futurewithvibe Model Usage Tracking
Track LLM usage, token consumption, and estimate costs.
Monitoring
Token Usage
Track tokens used in conversations:
- Input tokens
- Output tokens
- Total tokens
Cost Estimation
Anthropic Pricing (as of 2026)
- Claude Opus 4.5: $15 per MTok input, $75 per MTok output
- Claude Sonnet 3.5: $3 per MTok input, $15 per MTok output
OpenAI Pricing
- GPT-4 Turbo: $10 per MTok input, $30 per MTok output
- GPT-4: $30 per MTok input, $60 per MTok output
Usage Examples
User: "How many tokens have I used today?"
- Query session history
- Sum token counts
- Present totals
User: "Estimate my costs for this month"
- Gather usage statistics
- Calculate costs per model
- Present breakdown
User: "Which model is most efficient for my use case?"
- Compare token usage patterns
- Calculate cost per conversation
- Recommend most efficient model
Best Practices
- Log token usage with each API call
- Store usage data in session metadata
- Provide cost estimates before expensive operations
- Suggest model switching for cost optimization
Data Structure
usage = {
"model": "claude-opus-4-5",
"input_tokens": 1500,
"output_tokens": 500,
"cost_usd": 0.06, # Estimated
"timestamp": "2026-01-27T00:00:00Z"
}
Reporting
Generate usage reports:
- Daily/weekly/monthly summaries
- Per-model breakdown
- Cost trends
- Token efficiency metrics