportunities to switch to cheaper models for certain query types.
Per-Rule Breakdown
See which routing rules are driving spend:
- Total cost per rule
- Number of queries matching each rule
- Average cost per query for each rule
- Cache hit rate per rule
- Compression savings per rule
Use this to identify rules where compression or caching could be enabled to reduce costs.
Per-User Breakdown
Track individual user spending:
- Total cost per user
- Query count per user
- Average cost per query
- Top users by spend
Info: Use Virtual Keys to set per-user or per-team budget limits. When a budget is exceeded, the key is automatically deactivated.
Forecasting
The Cost Controls page includes a simple forecast based on current spending trends:
- Projected monthly spend
- Projected savings from caching and compression
- Budget alerts (when configured)
Compression Savings
Compression savings are tracked separately and show:
- Total tokens saved
- Dollar value of saved tokens
- Average compression ratio
- Compression rate (percentage of queries that were compressed)
Cost Optimization Tips
- Enable caching on high-volume routing rules with stable answers
- Enable compression for rules with long conversation histories
- Use Smart Routing to automatically select cost-efficient models
- Review the Model Risk Leaderboard to find cheaper models with acceptable quality
- Set Virtual Key budgets to prevent overspending by individual users or teams
- Monitor the Suggestions page for AI-powered cost optimization recommendations