Outcome
Ship an internal operator-facing analytics app that shows daily token usage, cost, latency, model mix, tool mix, and anomaly issues on Cloudflare Workers using D1-backed rollups.
Why this matters
Teams need daily visibility into AI usage without digging through logs, provider exports, or one-off spreadsheets.
Scope
- TanStack Start app deployed on Cloudflare Workers
- D1/Drizzle-backed rollup model
- Dense dashboard for overview, models, and tools
- Ingestion path for source usage data
- Query path optimized for daily aggregates
- Issue/anomaly surfacing for cost and reliability signals
Execution order
- Foundation and deployment wiring
- Ingestion contract
- Daily rollup materialization
- Live dashboard queries
- Drilldowns and persistent filters
- Issue detection and guardrails
- Auth and workspace scoping, if needed
- Hardening and polish
Definition of done
- Dashboard runs on Workers
- D1 schema and migrations are in repo
- Live data replaces mock overview data
- Team can trust day-level numbers for cost and usage decisions
Outcome
Ship an internal operator-facing analytics app that shows daily token usage, cost, latency, model mix, tool mix, and anomaly issues on Cloudflare Workers using D1-backed rollups.
Why this matters
Teams need daily visibility into AI usage without digging through logs, provider exports, or one-off spreadsheets.
Scope
Execution order
Definition of done