Living Systems — What's Running Now
Real-time status of the ecoPrimals sovereign mesh: active gates, deployed primals, capability routing, and live JupyterHub compute.
The Mesh Is Alive
The ecoPrimals ecosystem is not a description of future work. It is running. Seven gates are enrolled, three are meshed with bilateral direct TCP, and Songbird routes capability.call across all of them. This page shows what is actually deployed and operational.
Active Gates
| Gate | Status | Transport | What’s Running |
|---|---|---|---|
| sporeGate | 🟢 Online | LAN + WG | Sovereign CI, Caddy TLS, songBird mesh hub, Forgejo |
| eastGate | 🟢 Online | LAN + WG (10GbE) | Overwatch, primalSpring ( 886 tests), petalTongue |
| ironGate | 🟢 Meshed | LAN + WG | RTX 5070 Ti, JupyterHub (deploying), songBird |
| flockGate | 🟡 WAN | WG via golgi | Tower atomic evolution: songBird, bearDog, skunkBat dev |
| golgi | 🟢 Online | VPS | WireGuard hub, Forgejo host, depot, cascade timer |
| grapheneGate | 🟢 Online | ADB (USB) | Pixel 8a, Tower running, 14/14 aarch64-musl binaries |
| strandGate | ⬜ Joining | — | 64-core EPYC, 256GB — awaiting hardware enrollment |
Live Capabilities
When a gate starts Songbird, it announces its capabilities to the mesh. Other gates can then invoke any capability by name — songBird routes to the best available provider.
Currently Routed
| Capability | Provider | Path | Use |
|---|---|---|---|
http.proxy | sporeGate | LAN direct | HTTP routing to mesh services |
peer.connect | all meshed gates | bilateral TCP | Mesh peering, 0ms LAN |
capability.call | sporeGate → ironGate | LAN direct | Cross-gate compute dispatch |
build.release | sporeGate | local | Sovereign CI binary builds |
cascade.sync | golgi | WG | 15-min quorum cascade timer |
Deploying
| Capability | Provider | Status |
|---|---|---|
jupyter.execute | ironGate | JupyterHub deployment in progress |
compute.gpu | ironGate | RTX 5070 Ti ready, pending capability registration |
compute.cpu | strandGate | Awaiting hardware enrollment |
JupyterHub — Live Compute
JupyterHub on ironGate provides notebook access to the live mesh. Once deployed, the path is:
Browser → lab.primals.eco
→ bearDog :443 (ACME TLS)
→ songBird capability.call("jupyter")
→ ironGate :8000 (LAN direct, <1ms)
→ JupyterHub sessionWhat makes this different from a cloud notebook: your computation runs on sovereign hardware in a private lab. No telemetry. No vendor. The mesh handles routing — if ironGate goes offline, songBird can route to strandGate (once enrolled) or any future compute node. The notebook doesn’t know which gate ran it.
Example Workloads Available
| Workload | Hardware | Spring |
|---|---|---|
| 16S metagenomics pipeline | CPU | wetSpring |
| GROMACS metadynamics (CAZyme FEL) | RTX 5070 Ti GPU | hotSpring |
| Salmon RNA-seq quantification | CPU + NVMe | wetSpring |
| STAR alignment (large genomes) | 64-core EPYC (strandGate) | wetSpring |
| ET₀ irrigation modeling | CPU | airSpring |
| PK/PD compartmental modeling | CPU | healthSpring |
Each workload runs against the same infrastructure that produced the baseCamp results. Every run gets a provenance chain.
Sovereign CI Pipeline
All ? primals are continuously built from source on sporeGate’s Sovereign CI. The pipeline:
Developer pushes to Forgejo (git.primals.eco)
→ golgi cascade timer (15-min quorum)
→ sporeGate pulls, builds x86_64-musl + aarch64-musl
→ BLAKE3 checksums computed
→ Binaries published to depot (membrane.primals.eco/depot/)
→ Gates cascade + pull from depotWave 133a result: 30/30 ecobins built (15 x86_64 + 15 aarch64), 283 MB total, all checksummed and published. 11/14 primals build with zero manual intervention; 3 require documented workarounds (being converged).
Mesh Health
The mesh is self-healing. If a gate goes offline:
- Songbird detects peer loss via heartbeat timeout
- Capability routing tables update across all remaining peers
- Services that depended on the lost gate get routed to alternates
- When the gate returns,
peer.connectre-establishes bilateral trust
Key invariant: unplugging any single gate does not kill the network. The Flint edge router is the plasma membrane. Gates are ephemeral compute.
What’s Next
| Item | Gate | When |
|---|---|---|
| JupyterHub deploy + capability registration | ironGate | This wave |
| bearDog ACME gateway activation (Caddy retirement) | sporeGate | After flockGate fix |
| strandGate enrollment (mesh auto-absorb) | strandGate | When hardware arrives |
| Bidirectional relay (GitHub ↔ Forgejo parity) | golgi | cellMembrane deploy |
| Live mesh visualization (petalTongue wire-up) | eastGate | sporePrint + petalTongue |
Related
- Compute Access — tiers, hardware, how to connect
- Gate Mesh — Live Topology — mesh architecture and enrollment
- Reproduce Results — run the same pipelines on your hardware