feature: revise node states to be simpler (WIP)

README.md

@@ -303,7 +303,7 @@ sudo loginctl enable-linger "$USER"
 mesh-llm serve --model Qwen2.5-32B    # dashboard at http://localhost:3131
 ```
 
-Live topology, per-node GPU capacity, model picker, and built-in chat. Everything comes from `/api/status` (JSON) and `/api/events` (SSE).
+Live topology, per-node GPU capacity, model picker, and built-in chat. Live members show only the `Client`, `Standby`, `Loading`, and `Serving` badges. Wakeable provider-backed capacity is shown separately from topology and stays out of routing until it rejoins. Everything comes from `/api/status` (JSON) and `/api/events` (SSE).
 
 ## Multimodal Support
 
@@ -500,7 +500,7 @@ When a node is running, open:
 http://localhost:3131
 ```
 
-The console shows live topology, VRAM usage, loaded models, and built-in chat. It is backed by `/api/status` and `/api/events`.
+The console shows live topology with only `Client`, `Standby`, `Loading`, and `Serving` badges for live members, plus separate wakeable capacity, VRAM usage, loaded models, and built-in chat. Wakeable inventory is not part of topology peers or routing until it rejoins. It is backed by `/api/status` and `/api/events`.
 
 You can also try the hosted demo:
 

mesh-llm/docs/DESIGN.md

@@ -39,7 +39,7 @@ src/
 └── system/                  Hardware detection, benchmarking, self-update
 ```
 
-## Node Roles
+## Topology Roles
 
 ```rust
 enum NodeRole {
@@ -49,7 +49,7 @@ enum NodeRole {
 }
 ```
 
-Roles are exchanged via gossip. Preferred peers use `meshllm.node.v1` protobuf on QUIC ALPN `mesh-llm/1`; legacy peers may still negotiate `mesh-llm/0` and use the older JSON gossip payloads. A node transitions Worker → Host when elected.
+Roles are exchanged via gossip. Live-state badges are separate and use `Client`, `Standby`, `Loading`, and `Serving`. Preferred peers use `meshllm.node.v1` protobuf on QUIC ALPN `mesh-llm/1`; legacy peers may still negotiate `mesh-llm/0` and use the older JSON gossip payloads. A node transitions Worker → Host when elected.
 
 A newly connected peer is quarantined until it sends a valid `GossipFrame` with `gen = 1` (quarantine-until-gossip admission model). Only streams 0x01 (GOSSIP) and 0x05 (ROUTE_REQUEST) are accepted before admission. All other streams are rejected until the peer is admitted.
 

mesh-llm/docs/TESTING.md

@@ -162,7 +162,7 @@ mesh-llm serve --join <TOKEN>
 ```
 
 - Joiner scans the Hugging Face cache and picks an unserved model already on disk
-- Log: "Assigned to serve GLM-4.7-Flash (needed by mesh, already on disk)"
+- Log: "Selected to serve GLM-4.7-Flash (needed by mesh, already on disk)"
 
 ### 8. Lite client with multi-model
 
@@ -217,6 +217,20 @@ curl -X DELETE localhost:3131/api/runtime/models/Llama-3.2-1B-Instruct-Q4_K_M
 - Switching models highlights the serving node in topology view
 - Chat routes to selected model via API proxy
 
+### 11. Console live-state and wakeable capacity
+
+```bash
+cd mesh-llm/ui/
+npm run test:run
+npm run typecheck
+just build
+```
+
+- Live badges show only `Client`, `Standby`, `Loading`, and `Serving`
+- Wakeable capacity renders in a separate section from topology peers and live nodes
+- Wakeable entries do not appear in the topology peer list
+- Validation uses `npm run test:run`, `npm run typecheck`, and `just build`
+
 ## Mesh Identity
 
 ### 16. Mesh ID generation (originator)