Stage 7a-i: TP worker lifecycle scaffolding

Leader → worker process plumbing for tensor parallelism. The neuron
binary picks up two modes: default (the existing daemon, axum + HTTP)
and `--worker` (a bare RPC loop driven over stdin/stdout). The leader
spawns one worker per non-zero NCCL rank via tokio::process::Command
on the same binary path (production: /proc/self/exe; tests:
env!("CARGO_BIN_EXE_neuron")) and talks to each over newline-
delimited JSON.

Protocol (harness/tp/rpc.rs) is serde-tagged from the start —
WorkerRequest::{Ping, Init, NcclSanityCheck, Shutdown} and
WorkerResponse::{Pong, InitOk, NcclSanityResult, Bye, Error}, both
`#[serde(tag = "op", rename_all = "snake_case")]`. Adding ops in 7b/7c
is purely additive; unknown ops on the wire fail to parse (verified
in unit tests).

7a-i scope:
- WorkerPool::spawn(binary, world_size, devices) forks ranks 1..N as
  subprocesses, captures stdin/stdout, kills on drop.
- ping_all() round-trips a Ping to every worker and validates the
  returned rank.
- shutdown() sends Shutdown to each worker, awaits Bye, reaps.
- Worker mode: parse Ping/Shutdown, return Pong/Bye; Init and
  NcclSanityCheck return Error{kind="not_implemented_7a_i"} so a 7a-ii
  binary speaking the same wire is a drop-in replacement (the kind
  field signals "real NCCL lands in the next commit").
- CandleHarness::load_model refuses tensor_parallel > 1 with a clear
  message until 7b is in.

Three integration tests in tests/tp_worker_lifecycle.rs cover spawn/
ping/shutdown for 2- and 3-worker pools, plus the
not_implemented_7a_i contract test for Init. Seven rpc serde unit
tests assert the wire shape (op tags, field names, unknown-op
rejection). All pass on the dev host; no CUDA required.

Stage 7a-ii (next): the real NCCL Comm::from_rank wiring behind the
existing Init/NcclSanityCheck op surface, CUDA-gated. Verifiable on
beast's 2×5090.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

crates/neuron/src/main.rs

@@ -1,21 +1,52 @@
 use anyhow::Result;
 use clap::Parser;
-use neuron::{api, config::NeuronConfig, discovery, harness::HarnessRegistry, health, startup};
+use neuron::{
+    api,
+    config::NeuronConfig,
+    discovery,
+    harness::{HarnessRegistry, tp},
+    health, startup,
+};
 use std::sync::Arc;
 use std::time::Instant;
 use tokio::sync::RwLock;
 use tracing_subscriber::EnvFilter;
 
+/// Top-level CLI. The same binary runs as either the public neuron
+/// daemon (default) or a tensor-parallel worker subprocess (when
+/// `--worker` is set) spawned by the leader on the same host.
 #[derive(Parser)]
 #[command(name = "neuron")]
 #[command(about = "Per-node daemon for cortex inference clusters")]
 #[command(version)]
 struct Args {
-    /// Port to listen on (overrides config file).
+    /// Run in tensor-parallel worker mode. The leader process spawns
+    /// one of these per non-zero NCCL rank and drives it over
+    /// newline-delimited JSON on stdin/stdout. Worker mode skips
+    /// discovery, the HTTP listener, and the health poller — it's a
+    /// pure RPC loop.
+    #[arg(long, default_value_t = false)]
+    worker: bool,
+
+    /// NCCL rank for worker mode. Ignored when `--worker` is not set.
+    #[arg(long, default_value_t = 0)]
+    rank: u32,
+
+    /// Total NCCL world size for worker mode. Ignored when `--worker`
+    /// is not set.
+    #[arg(long, default_value_t = 1)]
+    tp_size: u32,
+
+    /// CUDA device index for worker mode. Ignored when `--worker` is
+    /// not set.
+    #[arg(long, default_value_t = 0)]
+    cuda_device: u32,
+
+    /// Port to listen on (overrides config file). Daemon mode only.
     #[arg(short, long)]
     port: Option<u16>,
 
-    /// Path to the neuron config file.
+    /// Path to the neuron config file. Daemon mode only.
     #[arg(short, long, default_value = "neuron.toml")]
     config: String,
 }
@@ -23,6 +54,7 @@ struct Args {
 #[tokio::main]
 async fn main() -> Result<()> {
     tracing_subscriber::fmt()
+        .with_writer(std::io::stderr)
         .with_env_filter(
             EnvFilter::try_from_default_env()
                 .unwrap_or_else(|_| EnvFilter::new("info,neuron=debug")),
@@ -31,6 +63,19 @@ async fn main() -> Result<()> {
 
     let args = Args::parse();
 
+    if args.worker {
+        return tp::worker::run(tp::worker::WorkerConfig {
+            rank: args.rank,
+            world_size: args.tp_size,
+            cuda_device: args.cuda_device,
+        })
+        .await;
+    }
+
+    daemon(args).await
+}
+
+async fn daemon(args: Args) -> Result<()> {
     let cfg = NeuronConfig::load(&args.config).unwrap_or_else(|e| {
         tracing::warn!(path = %args.config, error = %e, "config not found, using defaults");
         NeuronConfig::default()