29c64b340d
Chat-side integration of the LEZ-backed RLN mix protocol:
- src/chat/delivery/waku_client.nim: mount waku_mix with onchain
RLN spam protection wired to logos_core_client fetchers; gate
the first publish on (a) gifter status confirmation, (b)
cushion of 2 poll intervals after confirmation, and (c) proof
root stability in the local valid_roots window; wrap mix
lightpush in withTimeout so vanished SURB replies surface as
Err instead of pinning the send coroutine.
- src/chat/client.nim: surface sendBytes errors via asyncSpawn
wrapped try/except instead of discarding the future (was
hiding every mix-publish failure).
- chat-side gifter client invocation (RLN membership service
wire format, EIP-191 ethereum-allowlist auth).
- Background membership status watcher that reconciles the
optimistic leaf returned by the gifter against the chain's
authoritative leaf via the status RPC.
Simulation harness (simulations/mix_lez_chat/):
- Spin up sequencer + run_setup + 4 mix nodes (one of which
runs the gifter service) + chat sender + chat receiver.
- SIM_NETWORK={local,testnet}, SIM_SLIM for testnet (reuses
shipped config_account + cached payment_account), Docker
image + GHCR for cross-platform testing.
- Strict mix-pool readiness gate, kademlia + RLN root activity
checks, gifter EIP-191 auth fixture, slim-mode submodule
minimization.
- TREE_ID_HEX pinned to the canonical testnet deployment.
Submodule bumps:
- vendor/nwaku to 8e6ba04 (LEZ-backed RLN mix + 2-phase gifter).
- vendor/logos-lez-rln to 950f287 (SPEL RLN program + mix sim
infrastructure + canonical testnet deploy).
Docs:
- RUN_SLIM_TESTNET.md: slim sim recipe.
- cleanup/MODE_A_GIFTER_SLOT_BUG.md: per-signer nonce collision
postmortem driving the queue+worker fix.
3.2 KiB
Running the Mix + LEZ RLN Chat Simulation
End-to-end private chat between two logos-chat-module clients over a 4-node mix network with LEZ-backed RLN spam protection.
Two logoscore instances (sender + receiver) establish an X3DH key agreement via an out-of-band intro bundle, then exchange double-ratchet-encrypted messages routed through 3-hop Sphinx onion routes with per-hop RLN proof generation and verification. Node 0 mounts the rln_gifter service; nodes 1-3 and both chat clients register RLN memberships on-chain via the gifter protocol. The sender publishes via lightpushPublish(mixify=true), the mix exit node verifies the RLN proof before fanning out via gossipsub relay, and the receiver consumes the message via a Waku filter subscription.
macOS
Prereqs: nix (with flakes), Docker, cargo-risczero, SSH access to GitHub.
git clone -b feat/logos-delivery git@github.com:adklempner/logos-chat.git
cd logos-chat && bash simulations/mix_lez_chat/setup_and_run.sh
First run: ~15-25 min. Re-runs: bash simulations/mix_lez_chat/run_simulation.sh --fresh (~5 min).
Linux (native)
Prereqs: nix (with flakes), Docker, cargo-risczero, SSH access to GitHub.
git clone -b feat/logos-delivery git@github.com:adklempner/logos-chat.git
cd logos-chat && bash simulations/mix_lez_chat/setup_and_run.sh
Same as macOS. On x86_64 Linux this should work out of the box. On aarch64 Linux, guest zkVM binaries must be pre-built on another platform (rzup doesn't support aarch64-linux) and the wallet module nix build needs RISC0_SKIP_BUILD_KERNELS=1.
Linux (via Docker)
Prereqs: Docker.
git clone -b feat/logos-delivery git@github.com:adklempner/logos-chat.git
cd logos-chat && bash scripts/run_in_docker.sh
The pre-built image (ghcr.io/adklempner/logos-chat-sim) is pulled automatically (~8.5GB download). Guest zkVM binaries must exist on the host from a previous macOS/x86_64 build, or set GUEST_BINARIES_DIR.
Each sim run: ~10 min (clone + sequencer build + sim). To force a local image rebuild: REBUILD_IMAGE=1 bash scripts/run_in_docker.sh.
Pass criteria
ALL 15 CHECKS PASSED — 4 mix nodes mounted, gifter service, LEZ RLN active, sender+receiver initialized/started/mix-mounted, intro bundle created, messages sent and received.
LEZ backend
The simulation runs against the SPEL framework (logos-lez-rln sim-on-spel branch, based on feat/spel). The on-chain RLN programs use SPEL's #[lez_program] macro with 32-byte tree IDs, Borsh-encoded state, and PDA-based account derivation via combine_seeds.
Configuration
SIM_LOG_LEVEL=TRACE SIM_KADEMLIA_MIN_WAIT=10 bash simulations/mix_lez_chat/run_simulation.sh --fresh
Full list of SIM_* variables in simulations/mix_lez_chat/README.md.
If it fails
Re-run with fresh state:
bash simulations/mix_lez_chat/run_simulation.sh --fresh
Wallet/sequencer errors:
rm -f vendor/logos-lez-rln/dev/wallet_config.json vendor/logos-lez-rln/dev/storage.json
rm -f ~/.logos-lez-rln/payment_account_*.txt
Guest binary errors after updating submodules:
rm -rf vendor/logos-lez-rln/lez-rln/methods/guest/target
bash simulations/mix_lez_chat/setup_and_run.sh
Docker logs are rescued to ./docker-sim-logs/ on failure.