Integrations

AGPL analytics peer: dashboards over the Signals warehouse and SDG corpus.

A dashboard-capability engine on the lattice (gRPC :50451). This AGPL fork stays in its own tree — never vendored into Signals — and registers zndx.engine.v1 so launchers discover its UI the same way they discover Gaius or Hermes.

Platform production flows: one metadata service, YuniKorn pods, Airflow DAGs.

Signals-owned Metaflow so every federated engine shares run metadata, a RustFS artifact store, and Kubernetes tasks YuniKorn admits. Airflow is the production clock; Knative Eventing is the trigger plane.

OpenLineage UI on Atlas: the REST contract, not a second catalog.

Marquez-web is the human OpenLineage surface for Signals. It proxies /api/v1 and /api/v2beta to the Atlas fork; lineage events, jobs, datasets, and graphs live in AGE — the same store Ranger and sigint already use.

Nous Research agent runtime: AgentRTC sessions and weathership plugins.

Hermes Agent joins the lattice as capability=agent. Interactive WebRTC (Kyutai STT/TTS) stays on the engine; the GPU claim is a Signals coordination Activity on root.internal.inference.agent-rtc. Plugins under development attach weathership memory and context to that session.

Board-shaped collaboration via the official Miro AI MCP — direction, not a pin yet.

Miro's official AI tools expose boards over MCP (mcp.miro.com, OAuth). The weathership approach is to consume that MCP from Hermes and other engines — browse, diagram, and spec-extract onto boards — without vendoring miro-ai or standing up a second identity plane.