The production-grade starter for GenAI products

Multi-tenant SaaS foundations with the LLM layer done right: cost tracking, resilience, observability, infra as code — open source, MIT licensed.

Multi-tenancy built in
Organizations, roles and invitations — every query tenant-scoped at the repository layer.
LLM costs under control
Tokens, cost and latency tracked per tenant and per feature. Rate limiting included.
Production-grade plumbing
Retries, circuit breakers, streaming, OpenTelemetry, Terraform on GCP — the parts demos skip.
Evals from day one
A RAG evaluation harness with LLM-as-judge scoring, wired into the workflow.