@remcp/runtime 0.2.71 → 0.2.73

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@remcp/runtime",
-  "version": "0.2.71",
+  "version": "0.2.73",
   "description": "First-party ReMCP local device runtime: filesystem, terminal, computer-use, browser, diagnostics and document tools over MCP.",
   "type": "module",
   "license": "MIT",

package/src/extended/browser.mjs

@@ -470,15 +470,40 @@ function findExpression(args) {
       }
       return parts.join(' > ');
     }
+    function semanticRole(el) {
+      const explicit = (el.getAttribute('role') || '').trim().toLowerCase();
+      if (explicit) return explicit;
+      const tag = el.tagName.toLowerCase();
+      if (tag === 'button') return 'button';
+      if (tag === 'a' && el.hasAttribute('href')) return 'link';
+      if (tag === 'textarea') return 'textbox';
+      if (tag === 'select') return el.multiple || el.size > 1 ? 'listbox' : 'combobox';
+      if (tag === 'option') return 'option';
+      if (tag === 'img') return 'img';
+      if (tag === 'progress') return 'progressbar';
+      if (tag === 'meter') return 'meter';
+      if (tag === 'summary') return 'button';
+      if (tag === 'input') {
+        const type = String(el.type || 'text').toLowerCase();
+        if (['button','submit','reset','image'].includes(type)) return 'button';
+        if (type === 'checkbox') return 'checkbox';
+        if (type === 'radio') return 'radio';
+        if (type === 'range') return 'slider';
+        if (type === 'number') return 'spinbutton';
+        if (type === 'search') return 'searchbox';
+        if (!['hidden','file','color'].includes(type)) return 'textbox';
+      }
+      return '';
+    }
     return source.filter(el => {
       const style = getComputedStyle(el);
       const visible = style.visibility !== 'hidden' && style.display !== 'none' && el.getClientRects().length > 0;
       const hay = (el.innerText || el.textContent || el.getAttribute('aria-label') || '').trim();
-      const r = el.getAttribute('role') || '';
-      return visible && (!needle || hay.toLowerCase().includes(needle.toLowerCase())) && (!role || r.toLowerCase() === role.toLowerCase());
+      const r = semanticRole(el);
+      return visible && (!needle || hay.toLowerCase().includes(needle.toLowerCase())) && (!role || r === role.toLowerCase());
     }).slice(0, ${limit}).map(el => {
       const b = el.getBoundingClientRect();
-      return { selector: cssPath(el), tag: el.tagName.toLowerCase(), role: el.getAttribute('role') || '', name: el.getAttribute('aria-label') || el.getAttribute('name') || '', text: (el.innerText || el.textContent || '').trim().slice(0, 500), value: 'value' in el ? String(el.value).slice(0, 500) : '', x: b.x, y: b.y, width: b.width, height: b.height, disabled: Boolean(el.disabled) };
+      return { selector: cssPath(el), tag: el.tagName.toLowerCase(), role: semanticRole(el), name: el.getAttribute('aria-label') || el.getAttribute('name') || '', text: (el.innerText || el.textContent || '').trim().slice(0, 500), value: 'value' in el ? String(el.value).slice(0, 500) : '', x: b.x, y: b.y, width: b.width, height: b.height, disabled: Boolean(el.disabled) };
     });
   })()`;
 }

package/src/extended/desktop-linux.mjs

@@ -583,8 +583,28 @@ while queue:
   for i in range(e.childCount):queue.append((e.getChildAtIndex(i),app_name,pid,window_name,node_path+"."+str(i)))
  except:pass
 if target is None:raise Exception("UI element not found")
-if action in ("click","invoke","select","toggle"):
- target.queryAction().doAction(0)
+pointer_fallback=None
+if action in ("click","invoke"):
+ try:
+  applied=target.queryAction().doAction(0)
+  if applied is False: raise Exception("accessibility action returned false")
+ except Exception as action_error:
+  try:
+   extents=target.queryComponent().getExtents(pyatspi.DESKTOP_COORDS)
+   def _rect_value(rect,name,index):
+    try: return int(getattr(rect,name))
+    except:
+     try: return int(rect[index])
+     except: return 0
+   px=_rect_value(extents,"x",0);py=_rect_value(extents,"y",1)
+   pw=_rect_value(extents,"width",2);ph=_rect_value(extents,"height",3)
+   if pw<=0 or ph<=0: raise Exception("target has no usable screen bounds")
+   pointer_fallback={"x":px+pw//2,"y":py+ph//2,"width":pw,"height":ph,"reason":str(action_error)}
+  except Exception as bounds_error:
+   raise Exception("accessibility action failed: %s; coordinate fallback unavailable: %s"%(action_error,bounds_error))
+elif action in ("select","toggle"):
+ applied=target.queryAction().doAction(0)
+ if applied is False: raise Exception(action+" accessibility action returned false")
 elif action=="focus":
  if target.queryComponent().grabFocus() is False: raise Exception("focus accessibility action returned false")
 elif action=="set_value":
@@ -618,10 +638,28 @@ elif action in ("add_to_selection","remove_from_selection"):
   if not selection.selectChild(idx): raise Exception("could not add item to selection")
  else:
   if not selection.deselectChild(idx): raise Exception("could not remove item from selection")
-print(json.dumps({"action":action,"name":target.name or "","role":target.getRoleName(),**target_meta}))`;
+out={"action":action,"name":target.name or "","role":target.getRoleName(),**target_meta}
+if pointer_fallback is not None: out["pointer_fallback"]=pointer_fallback
+print(json.dumps(out))`;
   const result = await runFile('python3', ['-c', script], { label: 'AT-SPI action', timeout: 30_000, allowFailure: true });
   let payload = null;
   try { payload = JSON.parse(result.stdout || ''); } catch {}
+  if (payload?.pointer_fallback) {
+    const fallback = payload.pointer_fallback;
+    const x = Number(fallback.x), y = Number(fallback.y);
+    if (!Number.isFinite(x) || !Number.isFinite(y)) unavailable('Linux accessibility action', 'coordinate fallback returned invalid screen bounds');
+    await pointer({ action:'click', x, y, ...(args.backend ? { backend:args.backend } : {}) });
+    return jsonResult({
+      action,
+      id:id || null,
+      name:payload.name || '',
+      role:payload.role || '',
+      backend:'pointer_fallback',
+      x,
+      y,
+      fallback_reason:String(fallback.reason || 'accessibility action unavailable').slice(0, 500),
+    });
+  }
   if (payload && !payload.error) return text(JSON.stringify(payload));
   if (payload?.error) unavailable('Linux accessibility action', payload.error);
   if (result.code !== 0) {

package/src/extended/desktop.mjs

@@ -330,9 +330,21 @@ export async function uiSnapshot(args = {}) {
   const activeOnly = args.active_only != null
     ? args.active_only === true
     : (!explicitlyScoped && args.scope !== 'desktop');
-  const effective = { ...args, active_only:activeOnly };
-  const raw = await adapter.uiSnapshot(effective);
-  const value = resultValue(raw);
+  let effective = { ...args, active_only:activeOnly };
+  let raw = await adapter.uiSnapshot(effective);
+  let value = resultValue(raw);
+  const requestedDepth = Number(effective.max_depth);
+  const emptySnapshot = Array.isArray(value)
+    ? value.length === 0
+    : Array.isArray(value?.nodes) && value.nodes.length === 0;
+  if (process.platform === 'linux' && Number.isFinite(requestedDepth) && requestedDepth > 15 && emptySnapshot) {
+    // Some GNOME/AT-SPI providers collapse otherwise valid trees to an empty result
+    // when traversed too deeply. Preserve the caller's deep request first, but recover
+    // from that provider failure with the deepest stable fallback observed on Wayland.
+    effective = { ...effective, max_depth:15 };
+    raw = await adapter.uiSnapshot(effective);
+    value = resultValue(raw);
+  }
   return jsonResult(enrichUiPayload(value, effective));
 }
 export async function uiAction(args = {}) {
@@ -452,7 +464,7 @@ async function uiMatches(args = {}) {
   const scopedPid = resolved.pid != null ? resolved.pid : stableLinuxPid ? Number(stableLinuxPid) : null;
   const result = await uiSnapshot({
     max_nodes: resolved.max_nodes || 2000,
-    max_depth: resolved.max_depth || 16,
+    max_depth: resolved.max_depth || (process.platform === 'linux' ? 12 : 16),
     ...(scopedPid != null ? { pid: scopedPid } : {}),
     ...(resolved.app ? { app: resolved.app } : {}),
     ...((resolved.window_title || resolved.windowTitle) ? { window_title: resolved.window_title || resolved.windowTitle } : {}),
@@ -508,7 +520,7 @@ async function uiConditionMatches(args, condition) {
     if (!cached) {
       const snapshot = resultValue(await uiSnapshot({
         max_nodes:args.max_nodes || 2000,
-        max_depth:args.max_depth || 16,
+        max_depth:args.max_depth || (process.platform === 'linux' ? 12 : 16),
         ...(args.pid != null ? { pid:args.pid } : {}),
         ...(args.app ? { app:args.app } : {}),
         ...((args.window_title || args.windowTitle) ? { window_title:args.window_title || args.windowTitle } : {}),

package/src/extended/documents.mjs

@@ -336,6 +336,29 @@ export async function editDocument(args) {
   }
 }
 
+export function expandPdfPages(spec, maxPages = 10_000) {
+  const normalized = String(spec || '').replace(/\s+/g, '');
+  if (!normalized || !/^[0-9,\-]+$/.test(normalized)) throw new Error('pages must look like 1-3,5');
+  const pages = [];
+  for (const token of normalized.split(',')) {
+    if (!token) throw new Error('pages must look like 1-3,5');
+    const range = token.match(/^(\d+)-(\d+)$/);
+    if (range) {
+      const start = Number(range[1]);
+      const end = Number(range[2]);
+      if (!Number.isInteger(start) || !Number.isInteger(end) || start < 1 || end < start) throw new Error(`Invalid PDF page range: ${token}`);
+      if (pages.length + (end - start + 1) > maxPages) throw new Error(`PDF page selection exceeds ${maxPages} pages`);
+      for (let page = start; page <= end; page += 1) pages.push(page);
+      continue;
+    }
+    const page = Number(token);
+    if (!Number.isInteger(page) || page < 1) throw new Error(`Invalid PDF page: ${token}`);
+    pages.push(page);
+    if (pages.length > maxPages) throw new Error(`PDF page selection exceeds ${maxPages} pages`);
+  }
+  return pages;
+}
+
 function pdfAnnotations(buffer) {
   const raw = buffer.toString('latin1');
   const rows = [];
@@ -382,12 +405,38 @@ export async function pdfAction(args) {
     else unavailable('PDF split','install poppler-utils (pdfseparate)');
     return jsonResult({action,output_dir:outputDir,files:(await readdir(outputDir)).filter(name=>name.toLowerCase().endsWith('.pdf')).sort()});
   }
-  const pages=optionalString(args.pages); if(!pages || !/^[0-9,\-\s]+$/.test(pages)) throw new Error('pages must look like 1-3,5');
+  const pages=optionalString(args.pages);
+  const selectedPages=expandPdfPages(pages);
+  const normalizedPages=selectedPages.join(',');
   const output=await resolveSafePath(args.output,'output');
-  if(commandExists('qpdf')) await runFile('qpdf',[source,'--pages','.',pages.replace(/\s+/g,''),'--',output],{label:'PDF extract pages',timeout:60_000});
-  else if(commandExists('pdftk')) await runFile('pdftk',[source,'cat',...pages.replace(/\s+/g,'').split(','),'output',output],{label:'PDF extract pages',timeout:60_000});
-  else unavailable('PDF page extraction','install qpdf or pdftk');
-  return jsonResult({action,source,pages,output,bytes:(await stat(output)).size});
+  if(commandExists('qpdf')) {
+    await runFile('qpdf',[source,'--pages','.',pages.replace(/\s+/g,''),'--',output],{label:'PDF extract pages',timeout:60_000});
+  } else if(commandExists('pdftk')) {
+    await runFile('pdftk',[source,'cat',...pages.replace(/\s+/g,'').split(','),'output',output],{label:'PDF extract pages',timeout:60_000});
+  } else if(commandExists('pdfseparate')) {
+    const dir=await tempDir('remcp-pdf-pages-');
+    try {
+      const extracted=[];
+      for(let index=0;index<selectedPages.length;index+=1){
+        const page=selectedPages[index];
+        const pattern=path.join(dir,`selection-${index+1}-%d.pdf`);
+        await runFile('pdfseparate',['-f',String(page),'-l',String(page),source,pattern],{label:`PDF extract page ${page}`,timeout:60_000});
+        extracted.push(path.join(dir,`selection-${index+1}-${page}.pdf`));
+      }
+      if(extracted.length===1) {
+        await copyFile(extracted[0],output);
+      } else if(commandExists('pdfunite')) {
+        await runFile('pdfunite',[...extracted,output],{label:'PDF assemble extracted pages',timeout:60_000});
+      } else {
+        unavailable('PDF page extraction','pdfunite is required with pdfseparate when extracting multiple pages');
+      }
+    } finally {
+      await removeTemp(dir);
+    }
+  } else {
+    unavailable('PDF page extraction','install qpdf, pdftk, or poppler-utils (pdfseparate; pdfunite for multiple pages)');
+  }
+  return jsonResult({action,source,pages:normalizedPages,output,bytes:(await stat(output)).size});
 }
 
 export const documentHandlers = {

package/src/sessions.mjs

@@ -68,6 +68,10 @@ export function createProcessSession({ pid, child, command, shell }) {
     droppedLines: 0,
     bufferedChars: 0,
     cursor: 0,
+    // wait_for_process_output has its own consumption watermark. start_process may
+    // display initial output and advance the ordinary read cursor, but the first
+    // wait must still be able to match that retained output once.
+    waitCursor: 0,
     exitCode: null,
     signal: null,
     exited: false,

package/src/tools/terminal.mjs

@@ -150,6 +150,7 @@ export async function readProcessOutputTool(args, extra = {}) {
       }
     }
     slice = readNewOutput(session);
+    session.waitCursor = session.cursor;
     const first = Math.max(1, session.cursor - slice.length + 1);
     range = slice.length ? `${first}-${session.cursor} of ${totalLines(session)}` : `no new output (${totalLines(session)} lines total)`;
   }
@@ -181,9 +182,11 @@ export async function waitForProcessOutputTool(args, extra = {}) {
   const pattern = requireString(args.pattern, 'pattern');
   const matcher = compileWaiter(pattern);
   const timeoutMs = clampInteger(args.timeout_ms, 10000, 0, 120000);
-  const cursorStart = Math.max(0, session.cursor - session.droppedLines);
-  // Output that was already buffered before the call still counts: a pattern printed
-  // earlier should answer immediately instead of spinning for the whole timeout.
+  const waitCursor = Number.isInteger(session.waitCursor) ? session.waitCursor : session.cursor;
+  const cursorStart = Math.max(0, waitCursor - session.droppedLines);
+  // Output already retained before the call still counts once, even when start_process
+  // displayed it and advanced the ordinary read cursor. A separate wait watermark keeps
+  // retries from replaying the same historical match forever.
   let slice = session.lines.slice(cursorStart);
   const bufferedMatch = waiterMatches(slice, matcher);
   const deadline = Date.now() + timeoutMs;
@@ -196,6 +199,7 @@ export async function waitForProcessOutputTool(args, extra = {}) {
   }
   const matched = waiterMatches(slice, matcher);
   session.cursor = session.droppedLines + session.lines.length;
+  session.waitCursor = session.cursor;
   session.lastPartialRead = session.partial || null;
   const status = describeSession(session);
   const headline = matched
@@ -220,6 +224,7 @@ export async function interactWithProcessTool(args, extra = {}) {
   const input = typeof args.input === 'string' ? args.input : fail('input must be a string');
   const timeoutMs = clampInteger(args.timeout_ms, 1000, 0, 120000);
   session.cursor = session.droppedLines + session.lines.length;
+  session.waitCursor = session.cursor;
   session.lastPartialRead = null;
   throwIfAborted(extra.signal);
   const stdin = session.child?.stdin;
@@ -239,6 +244,7 @@ export async function interactWithProcessTool(args, extra = {}) {
     await waitForProcessActivity(session, Math.min(200, Math.max(20, deadline - Date.now())));
   }
   const slice = readNewOutput(session);
+  session.waitCursor = session.cursor;
   const status = describeSession(session);
   const output = slice.join('\n');
   return terminalStructured({