@staff0rd/assist 0.775.6 → 0.777.0

package/README.md

@@ -403,7 +403,7 @@ Rules are `- **<code>** — **<title>** — <text>` bullets, the title optional,
 
 Web sessions are owned by a long-lived daemon process, not the web server: the server is a thin client relaying WebSocket traffic to the daemon over a local IPC socket (`~/.assist/daemon/daemon.sock`; per-user named pipe `\\.\pipe\assist-sessions-daemon-<user>` on Windows). Restarting the web server leaves sessions running with scrollback intact. The daemon logs to `~/.assist/daemon/daemon.log` and auto-exits once no sessions remain and no client has connected for 60 seconds. See [docs/session-lifecycle.md](docs/session-lifecycle.md).
 
-When assist is installed as a git clone, each daemon keeps that install current in the background: it runs `assist watch wait --pull --build` in the clone lap after lap, with no session card, logging each lap to `daemon.log` (prefixed `auto-update:`) and to the clone's history in `~/.assist/watchers/`. If the clone cannot fast-forward, the daemon starts one claude session in it that rebases (or merges) onto the upstream without force-pushing or resetting, pushes, rebuilds and closes; the loop pauses until that session ends. A rebuilt version takes effect once the daemon and web server are restarted. Turn it off with `assist config set autoUpdate.enabled false -g`.
+When assist is installed as a git clone, each daemon keeps that install current in the background: it runs `assist watch wait --pull --build` in the clone lap after lap, with no session card, logging each lap to `daemon.log` (prefixed `auto-update:`) and to the clone's history in `~/.assist/watchers/`. If the clone cannot fast-forward, the daemon starts one claude session in it that rebases (or merges) onto the upstream without force-pushing or resetting, pushes, rebuilds and closes; the loop pauses until that session ends. A rebuilt version takes effect once the daemon and web server are restarted. The version badge in the web UI opens the Updates page, which lists every node with its running and built version, restart advice and history, and offers Restart per node or for every ready node, Pause/Resume (a pause persists across daemon restarts) and Check now (an immediate fetch). Turn it off with `assist config set autoUpdate.enabled false -g`.
 
 The topnav's **+** button (or Ctrl+N / Alt+N; Cmd+N on macOS) opens the new-session dialog, which replaces the old draft / bug / prompt / design topnav buttons. Its mode selector picks `draft`, `bug`, `prompt` or `design`; in `prompt` mode a harness selector under it picks Claude, Codex or pi when those are exposed, and Up/Down moves between the two rows. `design` launches an interactive `claude` session with the vendored design system prompt appended via `--append-system-prompt`. Left/Right change the mode or harness, and Tab steps through a selector's options before moving on to the next control.
 
@@ -434,7 +434,7 @@ Each assist install is a **node** with its own daemon and web server. A node can
 - `assist sessions nodes logs <name> [-n, --lines <count>] [--json]` — tail a linked node's `daemon.log`; naming this node reads the local log.
 - `assist sessions nodes restart <name> [--target daemon|webserver|both]` — the CLI form of **Restart <node>**: posts the node's `/api/restart`, waits up to 90s for its link to reconnect and prints its version.
 - `assist sessions nodes update <name>` — the CLI form of **Update assist on <node>**: over a connected link it opens an `assist update` card on the node, waits for its daemon to come back, then restarts its web server; otherwise it posts the node's `/api/self-update`. It then prints the node's new version.
-- `sessions.linkVersionCheck` — reaction to a version mismatch with a linked node: `block` (default) updates an older peer and holds the link until the versions match; `warn` proceeds anyway; `off` skips the check.
+- `sessions.linkVersionCheck` — reaction when a linked node's protocol range doesn't overlap this node's (links use the highest shared protocol, whatever the app versions, and never restart a peer): `block` (default) holds the link `version-blocked` until the older node is updated; `warn` proceeds anyway; `off` skips the check.
 
 `nodes restart` and `nodes update` log `nodes <action> <name> trace=<id>` here and send that trace id to the node, and when the node can't be reached or doesn't come back they name the first broken `nodes doctor` hop.