Daemon now loads `[shared].registry_url`, `[shared].token`, and the
`[daemon]` section from the unified file via maestro-config. The
registry-WS endpoint is derived from `[shared].registry_url` (scheme
swap + `/daemon`), so operators configure a single base URL instead of
duplicating the daemon-specific endpoint.
Env-var renames (all per the new `MAESTRO_<SECTION>_<FIELD>` convention):
- MAESTROD_REGISTRY_URL → MAESTRO_SHARED_REGISTRY_URL (and the value is
now an http(s) base, not the ws daemon endpoint)
- MAESTROD_DAEMON_URL → MAESTRO_DAEMON_URL
- MAESTROD_DAEMON_ID → MAESTRO_DAEMON_ID
- MAESTROD_BIND_ADDR → MAESTRO_DAEMON_BIND_ADDR
- MAESTROD_CONFIG → MAESTRO_CONFIG
- MAESTRO_TOKEN → MAESTRO_SHARED_TOKEN
`maestrod configure` now writes the unified file with read-modify-write
so any existing `[cli]` / `[registry]` sections on the same host are
preserved. The interactive prompts reflect the new shape: registry URL
is asked for as a base URL, and the token prompt names the new env var.
Missing-file behaviour is hard cutover: if the config doesn't exist and
not every required field comes from env, the daemon errors with a
message pointing at `maestrod configure` and explicitly noting that the
legacy paths (~/.config/maestrod/config.toml) are no longer read.
e2e harness updated to match: `Daemon::spawn` takes the registry's HTTP
base, and the spawned subprocess sees the renamed env vars.
CLI now reads `[shared].registry_url` and `[cli].cookie` from the unified
file via maestro-config instead of `~/.config/maestro/cli.toml`. The
two persistence paths are now distinct:
- `maestro set-registry <url>` → writes `[shared].registry_url`
- captured Set-Cookie on first contact → writes `[cli].cookie`
Both go through read-modify-write so they preserve any `[daemon]` /
`[registry]` content already in the file.
Env-var renames on the CLI flag layer:
- `--registry-url` reads `MAESTRO_SHARED_REGISTRY_URL` (was `MAESTRO_REGISTRY_URL`)
- `--config` reads `MAESTRO_CONFIG` (was `MAESTRO_CLI_CONFIG`)
Default config path is now `~/.config/maestro/config.toml`.
Introduces the shared on-disk schema that all binaries (CLI, daemon,
registry) will load from `~/.config/maestro/config.toml`. Sections:
[shared] registry_url, token — consumed by 2+ binaries
[cli] cookie
[daemon] daemon_id, hostname, daemon_url, bind_addr, state_dir, worktree_root
[registry] bind_addr
Exposes:
- MaestroConfigFile { shared, cli, daemon, registry } with strict load,
load-or-default, atomic save, and per-section empty-table suppression.
- default_config_path() / resolve_config_path() honoring MAESTRO_CONFIG.
- derive_daemon_ws_url() for the daemon's scheme-swap + /daemon suffix
off the shared http(s):// base URL.
- ConfigError::MissingFile with a hard-cutover message pointing the user
at `maestrod configure` (legacy paths are no longer read).
Consumers wired in follow-up commits.
Earlier today I picked option 1 from NAP-231's open question (run
`maestro-registry` locally on a different port). That was wrong: the
user-facing framing is "local web UI", and option 2 — a thin front-end
that serves the web bundle and reverse-proxies `/api/*` to whatever
registry the daemon is already registered with — is what actually
delivers that.
What the new crate does:
- `crates/maestro-web` ships a single binary `maestro-web`.
- It embeds the web bundle the same way `maestro-registry` does
(rust-embed against `web/dist/`).
- Mounts the bundle at `/` and a reverse proxy at `/api/*` (catch-all).
- HTTP proxy via `reqwest` — forwards method, headers (minus
hop-by-hop), body, and passes upstream response headers (including
`Set-Cookie`) back to the browser.
- WS proxy upgrades both sides and shuttles frames with
`tokio::select!`. Cookies on the initial `WebSocket` upgrade headers
are forwarded so the registry sees the same browser session.
- Reads `MAESTRO_REGISTRY` (upstream URL) and `MAESTRO_WEB_BIND_ADDR`
(default `127.0.0.1:8888`) from env.
- An integration test (`tests/proxy.rs`) stands up a mock upstream,
proxies an HTTP request with cookies through, and round-trips a text
WebSocket frame to verify the WS path.
Installer changes:
- Wizard now `cargo install`s `maestro-web` (not `maestro-registry`)
when the user opts into the web service.
- Unit template switched: `ExecStart=%h/.cargo/bin/maestro-web` plus
`Environment=MAESTRO_REGISTRY=<upstream>` +
`Environment=MAESTRO_WEB_BIND_ADDR=<bind>`.
- New `--web-registry <URL>` flag. Default in interactive mode: read
`registry_url` straight out of `~/.config/maestrod/config.toml`, so
the proxy lines up with what the daemon is using without further
prompting. Non-interactive runs without either of those error out
rather than guess.
Docs:
- `docs/INSTALL.md` updated unit body + flag table.
- `CLAUDE.md` repo-layout block now lists `maestro-web`.
148 tests pass workspace-wide (9 new in maestro-web, including the
proxy integration test; 15 in installer).
When a session terminates (the user runs `exit`, tmux is killed, or
`maestro kill` is called) the daemon's per-session broadcast channel
closes and `serve_attach` breaks out of its loop. The socket was then
dropped without a Close frame, so the client's `stream.next()` would
hang and the terminal stayed stuck in raw mode.
- daemon: send `Message::Close(None)` before returning from
`serve_attach` whenever the session itself ended (broadcast closed,
or session was killed between auth and subscribe).
- client: track remote-close, restore the TTY first, then print
`[maestro] session ended` so the message lands in cooked mode.
- tests: cover the kill path and the liveness-watcher reap path on the
daemon side.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
After the daemon-unit step, the wizard now offers a second user unit
(`maestro-web.service`) that runs `maestro-registry` on 127.0.0.1:8888
so the web UI is reachable in a browser locally.
Decisions:
- The unit runs `maestro-registry` directly — the web bundle is already
embedded in that binary, so a separate `maestro-web` binary would be
pure cost. `Description=` and the unit filename make the user-facing
framing clear.
- `Wants=maestrod.service` + `After=` (not `Requires=`): the registry
process doesn't actually need the daemon to start; the web UI just
has nothing to show without one. Soft dependency expresses that.
- Default loopback bind is `127.0.0.1:8888` to make it obvious which
port is the local UI vs a separately deployed registry on `:8080`.
Wizard surface:
- New prompt (default *no*) after the daemon-systemd step.
- New flags: `--no-web`, `--web-bind <addr>`, `--web-exec-start <path>`.
- Skipped automatically when systemd is absent or `--no-systemd` is set.
- The re-run/update path mirrors the daemon flow: refresh the unit,
daemon-reload, restart instead of start.
- `maestro-registry` is only `cargo install`ed when the user opts in,
keeping the no-web install footprint unchanged.
docs/INSTALL.md gets the new flags + a full sample unit body.
New `crates/maestro-e2e` exercises the full headless stack — registry,
two daemons, real HTTP+WS — as actual subprocesses.
Harness pieces:
- `Registry::spawn` and `Daemon::spawn` launch the real built binaries
with `env_clear()` plus an isolated `$HOME`, `XDG_CONFIG_HOME`,
`XDG_STATE_HOME`, and (for daemons) `TMUX_TMPDIR`. Drop guards kill
the child and clean the temp dirs.
- `Daemon::stop` + `Daemon::restart` swap the process without losing
the state dirs, so the recovery test exercises a real tmux
re-adoption.
- A `build::workspace_target_bin` helper re-invokes cargo from the
test process to build `maestrod` / `maestro-registry` on demand
(cargo only ships libs as path-dep of e2e by default).
- `TestClient` wraps reqwest + tokio-tungstenite for the public
HTTP+WS surface — list/spawn/kill/connect-info/attach + events WS.
Scenarios covered (`tests/e2e.rs`):
- list sessions from two daemons through one registry
- multi-attach: two WS clients see the same PTY output
- worktree with `cleanup_on_kill: true` is removed after kill
- daemon restart: catalog drops then repopulates from `recover()`
- concurrent spawns all get distinct ids
All five pass in ~0.7s in parallel; cleanup confirmed (no leftover
processes or temp dirs).
The daemon never noticed when a session's command exited and tmux tore
the session down — the in-memory `SessionManager` map plus the registry
catalog kept showing it as `Running`, and a follow-up `maestro attach`
would hang on an empty broadcast.
Adds a per-host background task that polls `tmux list-sessions` every
2s, diffs against the in-memory set, and reaps anything we still track
but tmux has lost: drops the record, deletes the sidecar + FIFO,
cleans up its worktree if requested, and emits `SessionEvent::Removed`
so the registry purges its catalog.
The task holds a `Weak<Inner>` so it exits cleanly once the last
`SessionManager` clone is dropped. Skips the tmux subprocess when the
in-memory map is empty.
The wizard now treats existing state as "we're updating, not installing":
- If `~/.config/maestrod/config.toml` exists, skip the configure prompt with
a clear notice. Sets `configured = true` so the systemd start logic still
makes sense.
- If the systemd user unit already exists, refresh the file in place, skip
the enable prompt (it's either already enabled or the user said no last
time), and ask to `restart` instead of `start` — so the new binary
actually takes effect.
Adds `SystemctlAction::Restart` and a test for it. Same `./install.sh` works
for both first install and subsequent updates; no `--update` flag.
Refs NAP-225.
New `maestro-installer` crate exposing the `maestro-install` binary. Bootstrap
flow: `cargo install --git ... maestro-installer && maestro-install` runs
`cargo install` for `maestro-daemon` and `maestro-cli`, optionally drives
`maestrod configure`, and on systemd hosts writes a user unit at
`~/.config/systemd/user/maestrod.service` plus reload/enable/start. Pure
builders (cargo install args, systemctl args, unit body) live in the lib and
are unit-tested; main.rs is the prompt glue.
Refs NAP-225.
The daemon's outbound registry client uses `connect_async`, which
errors with "TLS support not compiled in" against `wss://` URLs unless
the crate is built with a TLS feature. CLI got `rustls-tls-webpki-roots`
in NAP-216; daemon needs the same so it can talk to a TLS-terminated
registry like maestro-registry.n62.casa.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`configure::run` printed the "wrote" line whenever stdin was a TTY,
and `main::configure` printed it again unconditionally — so an
interactive run printed it twice. Library now stays silent; the binary
is the single owner of that confirmation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes NAP-224. Daemon onboarding now mirrors the registry's NAP-221 /
NAP-222 ergonomics: a fresh host needs three values, not eight.
Config (`crates/maestro-daemon/src/config.rs`):
- Split into `DaemonConfigFile` (on-disk, every field `Option<>`) and
`DaemonConfig` (resolved, all fields present).
- `load_config(path, EnvOverrides)` resolves env > file > default and
returns clear `Missing { field, env, path }` errors for the three
fields that genuinely require a human: `registry_url`, `daemon_url`,
`token`.
- Defaults: `bind_addr = 127.0.0.1:7070`; `state_dir` /
`worktree_root` from XDG; `hostname` from `gethostname(2)`;
`daemon_id` defaults to `hostname`.
- Treats a missing config file as empty so an env-only deploy works.
- Atomic file save (tmp + rename) for `configure` to share.
Env vars (mirrors NAP-222):
- `MAESTRO_TOKEN`
- `MAESTROD_REGISTRY_URL`, `MAESTROD_BIND_ADDR`,
`MAESTROD_DAEMON_URL`, `MAESTROD_DAEMON_ID`
- `MAESTROD_CONFIG` for the file path
`maestrod configure` (`configure.rs` + clap in `main.rs`):
- Interactive bootstrap when stdin is a TTY: prompts for the three
required fields, defaults the rest.
- Non-interactive mode (`--non-interactive` or non-TTY stdin) takes
values from flags only; errors out listing the first missing one.
- Refuses to overwrite an existing config without `--force`.
- Writes via the same atomic-save path used by config-file edits.
`maestrod` grew a clap CLI: `run` (default) and `configure` subcommands,
both honour `--config` / `$MAESTROD_CONFIG`.
Verified end-to-end:
- `maestrod configure --non-interactive --registry-url ... --daemon-url
... --token ...` writes a valid file and exits 0.
- `--non-interactive` without all required flags errors with the
missing field name.
- Existing file + no `--force` is rejected and the file is untouched.
- Workspace `cargo test`: 115 passed (was 97; +18 new daemon tests
across config + configure).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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>
Closes NAP-216. Thin client for the registry HTTP API + per-session
daemon WebSocket. Becomes the headless test harness for NAP-217 and
unblocks the demo-by-script path before the web client lands (NAP-218).
Crate layout (`crates/maestro-cli/`):
- `cli.rs` — clap-derive parser. Commands: `list`, `spawn`, `attach`,
`kill`, `set-registry`. Global `--registry-url` (env
`MAESTRO_REGISTRY_URL`), `--config`, repeatable `-v`.
- `config.rs` — `CliConfig` persisted at `~/.config/maestro/cli.toml`,
storing the default registry URL + the session cookie issued by the
registry's cookie middleware.
- `http.rs` — `RegistryClient` wrapping `reqwest`. Attaches the
persisted cookie, captures `Set-Cookie` responses (first-contact
bootstrap and rotation), atomically writes the new value back to
the config file.
- `sessions.rs` — `list/spawn/kill` + connect-info fetch. `list` prints
a per-daemon grouped table; `--json` for scripting.
- `attach.rs` — raw-TTY WS proxy. Termios cfmakeraw with an RAII guard
for restore-on-drop, SIGWINCH → `ClientControl::Resize` text frame,
configurable detach key (default Ctrl-q; accepts `^X` form), echoes
`Ping`/`Pong` for keepalive symmetry with the daemon side.
Workspace deps added: clap (derive+env), reqwest (rustls+cookies),
nix (term+signal).
Smoke-tested against the live registry at maestro-registry.n62.casa:
`maestro list` issued + persisted the cookie and printed the empty
catalog cleanly. 19 new unit tests (parser, config round-trip,
table-printing, cookie extraction, attach URL/detach-key parsing).
Full workspace `cargo test`: 97 passed.
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>
`tokio::fs::File` routes every `.read()` through `spawn_blocking`. On
an idle FIFO opened r+w, that thread blocks in `read()` forever — and
when the test runtime tries to shut down, the blocking-pool thread is
stuck on the syscall and the test binary hangs (which is exactly what
just took out CI run 11).
Switch to `std::fs::File` opened `O_RDWR | O_NONBLOCK` wrapped in
`tokio::io::unix::AsyncFd`. Reads now drive off epoll readiness with
`try_io`; no blocking-pool threads, no hung shutdown.
`Drop` still aborts the task so per-test cleanup is clean.
Completes the headless daemon stack for NAP-214.
- `config`: daemon TOML config (daemon_id, registry_url, bind_addr,
token, state_dir, worktree_root) with sensible XDG defaults.
- `jwt`: HS256 verifier; the daemon validates per-attach JWTs locally
with no callback to the registry. Also exposes `issue()` for tests.
- `pty::PtyBridge`: per-session FIFO + tokio broadcast pump. The FIFO is
opened r+w so it never sees zero writers — stays alive across
`pipe-pane` re-issues. Helper `pipe_pane_command()` builds the
`cat >> <fifo>` shell line.
- `session::SessionManager`: authoritative live-session store. Creates
worktrees (when requested), spawns tmux, mkfifo + pipe-pane, writes
sidecar; tears it all down on kill. Emits diff events on a
broadcast for the registry client. `recover()` adopts surviving
`maestro-*` tmux sessions on startup and re-pipes them.
- `registry_client::RegistryClient`: persistent outbound WS to the
registry with exponential backoff. Sends Hello → snapshot, drives a
select loop that pumps inbound CreateSession/KillSession RPCs into
the manager (replying with results), emits SessionAdded/Removed/
Updated diffs from the manager, and heartbeats every 10s.
- `attach`: axum router mounting `WS /sessions/:id/stream?token=<jwt>`.
Binary frames carry PTY bytes either way (subscribe to broadcast →
send to client; receive from client → tmux send-keys -l). Text
frames carry `ClientControl::{Resize,Ping,Pong}`.
- `bin/maestrod`: ties everything together, loads config from
`MAESTROD_CONFIG` env or `~/.config/maestrod/config.toml`,
recovers, spawns the registry client, binds the WS server.
Also adds `JwtClaims` to `maestro-protocol` so the daemon and the
future registry JWT issuer (NAP-215) agree on the wire shape.
Integration tests cover session manager round-trip + recovery + PTY
broadcast; attach endpoint JWT (bad/wrong-session/wrong-daemon),
PTY round-trip through real tmux, multi-attach broadcast, ping/pong;
registry client registration + snapshot + diff forwarding +
CreateSession RPC round-trip.
Closes NAP-214.
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.
Local primitives for `maestrod`, no network code yet:
- `tmux`: subprocess wrapper around new-session, kill-session,
list-sessions (filtered to maestro-*), send-keys -l, pipe-pane -o,
resize-window, kill-server.
- `sidecar`: atomic JSON metadata store keyed by session UUID, with
load_all for daemon restart recovery.
- `worktree`: `git worktree add -b` with collision-proof generated paths
and `git worktree remove --force` for cleanup_on_kill.
Integration tests use real tmux (isolated via unique socket + temp
TMUX_TMPDIR) and real git (tmpdir repo).
Closes NAP-212.
Introduces the maestro-protocol crate (NAP-211): the shared serde
types that the daemon, registry, and clients all serialize across
WebSocket boundaries. Covers:
- Session, SessionStatus
- CreateSessionRequest with optional WorktreeSpec
- RpcResult<T> reply envelope
- DaemonOutbound (Hello, snapshot, add/remove/update,
Heartbeat, request results)
- DaemonInbound (Ack, Error, CreateSession, KillSession)
- ClientControl text frames (Resize, Ping, Pong, Error)
Schema mirrors the wire-protocol tables in the design spec. Unit
tests round-trip every variant and assert tag/discriminator
stability so future renames break loudly.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
naps62-yolo
