ci: prove the relay-driven mesh lifecycle — discover, join, infer, deny — with real nodes

[michaelneale]

Summary

CI now proves the full Buzz shared-compute join story end to end: a member can discover another member's served model through the Buzz relay alone and run inference over the mesh, while a non-member gets nothing — no relay visibility and no mesh admission, even holding a leaked endpoint address.

This is deliberately different from mesh-llm's own CI smokes (which bootstrap two nodes with a hand-carried invite token / mdns): here the relay is the control plane, exactly like the desktop app:

1. Membership — identities A and B are added via buzz-admin (kind:13534 NIP-43 roster); C is not.
2. Advertise — each member publishes a client-signed kind:30003 discovery note carrying its MeshLLM owner binding and (for the serve node) serveTargets[].endpointAddr, covered by an endpoint-binding signature — the exact payload shape the desktop coordinator publishes.
3. Trust — the serve node derives its admission allowlist from the relay (statuses ∩ roster, owner_ids_from_events semantics) and starts with TrustPolicy::Allowlist.
4. Join — the client verifies owner + endpoint bindings and membership, then dials the relay-discovered endpoint (the desktop join-watcher's dial_endpoint_addr step). No out-of-band token.
5. Infer — a chat completion against the client's local OpenAI endpoint routes over QUIC to the serve node's model (CPU, SmolLM2-135M, ~105MB).
6. Deny — the stranger's REQ returns zero mesh statuses from the membership-gated relay, and dialing the leaked endpoint directly is refused mesh admission (verified: 503, no route).

What's in the PR

• crates/buzz-relay/examples/mesh_relay_lifecycle_smoke.rs — the harness. One process per node (mesh-llm keeps process-global state under ~/.mesh-llm), orchestrator + serve/client/stranger roles, byte-identical binding payloads to desktop/src-tauri/src/mesh_llm/identity.rs (called out with keep-in-sync comments).
• scripts/ci-mesh-lifecycle-smoke.sh — provisions a membership-gated relay (throwaway owner + signing identities via buzz-admin generate-key), runs the harness, cleans up. Fails fast if :3000 is already occupied (a stale open relay would mask gating).
• scripts/start-relay-for-tests.sh — gains opt-in NIP-43 membership env passthrough (BUZZ_REQUIRE_RELAY_MEMBERSHIP + RELAY_OWNER_PUBKEY + BUZZ_RELAY_PRIVATE_KEY). Default behavior unchanged.
• .github/workflows/mesh-lifecycle.yml — separate, path-filtered, non-required workflow (mesh paths + this harness + dispatch). Caches the mesh native runtime + HF model keyed on the mesh pin.

Relationship to mesh-llm's CI

Follows the shape mesh-llm's own CI proved stable (tiny CPU model, one runner, multiple real mesh-llm processes over real QUIC — cf. their ci-two-node-client-serving-smoke.sh), but swaps the token bootstrap for the relay-driven lifecycle, which is the part only Buzz can test.

Validation

Rebased onto main after #3467 (mesh v0.74) landed — the harness rides the repo's [dev-dependencies] pin and passed the full run against v0.74.0 locally (macOS, --profile ci):

PASS 1/5: relay-derived allowlist resolved
PASS 2/5: serve member ready with model: jc-builds/SmolLM2-135M-Instruct-Q4_K_M-GGUF:Q4_K_M
PASS 3/5: client member discovered + joined via relay
PASS 4/5: inference routed over the mesh: "PONG"
PASS 5/5: stranger denied by relay and by mesh admission
PASS: full relay-driven mesh lifecycle verified

cargo fmt --all --check and cargo clippy -p buzz-relay --all-targets -- -D warnings pass.

Notes / open questions

• Draft until the Mesh Lifecycle workflow has a green run on Actions (Linux runner; local validation was macOS. The native runtime install path is the same SDK-owned dynamic path the desktop uses, and mesh-llm's own CI runs the same model CPU-only on ubuntu-24.04).
• The harness follows the repo's mesh pin automatically, so it doubles as a canary for future mesh upgrades (it already caught the v0.73.1 → v0.74.0 bump during development).
• The stranger "deny" accepts either shape mesh-llm exhibits: no model visibility at all, or gossip visibility with inference refused (admission gates routing, not gossip).

[Chessing234]

cool ci. if the smoke is flaky on shared runners, maybe pin a longer join timeout in the workflow rather than retrying the whole job