Plastic Labs engineering is split across three layers:
- Core product — Honcho, the open-source memory primitive
- Managed SaaS — Groudon (multi-tenant orchestration) plus our reference conversational app Kyogre
- ML pipeline — Excadrill → Minccino → Machamp → Metagross
Start here for new engineers
- System Map — how the pieces fit together
- Onboarding — week-1 checklist (company-wide + the Engineering track)
- Pokédex — every codename, what it points to, where its repo lives
- Honcho ADR — the architectural decisions behind today’s Honcho (especially the Peer Paradigm)
Before you change anything that runs in production, read
the GitOps pipeline. Everything on GKE is
deployed by ArgoCD from git; nobody runs kubectl apply.
Reference
- Honcho Specs — active and shipped feature specifications
- Engineering System — tools, testing, debugging
- Infra — hardware, cloud, the 10.x IP plan, networking, telemetry
- Adding an internal service — reserve an internal IP and a
*.prod.internalname - Groudon Architecture — services, request routing, instance states
- Compute classes and spot — which machines the fleet runs on, why derivers are on spot
- Runbooks — all of them; most used: where-to-look, tenant-debugging, argocd-operations
- SDK Release Workflow · Docs Proxy