mirror of
https://github.com/logos-messaging/logos-delivery.git
synced 2026-07-21 04:00:27 +00:00
4a3db364e7
ServiceDiscovery.new(client=true) installs no inbound handler, so switch.mount always failed. Skip mount for clientMode and keep the protocol for outbound mix lookups only. Add a unit test and document relative CHAT_UI paths in the mixnet chat-ui launcher.
140 lines
6.0 KiB
Markdown
140 lines
6.0 KiB
Markdown
# E2E: logos-chat-ui over the local mixnet sim
|
||
|
||
Run two `logos-chat-ui` clients that exchange end-to-end messages **through the
|
||
5-node mix sim** (Sphinx routing + RLN spam protection + exit-node delivery).
|
||
|
||
This is a full from-scratch runbook: clone → build → provision → run → message.
|
||
|
||
> Want the **live testnet-0.2 mix fleet** instead of a local sim? The published
|
||
> `chat_ui_mix` app is **self-contained** (no setup) — see its
|
||
> [README → Try it](https://github.com/logos-co/logos-chat-ui/tree/feat/logos-testnetv02-mix#try-it).
|
||
|
||
The two clients adopt the **two provisioned chat memberships** the sim generates
|
||
(peer-IDs `…1Qx…` and `…RABscN`), via `CHAT_NODEKEY`. RLN keystores are
|
||
peer-ID-derived, so fixing the nodekey makes the right `rln_keystore_<peerId>.json`
|
||
resolve.
|
||
|
||
---
|
||
|
||
## 0. Prerequisites
|
||
|
||
- **Nix** with flakes enabled (the chat client is built/run via `nix run`).
|
||
- **Rust toolchain** (`rustc` + `cargo`, via [rustup](https://rustup.rs)) and the
|
||
standard dev tools (C compiler, GNU Make, Bash, Git) — needed to build the mix
|
||
node (`wakunode2`).
|
||
- macOS or Linux.
|
||
|
||
## 1. Repos & branches
|
||
|
||
Only **logos-delivery** needs to be cloned. The chat client (and its whole
|
||
dependency chain) is pulled automatically by `nix run`.
|
||
|
||
| repo | branch | role | how it's consumed |
|
||
|------|--------|------|-------------------|
|
||
| `logos-messaging/logos-delivery` | `feat/logos-testnetv02-mix` | mix sim (5 nodes) + this harness | clone + `make wakunode2` |
|
||
| `logos-co/logos-chat-ui` | `feat/logos-testnetv02-mix` | the GUI client | `nix run github:…` (pulls the rest) |
|
||
| `logos-co/logos-chat-module` | `feat/logos-testnetv02-mix` | `liblogoschat` module | transitive flake input |
|
||
| `logos-messaging/logos-chat` | `feat/logos-testnetv02-mix` | Nim SDK (in-process nwaku) | transitive flake input (`submodules=1`) |
|
||
|
||
## 2. Build & provision the sim
|
||
|
||
```bash
|
||
git clone --recurse-submodules -b feat/logos-testnetv02-mix \
|
||
https://github.com/logos-messaging/logos-delivery
|
||
cd logos-delivery
|
||
|
||
make wakunode2 # builds Nim + deps + librln + the mix node (first run is long)
|
||
|
||
cd simulations/mixnet
|
||
./build_setup.sh # generates rln_tree.db + 7 keystores (mix nodes + 2 chat slots)
|
||
```
|
||
|
||
> `build_setup.sh` sources the nwaku `env.sh`, so `nim` and `librln_v2.0.2.a`
|
||
> (built by `make wakunode2`) are already in place. The sim is configured with
|
||
> **cover traffic disabled** and **userMessageLimit = 100** (matches the chat's RLN
|
||
> default) in `config*.toml`.
|
||
|
||
## 3. Start the 5 mix nodes
|
||
|
||
Each in its own terminal, **bootstrap (`run_mix_node.sh`) first**:
|
||
|
||
```bash
|
||
cd simulations/mixnet
|
||
./run_mix_node.sh # bootstrap tcp/60001
|
||
./run_mix_node1.sh # tcp/60002
|
||
./run_mix_node2.sh # tcp/60003
|
||
./run_mix_node3.sh # tcp/60004
|
||
./run_mix_node4.sh # tcp/60005
|
||
```
|
||
|
||
Wait for `MixRlnSpamProtection started` + `Node setup complete` on each.
|
||
|
||
## 4. Start the two chat-ui clients
|
||
|
||
Two more terminals. The first run builds the whole chat chain via nix (slow once,
|
||
cached after):
|
||
|
||
```bash
|
||
cd simulations/mixnet/chat-ui
|
||
./run_chat_ui.sh A # ClientA, tcp/60010
|
||
./run_chat_ui.sh B # ClientB, tcp/60011
|
||
```
|
||
|
||
Each opens a GUI. The bottom status bar should show **`MIX 5/4`** (mix pool full)
|
||
with the send button enabled.
|
||
|
||
To iterate on a local chat-ui checkout instead of the pushed branch, clone it
|
||
**beside** `logos-delivery` and point `CHAT_UI` at it. The path is relative to
|
||
where you run the script (`run_chat_ui.sh` resolves it to an absolute path):
|
||
|
||
```bash
|
||
# from the same parent dir where you cloned logos-delivery:
|
||
git clone -b feat/logos-testnetv02-mix https://github.com/logos-co/logos-chat-ui
|
||
|
||
# then, back in simulations/mixnet/chat-ui:
|
||
CHAT_UI=../../../../logos-chat-ui ./run_chat_ui.sh A
|
||
```
|
||
|
||
## 5. Exchange messages
|
||
|
||
1. In **ClientA**: click **My Bundle** and copy the intro bundle.
|
||
2. In **ClientB**: **+ new** → paste A's bundle + a first message → create.
|
||
3. Send messages each way — they route A↔B **through the mixnet**.
|
||
|
||
> ⏱️ **Messages arrive in ~1 s.** A mix send takes ~0.8–1.1 s end-to-end: the Sphinx
|
||
> forward path reaches the exit hop in ~0.4–0.7 s, plus RLN proof time. (An earlier
|
||
> build had a hardcoded 30 s per-send "root convergence" wait — a leftover from
|
||
> dynamic on-chain membership — that made delivery take ~a minute; it was removed,
|
||
> since static RLN membership has a fixed Merkle root with nothing to converge.)
|
||
|
||
To confirm a message actually traversed the mix, watch a mix node log for
|
||
`onMessage - exit is destination` (the exit hop delivering to the recipient) — its
|
||
timestamp lands within ~0.5 s of when the message was sent.
|
||
|
||
## How the config flows
|
||
|
||
`run_chat_ui.sh` sources `env.sh` and sets the env `logos-chat-ui` reads into the
|
||
chat config:
|
||
|
||
| env | meaning |
|
||
|-----|---------|
|
||
| `CHAT_MIX_NODES` | the 5 mix nodes as `multiaddr:curve25519pubkey` — **both the node ID and the mix pubkey are hardcoded**; the chat does no mix-peer discovery |
|
||
| `CHAT_STATIC_PEER` | bootstrap node (relay the chat connects to) |
|
||
| `CHAT_CLUSTER_ID` / `CHAT_SHARD_ID` | `2` / `0` (must match the mix nodes) |
|
||
| `CHAT_NODEKEY` | adopt a provisioned chat identity so its RLN keystore resolves |
|
||
| `CHAT_MIX_REQUIRED` | force Required (mix) mode |
|
||
| `CHAT_UI` | flake to run (default: the pushed `feat/logos-testnetv02-mix` branch) |
|
||
|
||
## Notes / limitations
|
||
|
||
- **TRACE logging is heavy:** 5 mix nodes at `log-level=TRACE` (the default in
|
||
`config*.toml`) write multi-GB logs and burn CPU/disk — a resource concern, not a
|
||
correctness one. For long local runs, lower `log-level` or prefer fleet nodes.
|
||
- **Static membership only:** the two chat slots are fixed. More clients means
|
||
adding their peer-IDs to `setup_credentials.nim` + re-running `build_setup.sh`.
|
||
- **No mix discovery on the client:** every mix node must be listed in
|
||
`CHAT_MIX_NODES` (multiaddr **and** curve25519 mix pubkey). The mix *nodes*
|
||
discover each other via kad; the chat client does not.
|
||
- **Creds are not checked in** — `rln_tree.db` / `rln_keystore_*.json` are binary
|
||
and regenerated by `build_setup.sh`.
|