`Registry::deregister` removed the catalog entry unconditionally. If a daemon's
old WS handler hit heartbeat timeout *after* the daemon had reconnected on a
new socket, the stale handler's deregister wiped the freshly-installed live
entry and broadcast DaemonRemoved + SessionRemoved for all its sessions.
`register` now mints a per-connection fencing token (uuid) stored on the
DaemonEntry and returned to the handler, which holds it for the connection's
life. `deregister(daemon_id, token)` only tears down the entry when the stored
token matches, so a superseded handler's late deregister is a no-op.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
`dispatch_text` (registry) and `handle_inbound` (daemon) treated any
frame that didn't deserialize into a known `DaemonOutbound`/`DaemonInbound`
variant as fatal: the registry returned `false` (breaking the handler
loop and deregistering the daemon, wiping its catalog), and the daemon
propagated the serde error out of the read loop. A newer peer emitting a
not-yet-merged variant would therefore get stuck in a tight
disconnect/reconnect loop.
- Add a `#[serde(other)] Unknown` catch-all to both enums so unknown
`type` tags decode instead of erroring.
- Log-and-skip the `Unknown` variant on both sides.
- Also log-and-skip genuinely malformed frames rather than tearing down
the connection, so one bad frame can't wipe a daemon's catalog.
Tested: unknown/malformed frames keep the connection alive (registry
handler), and unknown `type` decodes to `Unknown` (protocol round-trip).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The registry serves no UI — that is maestro-web's job (connection-
orchestration spec §"UI hosting"). The rust-embed of web/dist was
legacy from NAP-218, before maestro-web was extracted (NAP-241), and it
made the registry (and thus the umbrella maestro binary) fail to compile
whenever web/dist was absent — e.g. a fresh cargo install --git after
the tracked .gitkeep was dropped. Remove web_assets.rs, the rust-embed
dep, and the / + /assets routes; the registry is now JSON/WS API only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
`registry::run` now accepts an `Option<PathBuf>` config override so the
umbrella binary's `--config` flag (and its clap-bound `MAESTRO_CONFIG`
env var) reach the registry, mirroring how the daemon already consumed
the override.
New `crates/maestro/` umbrella binary hosts the CLI, daemon, and registry
behind Cargo features `cli`, `daemon`, `registry` (default = all three).
Library crates keep their structure; per-binary `main.rs` files are removed
and their dispatch logic is exposed as public functions:
- `maestro_cli::dispatch(command, config, registry_url)`
- `maestro_daemon::cli::dispatch(config, cmd)` (with public `DaemonCmd`/`ConfigureArgs`)
- `maestro_registry::run()`
Subcommand surface:
- `maestro list / spawn / attach / kill / set-registry / policy` (unchanged)
- `maestro daemon [run|configure]` (was `maestrod`)
- `maestro registry` (was `maestro-registry`)
Hard cutover — old binary names are not preserved.
Verified: default-feature build, plus `--no-default-features --features {cli|daemon|registry}`
each build clean with `-D warnings`.
* DaemonEntry stores the daemon's most recent perimeter_id / private_addr
from heartbeats, with `update_perimeter` and `perimeter_for` accessors.
* issue_pair_attach_jwt mints the iss/sub/aud/sid/exp pair claim shape
(60s TTL) — verified by the target daemon.
* Registry::resolve_attach assembles AttachCandidates: T1 stays None
(NAP-235 canceled), T2 carries the target's perimeter when known, T3
stays None (NAP-238 pending), plus the pair JWT.
* WS handler routes the new DaemonOutbound::ResolveAttach to the new
method and persists perimeter info on every Heartbeat.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Each daemon connection now also subscribes to the registry's event broadcast.
A per-connection task pumps peer-filtered events through the existing
outbound mpsc as `CatalogEvent` frames; the daemon side mirrors them
locally (NAP-246).
Two pre-existing race fixes that NAP-246's extra traffic exposed:
* `create_session` / `kill_session` now eagerly mirror Ok results into
state.sessions (via the same `add_session` / `remove_session` calls the
daemon's async `SessionAdded` / `SessionRemoved` will hit later — both
are idempotent). HTTP callers no longer see a transient empty catalog
immediately after a successful POST.
* Catalog frames flow through the same select! arm as RPC frames so the
WS sink+stream coupling axum imposes doesn't stall the inbound reader.
Updates daemon_ws.rs tests to skip CatalogEvent frames when asserting on
RPC frames.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Convert DaemonOutbound::Heartbeat from unit to struct variant with two
optional fields advertising the daemon's T2 reachability. Old payloads
({"type":"Heartbeat"} with no fields) still parse — both fields default
to None — and a new daemon that doesn't populate them serializes back to
the same minimal shape.
Update the daemon's heartbeat ticker and the registry's match arm to the
new shape. Population from daemon config + registry persistence lands
with the broker logic in NAP-243.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The shared section is the implicit global scope, so its fields use
the bare `MAESTRO_<FIELD>` env-var spelling. Only `[daemon]` and
`[registry]` sections get section-prefixed env vars.
Renames:
- MAESTRO_SHARED_TOKEN → MAESTRO_TOKEN
- MAESTRO_SHARED_REGISTRY_URL → MAESTRO_REGISTRY_URL
`MAESTRO_DAEMON_*` and `MAESTRO_REGISTRY_BIND_ADDR` are unchanged.
This restores the canonical spec name (`MAESTRO_TOKEN`) that operators
already use in their deployment scripts — NAP-233 had renamed it to
`MAESTRO_SHARED_TOKEN`, which silently broke daemon auth against
existing registry deployments still setting `MAESTRO_TOKEN` in env.
Constants renamed alongside: `ENV_SHARED_TOKEN` → `ENV_TOKEN`,
`ENV_SHARED_REGISTRY_URL` → `ENV_REGISTRY_URL`. `EnvOverrides` fields
shed their `shared_` prefix to match.
Touches: daemon/cli/registry crates, e2e harness, design spec.
Registry now reads `[shared].token` and `[registry].bind_addr` from the
unified file via maestro-config. First-run token generation does a
read-modify-write so `[shared]` / `[cli]` / `[daemon]` sections that
already exist on a co-located host stay intact.
Env-var renames:
- MAESTRO_TOKEN → MAESTRO_SHARED_TOKEN
- MAESTRO_REGISTRY_CONFIG → MAESTRO_CONFIG
`MAESTRO_REGISTRY_BIND_ADDR` is unchanged. The startup banner now points
the operator at `MAESTRO_SHARED_TOKEN` when announcing a freshly
generated token.
e2e harness updated to use the new env var; auth.rs doc-comment updated
to reflect the new env-var name.
NAP-218. Add a node:20 stage to the registry Dockerfile that runs
`pnpm install && pnpm build` against `web/`, then copies the
resulting `dist/` into the Rust builder so rust-embed picks it up.
Single-binary deploy still — no separate static-asset volume.
`.dockerignore` keeps `web/node_modules` and `web/dist` out of the
build context (the node stage produces them fresh) while letting the
sources through.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
NAP-218. The registry's `GET /` now serves the built web bundle from
`web/dist/` (embedded into the binary at compile time via rust-embed)
instead of the NAP-215 placeholder, and `GET /assets/*` serves the
hashed JS/CSS chunks Vite produces.
When `web/dist` is empty (fresh checkout without `pnpm build`), a
small fallback page tells the operator to run the JS build. That
keeps `cargo run -p maestro-registry` usable without a node
toolchain installed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Lets `maestro-registry` run in container deployments (e.g. Dokploy)
without mounting a writable config volume.
- `MAESTRO_TOKEN`: when set, bypasses `registry.toml` entirely (no
read, no write). The container needs no persistent volume just to
seed the shared secret — pass it as a platform secret instead.
- `MAESTRO_REGISTRY_BIND_ADDR`: overrides the file's `bind_addr`.
Lets the container bind `0.0.0.0:$PORT` without rewriting TOML.
- New `load_config(path, env_token, env_bind_addr)` in `token.rs`
centralises the env > file > default precedence. Takes explicit
`Option`s so tests don't touch process-global env.
- `TokenError::InvalidBindAddr` carries the raw input + parse error
so misconfigured deploys log a useful diagnostic.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes NAP-221. Adds the `maestro-registry` bin so the registry can be
spawned as a child process (required by the NAP-217 E2E harness) and
run as a long-lived service.
- New `crates/maestro-registry/src/main.rs` with a `#[tokio::main]`
entry point: loads config via `ensure_token` (override path via
`MAESTRO_REGISTRY_CONFIG`), binds the configured `bind_addr`, serves
the production router, and shuts down cleanly on Ctrl-C. Mirrors the
`maestrod` wiring (EnvFilter tracing, ExitCode reporting).
- `[lib]` + `[[bin]]` sections in `crates/maestro-registry/Cargo.toml`
so the crate exposes both surfaces (same pattern as `maestro-daemon`).
- `RegistryConfig` gains a `bind_addr: SocketAddr` field with
`serde(default)` to default to `127.0.0.1:8080` per spec §Registry.
Existing TOML files without the field still load.
- `default_config_path` re-exported from the crate root for the bin.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the client-facing surface on the registry (NAP-215). Endpoints:
- GET /api/daemons — list connected daemons (with daemon_url)
- GET /api/sessions — flat session catalog
- POST /api/sessions — forwards CreateSession to chosen daemon, 504 on
10s daemon timeout
- DELETE /api/sessions/:id — forwards KillSession
- GET /api/sessions/:id/connect-info — returns {daemon_url, jwt} with a
5min HS256-signed attach JWT
- WS /api/events — snapshot + delta broadcast of catalog and presence
- GET / placeholder served until NAP-218 web bundle lands
Cookie auth middleware mints a long-lived httpOnly JWT cookie on first
request and reissues silently on missing/tampered cookies. Per spec the
perimeter is the operator's network — the cookie is a second factor.
Protocol additions: daemon_url on Hello, DaemonInfo, ConnectInfo,
RegistryEvent. Daemon config + RegistryClient now carry daemon_url.
Registry state now exposes an event broadcast channel (subscribe gives
a snapshot + tokio::sync::broadcast::Receiver) and helper lookups
daemon_for_session / daemon_url.
Tests: 11 reqwest-driven HTTP integration tests cover every endpoint,
the cookie flow, 504 timeout path, worktree forwarding, and events WS
propagation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Registry-side primitives for the daemon↔registry conversation. No
client-facing HTTP yet (that's NAP-215).
- `state::Registry`: in-memory store of connected daemons + a session
catalog keyed by (daemon_id, session_id). Tracks pending CreateSession
/ KillSession RPCs so a single daemon connection can multiplex them.
- `handler::handle_socket`: per-connection loop that authenticates the
Hello, applies snapshot/diff/heartbeat frames, and resolves matching
*Result frames. Heartbeat watchdog is a `tokio::time::timeout` around
the read side — no separate janitor task.
- `server::router_with_timeout`: axum router mounting `WS /daemon`. The
timeout knob lets tests verify the 30s drop path in milliseconds.
- `token::ensure_token`: first-run bootstrap that generates a 256-bit
hex token, prints it once to stderr, and persists it to
`~/.config/maestro/registry.toml`.
- Outbound `create_session` / `kill_session` exposed as the internal
API NAP-215's HTTP handlers will call.
Integration tests use a real axum server bound to `:0` and drive it via
`tokio-tungstenite`, covering: bad-token reject, snapshot accept, diff
apply (add/update/remove), heartbeat timeout dropping the daemon,
re-registration replacing the catalog, and a full CreateSession RPC
round-trip.
Closes NAP-213.
naps62-yoloandClaude Opus 4.8