FinOps Dashboard
Real costs for the Raku platform. All numbers USD/month.
Cost Breakdown
OpenAI Embeddings
Estimatetext-embedding-3-small @ $0.02/1M tokens. ~5 tenants × 100 memories × 500 tokens = 250K tokens/mo
Embeddings are the #1 variable cost per Ware. text-embedding-3-small at $0.02/1M tokens. At 1000 memories per tenant per month = $0.01/tenant.
Fly.io Compute
ActualHobby plan: shared-cpu-1x, 256MB RAM, 1GB encrypted volume. Free allowance covers current usage
Fixed cost shared across all tenants. At 100 tenants = $0/tenant (hobby). At scale: $5.70/mo (dedicated-cpu-1x) / 100 tenants = $0.06/tenant.
GitHub Actions CI/CD
EstimateFree tier: 2,000 min/mo (Pro: 3,000). Each CI run ~3-5 min. ~30 pushes/mo = ~120 min
CI cost scales with push frequency. Each CI run ~4 minutes × 30 pushes/mo = 120 min/mo. Well within free tier (2,000 min). Overage: $0.008/min.
Cloudflare
ActualFree tier: Pages, DNS, SSL, email routing, R2 (10GB free). Zero cost at current scale
CDN, DNS, SSL, email routing all free. Scales to millions of requests before paid tier kicks in.
Portkey AI Gateway
Coming SoonComing June. Free tier: 10K requests/mo. Will track per-tenant tokens and model costs
LLM calls are the #2 variable cost. Cortex routing to cheapest capable model is the margin strategy.
Unit Economics Calculator
Cost Trend (Projected)
Projected costs based on tenant growth from 5 (Apr) to 100 (Sep). Assumes conservative LLM usage pattern.
Cortex Tier Pricing Model
Cortex routes each request to the cheapest capable model. Tier 0-1 are free (cached/rule-based). This is the margin strategy.
Portkey AI Gateway integration coming June. Will enable real per-tenant token tracking and automatic tier selection.