openmausbot 0.1.94 → 0.1.96

package/dist-server/companion/src/listener.js

@@ -12,20 +12,34 @@
 import { execFile } from "node:child_process";
 import { homedir, networkInterfaces } from "node:os";
 import { delimiter, join } from "node:path";
-/** Interfaces that exist to tunnel, bridge or mesh traffic — utun (Tailscale
- * and every other VPN), vmnet/bridge (VMs, containers, internet sharing),
- * awdl/llw (AirDrop's side channels), feth/tap/tun. Their addresses stay in
- * the list, because the tailnet one is exactly what a phone off-network
- * dials — but a phone on the same wifi can reach none of them, so none of
- * them may come first. */
-const VIRTUAL_INTERFACES = /^(utun|tun|tap|bridge|vmnet|awdl|llw|feth)/;
-/** Lower sorts earlier. `en0`, `en1`, … are macOS's built-in wifi and
- * ethernet — the networks a phone is actually standing on. */
+/** Interfaces that exist to tunnel, bridge, virtualize or mesh traffic.
+ * Their addresses stay in the list — the tailnet one is exactly what a phone
+ * off-network dials, and a Hyper-V external switch moves the PC's real address
+ * onto a "vEthernet (…)" adapter — but a phone on the same Wi-Fi can reach
+ * none of the usual ones, so none of them may come first.
+ *
+ * Matched by name, case-insensitively, because that is all Node reports:
+ * - macOS: utun (every VPN, Tailscale), bridge/vmnet (VMs, internet
+ *   sharing), awdl/llw (AirDrop), feth, anpi.
+ * - Linux: docker0, br-<id> (Docker networks), veth, virbr (libvirt),
+ *   vboxnet, cni/flannel/lxc/lxd, tun/tap, wg, zt (ZeroTier), tailscale0.
+ * - Windows names adapters rather than numbering them: "vEthernet (WSL)",
+ *   "vEthernet (Default Switch)", "VirtualBox Host-Only Network", "VMware
+ *   Network Adapter VMnet8", "Tailscale", "ZeroTier One", anything with VPN
+ *   or Bluetooth in its name. */
+const VIRTUAL_INTERFACE = /^(utun|tun|tap|bridge|br-|vmnet|vboxnet|virbr|veth|docker|cni|flannel|lxc|lxd|zt|wg|tailscale|awdl|llw|feth|anpi)|vethernet|virtual|vmware|hyper-v|wsl|zerotier|vpn|loopback|bluetooth/i;
+/** The networks a phone is actually standing on: macOS en0/en1, Linux
+ * eth0/enp3s0/eno1/wlan0/wlp2s0, Windows "Wi-Fi", "Ethernet 2", and the
+ * German/Chinese "WLAN". */
+const PHYSICAL_INTERFACE = /^(en|eth|wl)[a-z0-9]*$|^(wi-?fi|wlan|ethernet)\b/i;
+/** Lower sorts earlier: physical, then anything unrecognized (a localized
+ * Windows name, a hotspot), then virtual. Virtual is checked first so
+ * "vEthernet" is never mistaken for "Ethernet". */
 const interfaceRank = (name) => {
-    if (/^en\d+$/.test(name))
-        return 0;
-    if (VIRTUAL_INTERFACES.test(name))
+    if (VIRTUAL_INTERFACE.test(name))
         return 2;
+    if (PHYSICAL_INTERFACE.test(name))
+        return 0;
     return 1;
 };
 /** Every IPv4 address a phone on the same network could dial, most reachable
@@ -33,12 +47,19 @@ const interfaceRank = (name) => {
  * will reach us.
  *
  * Ranked, not merely collected: `networkInterfaces()` promises nothing about
- * order, callers put the first non-tailnet entry into the pairing QR, and on
- * a Mac with a VPN or a VM running the first entry can be a utun or bridge100
- * address the phone cannot route to. Real interfaces lead, tunnels and
- * bridges trail; the sort is stable, so enumeration order still breaks ties.
+ * order, callers put the first non-tailnet entry into the pairing QR, and the
+ * first entry can be an address the phone cannot route to — a utun or
+ * bridge100 on a Mac, WSL's vEthernet on Windows, docker0 on Linux. Real
+ * interfaces lead, virtual ones trail; the sort is stable, so enumeration
+ * order still breaks ties.
  * The parameter exists for tests — the interface table is the machine's. */
 export function lanAddresses(interfaces = networkInterfaces()) {
+    return lanInterfaces(interfaces).map((entry) => entry.address);
+}
+/** `lanAddresses`, each with the interface it is on, in the same order. The
+ * name is what Windows calls the adapter ("Wi-Fi"), which is how the pairing
+ * panel asks whether Windows has that network down as Public. */
+export function lanInterfaces(interfaces = networkInterfaces()) {
     const found = [];
     for (const [name, entries] of Object.entries(interfaces)) {
         for (const entry of entries ?? []) {
@@ -46,10 +67,10 @@ export function lanAddresses(interfaces = networkInterfaces()) {
                 continue;
             if (entry.address.startsWith("169.254."))
                 continue;
-            found.push({ rank: interfaceRank(name), address: entry.address });
+            found.push({ rank: interfaceRank(name), name, address: entry.address });
         }
     }
-    return found.sort((a, b) => a.rank - b.rank).map((entry) => entry.address);
+    return found.sort((a, b) => a.rank - b.rank).map(({ name, address }) => ({ name, address }));
 }
 /** Tailscale hands its nodes an address in 100.64.0.0/10 — the CGNAT range
  * RFC 6598 set aside, which is why it never collides with a home network.

package/dist-server/companion/src/routes.js

@@ -35,6 +35,22 @@ export const CLOUD_DESKTOP_CONTROL_ROUTE = {
     method: "POST",
     path: /^\/api\/bots\/[\w-]+\/computer\/(?:control|screenshot|viewer-close)$/,
 };
+/** Driving a bot's own browser: the frame stream, and the channel that sends
+ * pointer and keyboard events into it.
+ *
+ * Deliberately separate from the cloud-desktop classifier above rather than
+ * folded into it. A cloud desktop is a disposable VM; a bot's browser is
+ * normally signed into the person's real accounts, with their cookies and
+ * their sessions. A device trusted with a throwaway VM must not acquire the
+ * second permission because the two looked similar from here. */
+export const BROWSER_LIVE_ROUTE = {
+    method: "GET",
+    path: /^\/api\/bots\/[\w-]+\/browser\/live$/,
+};
+export const BROWSER_ACTION_ROUTE = {
+    method: "POST",
+    path: /^\/api\/bots\/[\w-]+\/browser\/action$/,
+};
 /** The Local VM's live desktop, relayed by the sidecar like a VPS viewer.
  * The harness grants it only while a person holds that bot's computer. */
 export const LOCAL_VM_JOIN_ROUTE = {
@@ -62,6 +78,13 @@ export function isCloudDesktopAccess(method, path) {
         || (method === LOCAL_VM_SCREENSHOT_ROUTE.method && LOCAL_VM_SCREENSHOT_ROUTE.path.test(path))
         || (method === LOCAL_VM_JOIN_ROUTE.method && LOCAL_VM_JOIN_ROUTE.path.test(path));
 }
+/** Both halves of browser control, behind one capability. Watching the frames
+ * and driving them are the same permission: a viewer that can see a logged-in
+ * session is already past the line the capability is drawn at. */
+export function isBrowserControlAccess(method, path) {
+    return (method === BROWSER_LIVE_ROUTE.method && BROWSER_LIVE_ROUTE.path.test(path))
+        || (method === BROWSER_ACTION_ROUTE.method && BROWSER_ACTION_ROUTE.path.test(path));
+}
 /** Every request the iOS app makes, and nothing else.
  *
  * Ids are `[\w-]+`, matching the harness's own route patterns. The paths
@@ -111,6 +134,17 @@ const ALLOWED = [
     // recent activity. No settings, no transcript — read on open and on
     // pull-to-refresh.
     { method: "GET", path: /^\/api\/bots\/[\w-]+\/overview$/ },
+    // Read-only: what the bot did, with the outcome, built from logs that
+    // already exist (server/activity.ts). No settings reachable through it.
+    { method: "GET", path: /^\/api\/bots\/[\w-]+\/activity$/ },
+    // The section's shared team memory: people, places, decisions and terms.
+    // Reading, adding, answering a proposal, editing a detail, and removing
+    // are all content edits, not execution policy — the same line the profile
+    // and sidebar-section routes draw. The section rides as a query string.
+    { method: "GET", path: /^\/api\/team-memory$/ },
+    { method: "POST", path: /^\/api\/team-memory$/ },
+    { method: "PATCH", path: /^\/api\/team-memory\/[\w-]+$/ },
+    { method: "DELETE", path: /^\/api\/team-memory\/[\w-]+$/ },
     // Paired-safe profile subset. The harness route itself rejects fields
     // outside identity, standing instructions (soul, byte-capped), avatar,
     // notifications, and voice preferences.
@@ -123,6 +157,11 @@ const ALLOWED = [
     // second, per-device capability check before it reaches the harness.
     CLOUD_DESKTOP_JOIN_ROUTE,
     CLOUD_DESKTOP_CONTROL_ROUTE,
+    // A bot's browser, watched and driven from the phone. Like the cloud
+    // desktop above, the proxy applies a second per-device capability check
+    // before either of these reaches the harness.
+    BROWSER_LIVE_ROUTE,
+    BROWSER_ACTION_ROUTE,
     // A picture of the Local VM — not its lifecycle, which stays on the host.
     // Gated per device by the proxy like the cloud desktop above.
     LOCAL_VM_SCREENSHOT_ROUTE,

package/dist-server/container-mcp.js

@@ -15816,10 +15816,6 @@ var remoteMutationSchema = remoteEntrySchema.extend({
   }).optional()
 });
 
-// server/cloud-overflow.ts
-var CLOUD_SEAT_IDLE_STOP_MS = 5 * 6e4;
-var CLOUD_SEAT_SLEEP_RETRY_MAX_MS = 10 * 6e4;
-
 // server/config.ts
 var optionalText = external_exports.string().optional();
 var SSH_ALIAS = /^[A-Za-z0-9][A-Za-z0-9_.-]{0,127}$/;
@@ -16017,17 +16013,6 @@ var featureConfigSchema = external_exports.object({
    * conversations that share words with the message ride into the turn.
    * Read-only; on unless explicitly switched off — see autoRecallEnabled. */
   autoRecall: external_exports.boolean().optional(),
-  /** Idle release for computer claims (#1653): a desktop seat that stays
-   * screen-quiet for 90 seconds is released to waiting turns while its
-   * holder's turn still lives; the previous holder re-claims directly
-   * for 10 minutes and yields to an occupied seat. Off until baked; see
-   * computerClaimIdleReleaseEnabled for how to enable it by hand. */
-  computerClaimIdleRelease: external_exports.boolean().optional(),
-  /** Consented cloud overflow for local computer waits (#1655): a wait
-   * may offer a per-second-billed cloud seat, but only through an explicit
-   * consent card or a configured allowlist — never a silent Auto default.
-   * Off until baked; see cloudOverflowEnabled for how to enable it by hand. */
-  cloudOverflow: external_exports.boolean().optional(),
   /** Opt-in shared skills library: one store at the data dir that bots
    * read by assignment instead of per-workspace copies. Off until
    * explicitly enabled — see skillsLibraryEnabled. */
@@ -16153,8 +16138,12 @@ var appConfigSchema = external_exports.object({
    * provider is Boat now and the key is kept for compatibility. */
   box: external_exports.object({ token: optionalText }).optional(),
   vps: vpsConfigSchema.optional(),
-  /** Optional OpenCode key; persisted write-only and passed only to its child. */
-  opencodeGo: external_exports.object({ apiKey: optionalText }).optional(),
+  /** Optional OpenCode key; persisted write-only and passed only to its child.
+   * `providerKeys`: keys for OpenCode's other providers (Venice, Groq…) by the
+   * environment name OpenCode reads, also write-only and only for its child.
+   * Settings changes go through mergeOpenCodeProviderKeys; the stored copy
+   * is read loosely (storedAppConfigSchema). */
+  opencodeGo: external_exports.object({ apiKey: optionalText, providerKeys: external_exports.record(external_exports.string(), external_exports.string()).optional() }).optional(),
   /** Voice settings and the selected voice id. `provider` picks the
    * engine: "elevenlabs" (default; needs `key`), "fish" (needs its own
    * `fishKey`; `fishModel` picks its speech model), "system" (the Mac's
@@ -16234,15 +16223,6 @@ var appConfigSchema = external_exports.object({
   /** The authorization decision log (server/decision-log.ts): days of month
    * files kept, at least; OMB_DECISION_RETENTION_DAYS wins when set. */
   decisions: external_exports.object({ retentionDays: external_exports.number().int().min(1).max(3650).optional() }).strict().optional(),
-  /** #1655 cloud-overflow settings. perSecondCostUsd is the operator's own
-   * verified rate: with no price configured the feature stays inert rather
-   * than show an invented one. allowlistedThreads carries standing consent
-   * for exact thread ids. */
-  cloudOverflow: external_exports.object({
-    perSecondCostUsd: external_exports.number().positive().max(10).optional(),
-    idleStopMs: external_exports.number().int().positive().max(24 * 60 * 6e4).optional(),
-    allowlistedThreads: external_exports.array(external_exports.string().min(1)).max(1e3).optional()
-  }).strict().optional(),
   localVm: localVmConfigSchema.optional(),
   features: featureConfigSchema.optional(),
   onboarding: onboardingConfigSchema.optional(),
@@ -16257,7 +16237,14 @@ var appConfigSchema = external_exports.object({
   mcpServers: external_exports.record(external_exports.string(), external_exports.unknown()).optional()
 });
 var storedAppConfigSchema = appConfigSchema.extend({
-  browserProfiles: storedBrowserProfilesSchema.optional()
+  browserProfiles: storedBrowserProfilesSchema.optional(),
+  /** Read loosely, like mcpServers: a hand-edited provider key that isn't
+   * text is dropped here, and a bad name or key later
+   * (openCodeProviderKeys), never the whole file. */
+  opencodeGo: external_exports.object({
+    apiKey: optionalText,
+    providerKeys: external_exports.record(external_exports.string(), external_exports.unknown()).catch({}).transform((keys) => Object.fromEntries(Object.entries(keys).filter((entry) => typeof entry[1] === "string"))).optional()
+  }).optional()
 });
 var appConfigPatchSchema = appConfigSchema.omit({ instances: true, mcpServers: true, cliStartup: true, customDomain: true }).extend({ threads: threadsPatchSchema.optional(), newBots: newBotsPatchSchema.optional() });
 var jsonObjectSchema = external_exports.record(external_exports.string(), external_exports.json());
@@ -16305,7 +16292,8 @@ function resolveProxy(relative) {
   return existsSync2(source) ? source : join3(SERVER_ROOT, `${relative}.js`);
 }
 var SPAWNED_PROXIES = {
-  browser: resolveProxy("browser-proxy"),
+  // The built-in browser and the cloud computer: argv[2] picks which.
+  harnessMcp: resolveProxy("harness-mcp-proxy"),
   localComputer: resolveProxy("local-computer-proxy"),
   permission: resolveProxy("permission-proxy"),
   containerMcp: resolveProxy("container-mcp"),
@@ -16337,7 +16325,6 @@ var VM_WORKSPACE_DIR = join4(DATA_DIR, "vm-home");
 var DISPLAY = ":1";
 var CUA_EXECUTABLE = "/usr/local/libexec/openmausbot/cua-driver";
 var MEMORY_BYTES = 4 * 1024 * 1024 * 1024;
-var SHM_BYTES = 512 * 1024 * 1024;
 function cuaExecArgs(args, options = {}) {
   return [
     "exec",