mixdog 1.0.8 → 1.0.10

package/src/runtime/computer-bridge/client.mjs

@@ -13,13 +13,20 @@ import {
   toComputerHostCommand,
   validateComputerToolArgs,
 } from './action-schema.mjs';
-import { computerResultRecovery, formatComputerToolError } from './error-recovery.mjs';
+import { formatComputerToolError } from './error-recovery.mjs';
+import {
+  canonicalComputerResultIsError,
+  canonicalComputerResultText,
+  computerImageDimensions,
+} from './result-canonical.mjs';
 import { computerErrorCode } from './error-code.mjs';
 import { bridgeDiscoveryChanged, readBridgeDiscovery } from '../bridge-discovery.mjs';
 import { MAX_COMPUTER_REQUEST_BYTES, readComputerBridgeJson, validateComputerReply } from './limits.mjs';
 import { computerActionHas } from './actions.mjs';
 import { continuePendingComputerWork, isPendingComputerWork } from './pending-continuation.mjs';
 
+export { canonicalComputerResultText };
+
 const DISCOVERY_FILE = 'computer-bridge.json';
 // Desktop UI Automation queries and input dispatch can be slow; sit above the
 // bridge's own per-action timeouts so its specific error wins over a bare abort.
@@ -48,186 +55,6 @@ const activeComputerExecutions = new Set();
 const BRIDGE_UNAVAILABLE_MESSAGE =
   'computer use is unavailable; open the Mixdog desktop app and enable Computer Use in settings';
 
-const ACT_STEP_STATUSES = new Set(['succeeded', 'failed', 'skipped', 'pending', 'uncertain']);
-
-function canonicalizeActResult(value, args) {
-  value.completed_actions = value.completed_steps;
-  value.total_actions = value.total_steps;
-  const canonicalStep = (row, index) => {
-    const normalized = { ...row };
-    normalized.type = args.input?.actions?.[index]?.type || normalized.action;
-    if (!ACT_STEP_STATUSES.has(normalized.status)) {
-      normalized.status = normalized.ok === false ? 'failed' : 'succeeded';
-    }
-    delete normalized.action;
-    delete normalized.ok;
-    return normalized;
-  };
-  value.actions = Array.isArray(value.steps) ? value.steps.map(canonicalStep) : value.steps;
-  // A single-window act reports the same transition on the act and its step.
-  const transition = JSON.stringify(value.window_transition);
-  for (const row of Array.isArray(value.actions) ? value.actions : []) {
-    if (transition && JSON.stringify(row?.window_transition) === transition) delete row.window_transition;
-  }
-  delete value.completed_steps;
-  delete value.total_steps;
-  delete value.steps;
-}
-
-// Host diagnostics stay in the host's run record; the model acts on none of them.
-const HOST_DIAGNOSTIC_FIELDS = ['timings_ms', 'capture_attempts'];
-
-function dropHostDiagnostics(value) {
-  if (!value || typeof value !== 'object') return;
-  if (Array.isArray(value)) {
-    for (const item of value) dropHostDiagnostics(item);
-    return;
-  }
-  for (const field of HOST_DIAGNOSTIC_FIELDS) delete value[field];
-  for (const item of Object.values(value)) dropHostDiagnostics(item);
-}
-
-// Elements and OCR text reach the model one line each, the shape browser
-// snapshots use, because a field name costs more than the value it labels:
-// `#mark [ref] Role "name" value="…" state=… source=… disabled focused
-// @x,y,width,height actions`. Flags appear only when they differ from the
-// common case the capture description states (an enabled UIA element), and
-// the geometry a som frame repeats as x/y/width/height, center and
-// screen_bounds collapses into its one bounds box.
-const DERIVED_GEOMETRY_FIELDS = new Set([
-  'x',
-  'y',
-  'width',
-  'height',
-  'center_x',
-  'center_y',
-  'center',
-  'screen_bounds',
-]);
-
-const quotedText = (text) => JSON.stringify(String(text));
-const bareText = (text) => (/^[^\s"]+$/.test(String(text)) ? String(text) : quotedText(text));
-const isUnsetField = (field) =>
-  field === undefined || field === null || field === '' || field === false || (Array.isArray(field) && !field.length);
-
-function boundsText(box) {
-  return box.length === 4 && box.every((edge) => Number.isFinite(edge)) ? `@${box.join(',')}` : null;
-}
-
-function extraFieldTexts(rest) {
-  return Object.entries(rest)
-    .filter(([key, field]) => !DERIVED_GEOMETRY_FIELDS.has(key) && !isUnsetField(field))
-    .map(([key, field]) => `${key}=${typeof field === 'string' ? bareText(field) : JSON.stringify(field)}`);
-}
-
-function elementLine(element) {
-  if (!element || typeof element !== 'object') return element;
-  const { mark, ref, role, name, value, state, source, enabled, has_keyboard_focus, bounds, actions, ...rest } =
-    element;
-  return [
-    isUnsetField(mark) ? null : `#${mark}`,
-    isUnsetField(ref) ? null : `[${ref}]`,
-    bareText(role || 'Unknown'),
-    quotedText(name ?? ''),
-    isUnsetField(value) ? null : `value=${quotedText(value)}`,
-    isUnsetField(state) || state === source ? null : `state=${bareText(state)}`,
-    isUnsetField(source) || source === 'uia' ? null : `source=${bareText(source)}`,
-    enabled === false ? 'disabled' : null,
-    has_keyboard_focus === true ? 'focused' : null,
-    boundsText(Array.isArray(bounds) ? bounds : [rest.x, rest.y, rest.width, rest.height]),
-    Array.isArray(actions) && actions.length ? actions.join(',') : null,
-    ...extraFieldTexts(rest),
-  ]
-    .filter(Boolean)
-    .join(' ');
-}
-
-function ocrTextLine(row) {
-  if (!row || typeof row !== 'object') return row;
-  const { mark, text, line, x, y, width, height, ...rest } = row;
-  return [
-    isUnsetField(mark) ? null : `#${mark}`,
-    quotedText(text ?? ''),
-    boundsText([x, y, width, height]),
-    isUnsetField(line) ? null : `line=${line}`,
-    ...extraFieldTexts(rest),
-  ]
-    .filter(Boolean)
-    .join(' ');
-}
-
-function renderObservationLines(frame) {
-  if (!frame || typeof frame !== 'object') return;
-  if (Array.isArray(frame.elements)) frame.elements = frame.elements.map(elementLine);
-  for (const key of ['lines', 'words']) {
-    if (Array.isArray(frame.ocr?.[key])) frame.ocr[key] = frame.ocr[key].map(ocrTextLine);
-  }
-}
-
-export function canonicalComputerResultText(text, args) {
-  if (args.action === 'clipboard' && args.input?.operation === 'read') return text;
-  let value;
-  try {
-    value = JSON.parse(text);
-  } catch {
-    return text;
-  }
-  if (!value || typeof value !== 'object' || Array.isArray(value) || typeof value.action !== 'string') {
-    return text;
-  }
-  value.action = args.action;
-  if (args.action === 'list') value.kind = args.input?.kind;
-  if (args.action === 'capture') value.mode = args.input?.mode || 'state';
-  if (args.action === 'window') value.operation = args.input?.operation;
-  if (args.action === 'clipboard') value.operation = args.input?.operation;
-  if (args.action === 'act') canonicalizeActResult(value, args);
-  if (value.capture_after && value.observation === undefined) {
-    value.observation = value.capture_after;
-    delete value.capture_after;
-  }
-  dropHostDiagnostics(value);
-  renderObservationLines(value);
-  renderObservationLines(value.observation);
-  if (value.ok === false && value.recovery === undefined) {
-    const recovery = computerResultRecovery(value, args);
-    if (recovery) value.recovery = recovery;
-  }
-  return JSON.stringify(value);
-}
-
-// A capture payload reports the frame's real pixel size (`width`/`height`,
-// nested under `observation` for action replies). Carrying it on the image
-// block lets the context estimator bill a screenshot at its true vision cost
-// instead of the flat unknown-image allowance. Provider normalizers rebuild
-// the wire block from `type`/`source` alone, so these fields never reach an
-// API and never shift the cached prefix.
-function computerImageDimensions(text) {
-  let value;
-  try {
-    value = JSON.parse(text);
-  } catch {
-    return null;
-  }
-  for (const frame of [value, value?.observation]) {
-    const width = Number(frame?.width);
-    const height = Number(frame?.height);
-    if (width > 0 && height > 0 && Number.isFinite(width) && Number.isFinite(height)) return { width, height };
-  }
-  return null;
-}
-
-function canonicalComputerResultIsError(text, args) {
-  if (args.action === 'clipboard' && args.input?.operation === 'read') return false;
-  try {
-    const value = JSON.parse(text);
-    return Boolean(
-      value && typeof value === 'object' && !Array.isArray(value) && value.action === args.action && value.ok === false
-    );
-  } catch {
-    return false;
-  }
-}
-
 /** Sync gate for the session tool surface (featureDisallowedTools). */
 export function computerBridgeAvailableSync() {
   return readBridgeDiscovery(DISCOVERY_FILE) !== null;
@@ -541,8 +368,8 @@ async function abortComputerSession(sessionId) {
   return aborted;
 }
 
-/** End the visible Computer Use execution at agent-turn settlement without
- * dropping the warm worker and observation refs kept for a possible follow-up. */
+/** End the visible Computer Use execution at agent-turn settlement and hand the
+ * desktop back, keeping the warm worker and observation refs for a possible follow-up. */
 export async function endComputerExecution(sessionId) {
   const id = String(sessionId || '').trim();
   if (!id || !activeComputerExecutions.has(id)) return false;

package/src/runtime/computer-bridge/core-actions.mjs

@@ -276,7 +276,8 @@ function actionArgumentsError(action, type, label, frameId) {
     if (hasOwn(action, 'ref') && hasOwn(action, 'element')) return `${label} accepts only one of ref or element`;
     return typeof action.value === 'string' ? null : `${label} requires value`;
   }
-  if (type === 'key' || type === 'key_down' || type === 'key_up') return typeof action.keys === 'string' ? null : `${label} requires keys`;
+  if (type === 'key' || type === 'key_down' || type === 'key_up')
+    return typeof action.keys === 'string' ? null : `${label} requires keys`;
   if (type === 'wait') {
     const { duration } = action;
     const valid = typeof duration === 'number' && Number.isFinite(duration) && duration >= 0 && duration <= 5;

package/src/runtime/computer-bridge/error-recovery.mjs

@@ -21,11 +21,7 @@ function targetLabel(args) {
 const LIST_WINDOW_CODES = new Set(['ambiguous_window_target', 'window_stale', 'window_target_not_found']);
 const STALE_TARGET_CODES = new Set(['stale_frame', 'stale_target', 'target_mismatch']);
 const USER_YIELD_CODES = new Set(['computer_user_control_active', 'computer_user_takeover', 'user_input_active']);
-const DIAGNOSE_CODES = new Set([
-  'input_observation_unavailable',
-  'input_recovery_unconfirmed',
-  'computer_cursor_unavailable',
-]);
+const DIAGNOSE_CODES = new Set(['input_observation_unavailable', 'computer_cursor_unavailable']);
 const CLEANUP_CODES = new Set([
   'computer_cleanup_pending',
   'computer_abort_cleanup_unconfirmed',
@@ -47,6 +43,40 @@ function recaptureLeaseGuidance(code, target) {
   return `The target lease is now available. Capture ${target} again before issuing any input.`;
 }
 
+function timeoutRecovery(code, args, target) {
+  if (READ_ONLY_ACTIONS.has(String(args?.action || ''))) {
+    return {
+      code,
+      next: 'capture',
+      guidance: `The read exceeded its bounded budget and sent no input, so no cleanup or user-control guard applies. Capture ${target} with mode="state" or "som": pixels and OCR stay usable while that window's accessibility provider is slow.`,
+    };
+  }
+  return {
+    code,
+    next: 'diagnose',
+    guidance: `The command may have executed before timing out. Do not repeat it or switch delivery modes. Diagnose the host first; cleanup and user-control guards must clear through verified recovery. Then capture ${target} and inspect the effect before issuing any new input.`,
+  };
+}
+
+function foregroundRecovery(code, target) {
+  if (code === 'foreground_changed') {
+    return {
+      code,
+      next: 'user',
+      guidance:
+        'Focus changed during dispatch. Do not pull it back or retry input automatically. Check user control, then obtain fresh state when control is available.',
+    };
+  }
+  if (code === 'foreground_unavailable') {
+    return {
+      code,
+      next: 'user',
+      guidance: `The system did not grant foreground focus (macOS refuses it while another app is active). Ask the user to activate ${target}, or use a foreground click on the window, which activates it; then capture fresh state. Do not substitute background input or repeat the failed gesture.`,
+    };
+  }
+  return undefined;
+}
+
 function recoveryForCode(code, args) {
   const target = targetLabel(args);
   if (code === 'computer_background_cleanup_unconfirmed') {
@@ -82,6 +112,14 @@ function recoveryForCode(code, args) {
         "Computer Use yielded to the user. Call wait_for_user for bounded waiting. The user's Resume on the overlay continues the pause, and ordinary physical input may also resume after the host-configured quiet interval (default 5 seconds); explicit stops and uncertain cleanup/observation require the user. Stop cancels the task. Timeout does not authorize input. After resumed, capture fresh state; never replay interrupted input.",
     };
   }
+  if (code === 'input_recovery_unconfirmed') {
+    return {
+      code,
+      next: 'wait_for_user',
+      guidance:
+        'The desktop state after the action could not be verified, so Computer Use stopped the session and handed the desktop to the user. Do not repeat the mutation. Call wait_for_user: the host resumes by itself once the desktop has been quiet for its interval, or when the user presses Resume. After resumed, capture fresh state and inspect the effect before any new input.',
+    };
+  }
   if (DIAGNOSE_CODES.has(code)) {
     return {
       code,
@@ -105,20 +143,7 @@ function recoveryForCode(code, args) {
       guidance: `No input was sent. Foreground delivery needs an observation taken while the user's own input was idle. Let their input settle and capture ${target} again, or send this action through background delivery when it supports one.`,
     };
   }
-  if (code === 'computer_command_timeout') {
-    if (READ_ONLY_ACTIONS.has(String(args?.action || ''))) {
-      return {
-        code,
-        next: 'capture',
-        guidance: `The read exceeded its bounded budget and sent no input, so no cleanup or user-control guard applies. Capture ${target} with mode="state" or "som": pixels and OCR stay usable while that window's accessibility provider is slow.`,
-      };
-    }
-    return {
-      code,
-      next: 'diagnose',
-      guidance: `The command may have executed before timing out. Do not repeat it or switch delivery modes. Diagnose the host first; cleanup and user-control guards must clear through verified recovery. Then capture ${target} and inspect the effect before issuing any new input.`,
-    };
-  }
+  if (code === 'computer_command_timeout') return timeoutRecovery(code, args, target);
   if (code.startsWith('menu_')) {
     return {
       code,
@@ -126,21 +151,8 @@ function recoveryForCode(code, args) {
       guidance: `Capture ${target} again; empty accessibility automatically uses OCR. Use a fresh OCR mark or frame point and do not retry the same menu path unchanged.`,
     };
   }
-  if (code === 'foreground_changed') {
-    return {
-      code,
-      next: 'user',
-      guidance:
-        'Focus changed during dispatch. Do not pull it back or retry input automatically. Check user control, then obtain fresh state when control is available.',
-    };
-  }
-  if (code === 'foreground_unavailable') {
-    return {
-      code,
-      next: 'user',
-      guidance: `Windows did not grant foreground focus. Ask the user to activate ${target}, then capture fresh state. Do not substitute background input or repeat the failed gesture.`,
-    };
-  }
+  const foreground = foregroundRecovery(code, target);
+  if (foreground) return foreground;
   if (code.startsWith('background_')) {
     return {
       code,

package/src/runtime/computer-bridge/result-canonical.mjs

@@ -0,0 +1,181 @@
+import { computerResultRecovery } from './error-recovery.mjs';
+
+const ACT_STEP_STATUSES = new Set(['succeeded', 'failed', 'skipped', 'pending', 'uncertain']);
+
+function canonicalizeActResult(value, args) {
+  value.completed_actions = value.completed_steps;
+  value.total_actions = value.total_steps;
+  const canonicalStep = (row, index) => {
+    const normalized = { ...row };
+    normalized.type = args.input?.actions?.[index]?.type || normalized.action;
+    if (!ACT_STEP_STATUSES.has(normalized.status)) {
+      normalized.status = normalized.ok === false ? 'failed' : 'succeeded';
+    }
+    delete normalized.action;
+    delete normalized.ok;
+    return normalized;
+  };
+  value.actions = Array.isArray(value.steps) ? value.steps.map(canonicalStep) : value.steps;
+  // A single-window act reports the same transition on the act and its step.
+  const transition = JSON.stringify(value.window_transition);
+  for (const row of Array.isArray(value.actions) ? value.actions : []) {
+    if (transition && JSON.stringify(row?.window_transition) === transition) delete row.window_transition;
+  }
+  delete value.completed_steps;
+  delete value.total_steps;
+  delete value.steps;
+}
+
+// Host diagnostics stay in the host's run record; the model acts on none of them.
+const HOST_DIAGNOSTIC_FIELDS = ['timings_ms', 'capture_attempts'];
+
+function dropHostDiagnostics(value) {
+  if (!value || typeof value !== 'object') return;
+  if (Array.isArray(value)) {
+    for (const item of value) dropHostDiagnostics(item);
+    return;
+  }
+  for (const field of HOST_DIAGNOSTIC_FIELDS) delete value[field];
+  for (const item of Object.values(value)) dropHostDiagnostics(item);
+}
+
+// Elements and OCR text reach the model one line each, the shape browser
+// snapshots use, because a field name costs more than the value it labels:
+// `#mark [ref] Role "name" value="…" state=… source=… disabled focused
+// @x,y,width,height actions`. Flags appear only when they differ from the
+// common case the capture description states (an enabled UIA element), and
+// the geometry a som frame repeats as x/y/width/height, center and
+// screen_bounds collapses into its one bounds box.
+const DERIVED_GEOMETRY_FIELDS = new Set([
+  'x',
+  'y',
+  'width',
+  'height',
+  'center_x',
+  'center_y',
+  'center',
+  'screen_bounds',
+]);
+
+const quotedText = (text) => JSON.stringify(String(text));
+const bareText = (text) => (/^[^\s"]+$/.test(String(text)) ? String(text) : quotedText(text));
+const isUnsetField = (field) =>
+  field === undefined || field === null || field === '' || field === false || (Array.isArray(field) && !field.length);
+
+function boundsText(box) {
+  return box.length === 4 && box.every((edge) => Number.isFinite(edge)) ? `@${box.join(',')}` : null;
+}
+
+function extraFieldTexts(rest) {
+  return Object.entries(rest)
+    .filter(([key, field]) => !DERIVED_GEOMETRY_FIELDS.has(key) && !isUnsetField(field))
+    .map(([key, field]) => `${key}=${typeof field === 'string' ? bareText(field) : JSON.stringify(field)}`);
+}
+
+function elementLine(element) {
+  if (!element || typeof element !== 'object') return element;
+  const { mark, ref, role, name, value, state, source, enabled, has_keyboard_focus, bounds, actions, ...rest } =
+    element;
+  return [
+    isUnsetField(mark) ? null : `#${mark}`,
+    isUnsetField(ref) ? null : `[${ref}]`,
+    bareText(role || 'Unknown'),
+    quotedText(name ?? ''),
+    isUnsetField(value) ? null : `value=${quotedText(value)}`,
+    isUnsetField(state) || state === source ? null : `state=${bareText(state)}`,
+    isUnsetField(source) || source === 'uia' ? null : `source=${bareText(source)}`,
+    enabled === false ? 'disabled' : null,
+    has_keyboard_focus === true ? 'focused' : null,
+    boundsText(Array.isArray(bounds) ? bounds : [rest.x, rest.y, rest.width, rest.height]),
+    Array.isArray(actions) && actions.length ? actions.join(',') : null,
+    ...extraFieldTexts(rest),
+  ]
+    .filter(Boolean)
+    .join(' ');
+}
+
+function ocrTextLine(row) {
+  if (!row || typeof row !== 'object') return row;
+  const { mark, text, line, x, y, width, height, ...rest } = row;
+  return [
+    isUnsetField(mark) ? null : `#${mark}`,
+    quotedText(text ?? ''),
+    boundsText([x, y, width, height]),
+    isUnsetField(line) ? null : `line=${line}`,
+    ...extraFieldTexts(rest),
+  ]
+    .filter(Boolean)
+    .join(' ');
+}
+
+function renderObservationLines(frame) {
+  if (!frame || typeof frame !== 'object') return;
+  if (Array.isArray(frame.elements)) frame.elements = frame.elements.map(elementLine);
+  for (const key of ['lines', 'words']) {
+    if (Array.isArray(frame.ocr?.[key])) frame.ocr[key] = frame.ocr[key].map(ocrTextLine);
+  }
+}
+
+export function canonicalComputerResultText(text, args) {
+  if (args.action === 'clipboard' && args.input?.operation === 'read') return text;
+  let value;
+  try {
+    value = JSON.parse(text);
+  } catch {
+    return text;
+  }
+  if (!value || typeof value !== 'object' || Array.isArray(value) || typeof value.action !== 'string') {
+    return text;
+  }
+  value.action = args.action;
+  if (args.action === 'list') value.kind = args.input?.kind;
+  if (args.action === 'capture') value.mode = args.input?.mode || 'state';
+  if (args.action === 'window') value.operation = args.input?.operation;
+  if (args.action === 'clipboard') value.operation = args.input?.operation;
+  if (args.action === 'act') canonicalizeActResult(value, args);
+  if (value.capture_after && value.observation === undefined) {
+    value.observation = value.capture_after;
+    delete value.capture_after;
+  }
+  dropHostDiagnostics(value);
+  renderObservationLines(value);
+  renderObservationLines(value.observation);
+  if (value.ok === false && value.recovery === undefined) {
+    const recovery = computerResultRecovery(value, args);
+    if (recovery) value.recovery = recovery;
+  }
+  return JSON.stringify(value);
+}
+
+// A capture payload reports the frame's real pixel size (`width`/`height`,
+// nested under `observation` for action replies). Carrying it on the image
+// block lets the context estimator bill a screenshot at its true vision cost
+// instead of the flat unknown-image allowance. Provider normalizers rebuild
+// the wire block from `type`/`source` alone, so these fields never reach an
+// API and never shift the cached prefix.
+export function computerImageDimensions(text) {
+  let value;
+  try {
+    value = JSON.parse(text);
+  } catch {
+    return null;
+  }
+  for (const frame of [value, value?.observation]) {
+    const width = Number(frame?.width);
+    const height = Number(frame?.height);
+    if (width > 0 && height > 0 && Number.isFinite(width) && Number.isFinite(height)) return { width, height };
+  }
+  return null;
+}
+
+export function canonicalComputerResultIsError(text, args) {
+  if (args.action === 'clipboard' && args.input?.operation === 'read') return false;
+  try {
+    const value = JSON.parse(text);
+    return Boolean(
+      value && typeof value === 'object' && !Array.isArray(value) && value.action === args.action && value.ok === false
+    );
+  } catch {
+    return false;
+  }
+}

package/src/runtime/local-provider/asset-installer.mjs

@@ -143,18 +143,26 @@ export async function downloadVerifiedLocalAsset(
   return destination;
 }
 
-async function extractZip(zipPath, destination, signal) {
-  signal?.throwIfAborted();
+// The Windows builds ship as zips (PowerShell's Expand-Archive); the macOS
+// build is a gzipped tarball, which the system tar unpacks.
+function extractCommand(archive, destination) {
+  if (!archive.endsWith('.zip')) return ['tar', ['-xzf', archive, '-C', destination], process.env];
   const command = [
     "$ErrorActionPreference = 'Stop'",
     'Expand-Archive -LiteralPath $env:MIXDOG_LOCAL_ARCHIVE -DestinationPath $env:MIXDOG_LOCAL_DESTINATION -Force',
   ].join('; ');
-  const child = spawn('powershell', ['-NoProfile', '-NonInteractive', '-Command', command], {
-    env: {
-      ...process.env,
-      MIXDOG_LOCAL_ARCHIVE: zipPath,
-      MIXDOG_LOCAL_DESTINATION: destination,
-    },
+  return [
+    'powershell',
+    ['-NoProfile', '-NonInteractive', '-Command', command],
+    { ...process.env, MIXDOG_LOCAL_ARCHIVE: archive, MIXDOG_LOCAL_DESTINATION: destination },
+  ];
+}
+
+async function extractArchive(archive, destination, signal) {
+  signal?.throwIfAborted();
+  const [command, args, env] = extractCommand(archive, destination);
+  const child = spawn(command, args, {
+    env,
     windowsHide: true,
     signal,
     stdio: ['ignore', 'pipe', 'pipe'],
@@ -184,7 +192,9 @@ async function installRuntimeInternal({ dataDir, onProgress, fetchFn, signal })
   signal?.throwIfAborted();
   const entry = localProviderRuntimePlatformEntry();
   if (!entry || !localProviderCatalogStatus({ dataDir, hardware }).available) {
-    throw new Error('[local-provider] Windows x64 with a compatible NVIDIA RTX GPU is required');
+    throw new Error(
+      '[local-provider] Windows x64 with a compatible NVIDIA RTX GPU, or an Apple Silicon Mac with enough memory, is required'
+    );
   }
   const target = localProviderRuntimeDirectory(dataDir);
   const executable = join(target, entry.executable);
@@ -217,7 +227,7 @@ async function installRuntimeInternal({ dataDir, onProgress, fetchFn, signal })
         totalBytes: entry.downloadBytes,
         percent: Math.min(99, Math.round(((completedBytes + asset.size) / entry.downloadBytes) * 100)),
       });
-      await extractZip(archive, staging, signal);
+      await extractArchive(archive, staging, signal);
       completedBytes += asset.size;
     }
     if (!existsSync(join(staging, entry.executable))) {

package/src/runtime/local-provider/catalog.mjs

@@ -4,7 +4,7 @@ import { fileURLToPath } from 'node:url';
 
 import { resolvePluginData } from '../shared/plugin-paths.mjs';
 import { localProviderDiskStatus, partialAssetBytes } from './asset-storage.mjs';
-import { localProviderHardwareStatus } from './hardware.mjs';
+import { localProviderHardwareStatus, localProviderPlatformKey } from './hardware.mjs';
 import { registeredLocalModels } from './registered-models.mjs';
 import { localModelState } from './model-state.mjs';
 import { localContextSettings } from './context-settings.mjs';
@@ -24,12 +24,8 @@ export function localProviderModelRoot(dataDir = resolvePluginData()) {
   return join(dataDir, 'local-provider', 'models');
 }
 
-function localProviderRuntimePlatformKey(platform = process.platform, arch = process.arch) {
-  return platform === 'win32' && arch === 'x64' ? 'win32-x64-nvidia' : '';
-}
-
 export function localProviderRuntimePlatformEntry(options = {}) {
-  const key = localProviderRuntimePlatformKey(options.platform, options.arch);
+  const key = localProviderPlatformKey(options.platform, options.arch);
   const entry = key ? LOCAL_PROVIDER_MANIFEST.runtime?.platforms?.[key] : null;
   return entry ? { key, ...entry } : null;
 }
@@ -109,7 +105,8 @@ export function localProviderCatalogStatus({
   const executable = localProviderRuntimeExecutable(dataDir, { platform, arch });
   const models = localProviderModelEntries(dataDir).map((entry) => publicModel(entry, dataDir, hardware));
   return {
-    available: Boolean(platformEntry && hardware?.gpu?.vendor === 'NVIDIA' && models.some((entry) => entry.compatible)),
+    platformSupported: Boolean(platformEntry),
+    available: Boolean(platformEntry && hardware?.supported && models.some((entry) => entry.compatible)),
     runtime: {
       installed: Boolean(executable && existsSync(executable)),
       version: LOCAL_PROVIDER_MANIFEST.runtime.version,

package/src/runtime/local-provider/data/manifest.json

@@ -23,6 +23,19 @@
             "sha256": "8c79a9b226de4b3cacfd1f83d24f962d0773be79f1e7b75c6af4ded7e32ae1d6"
           }
         ]
+      },
+      "darwin-arm64-metal": {
+        "executable": "llama-b10621/llama-server",
+        "backend": "Metal",
+        "downloadBytes": 10954823,
+        "assets": [
+          {
+            "name": "llama-b10621-bin-macos-arm64.tar.gz",
+            "url": "https://github.com/ggml-org/llama.cpp/releases/download/b10621/llama-b10621-bin-macos-arm64.tar.gz",
+            "size": 10954823,
+            "sha256": "429c8270608600188035e5e92f7d78dffb7900904fe7dd7e6a84f48068cd13cf"
+          }
+        ]
       }
     }
   },
@@ -30,7 +43,7 @@
     {
       "id": "qwen3.8-27b-q4-k-m",
       "name": "Qwen3.8 27B Q4_K_M",
-      "description": "Recommended coding and tool-use model for NVIDIA RTX GPUs with 24 GB VRAM.",
+      "description": "Recommended coding and tool-use model for GPUs with 24 GB of memory.",
       "filename": "Qwen3.8-27B-Q4_K_M.gguf",
       "url": "https://huggingface.co/ggml-org/Qwen3.8-27B-GGUF/resolve/0669b98607d47046c7c2b3f801011d54a08cfccf/Qwen3.8-27B-Q4_K_M.gguf?download=true",
       "revision": "0669b98607d47046c7c2b3f801011d54a08cfccf",