Release 0.6.24

docs/release-notes/0.6.24.md

@@ -0,0 +1,53 @@
+**This release adds dashboard control of the host Apollo's USB forwarding, keeps USB forwarding off for seats, and fixes a 30-second delay when the service stops with a dashboard open.**
+
+## What changes
+
+- **A USB forwarding panel on the host card** (#135). It shows whether the console Apollo's USB tunnel is on, what is forwarded now, and which clients are paired. You can turn the tunnel on or off and block or allow a client's USB forwarding.
+  - It talks to the console Apollo through a local control pipe. That needs ApolloVibe `2026.6.1-ms7` or newer, with "USB Forwarding Control Pipe" ticked in Apollo's Network tab. Without that, nothing changes.
+  - MultiSeat never edits Apollo's config. A switch made in the panel lasts until that Apollo restarts, and the panel says so.
+  - **Only the person at the console can use it.** The API has no key on a host that chose that, and every seat user can reach it on loopback, so the whole `/api/host/usb` group, reading included, is refused to anyone else. A dashboard on another PC, or inside a seat, sees a USB card that says it cannot control USB forwarding.
+- **Seats never forward USB** (#133). Each seat's generated Apollo config now writes `usb_forwarding_enabled = disabled`, and the seat firewall rules never open the USB tunnel port (main port + 7). The `usbip-win2` driver attaches a device to the whole machine, so a device forwarded into one seat would show up for every account on the host.
+- **An opt-in driver install step** (#133): `install-prerequisites.ps1 -WithUsbForwarding`. It installs `usbip-win2` v0.9.8.1 after checking the file's SHA-256 and the installer's signature and signer name. It then checks that the installed drivers chain to Microsoft. It is never part of a default or unattended run and it never reboots; a reboot is needed before the driver works, and the script leaves that to you. `-UsbForwardingDryRun` prints every step and changes nothing.
+- **The service stops promptly while a dashboard is open** (#131). On 2026-10-21 a stop took about 31 seconds because an open dashboard socket held up shutdown. The socket now closes when the host stops. In the new test, a stop that took 30.0 seconds before the fix takes about 0.1 seconds.
+- Docs: audio for the `PerSession` default, with VoiceMeeter only for `SharedHost` (#130). A new test kills a stray Apollo using real processes (#132).
+
+## Who this affects
+
+- **Everyone:** the faster service stop.
+- **Nobody else, unless they opt in:** the USB panel needs Apollo's control pipe turned on, and the driver step needs its own switch. Seats are forced to USB off whatever the host does.
+- USB forwarding itself needs ApolloVibe `2026.6.1-ms7` and MoonlightVibe 6.3.11 or newer. It is experimental. Those releases carry the setup steps.
+
+## How this was checked
+
+- **Unit tests:** CI passed on `master` for every change here. The suite on the last change finished with 1927 passed, 18 skipped and 0 failed, and the dashboard tests passed 82 of 82. Each protection in the panel's access check, and each check in the driver step, was removed on purpose and a test failed every time.
+- **An independent review** of the driver step found no Critical issues. Its two High findings and its Medium findings were fixed and tested.
+- **USB forwarding end to end** was tested on 2026-10-29 with ApolloVibe `ms7`'s `sunshine.exe`, MoonlightVibe 6.3.11 and a USB stick. That test did not go through the MultiSeat panel or the driver step.
+
+## Not verified
+
+- **The USB panel in a real browser.** It has not been compared between a browser on the console and one inside a seat. The access rule is tested with this process as the caller, not from a live seat session.
+- **A seat user moving their session to the console** (`tscon`) is covered by a unit test of the rule, not measured on a live seat.
+- **The driver install step** has not run for real: no real download, and no check on a freshly installed driver. If the driver check is too strict it refuses, which is safe.
+- **The faster stop** is measured by the test, not by a stop on a live host.
+
+## Known limitations
+
+- **After a Quit, the Moonlight client still lists the seat but cannot connect** until Apollo is started from the dashboard.
+- **On at least one host the scale override is still ignored** (issue #93). This release does not change that.
+- Seats capture the remote desktop surface, not a dedicated virtual display. That is a Windows constraint.
+- Steam isolation is experimental and off by default.
+
+## Install
+
+```powershell
+.\prerequisites\install-prerequisites.ps1     # drivers - needed either way
+.\scripts\install-service.ps1 -FromZip .
+```
+
+Afterwards, confirm what landed:
+
+```powershell
+& 'C:\Program Files\MultiSeat\MultiSeat.Service.exe' --config
+```
+
+It should report `0.6.24`.

version.txt

@@ -1 +1 @@
-0.6.23
+0.6.24