infinicode 2.8.51 → 2.8.53

package/dist/kernel/federation/dashboard-html.js

@@ -5448,14 +5448,16 @@ button.act:disabled{opacity:0.5;cursor:default;}
     // LiveKit is the authoritative camera feed. The standalone OpenCV MJPEG
     // server is opt-in because running it beside preview_agent would open the
     // same Windows camera twice and make both feeds unreliable.
-    function tryVisionOverlay(n){
-      var img=$('pk-vision-img'); if(!img) return;
-      if(String(window.localStorage.getItem('robopark.visionOverlay')||'').toLowerCase()!=='true'){
-        visionOn=false; img.src=''; img.style.display='none'; return;
-      }
-      var row=schedRow(n), ip=row&&row.lan_ip;
-      if(!ip){ visionOn=false; img.src=''; img.style.display='none'; return; }
-      var port=(row&&row.vision_port)||5000;
+    function tryVisionOverlay(n, force){
+      var img=$('pk-vision-img'); if(!img) return;
+      if(!force && String(window.localStorage.getItem('robopark.visionOverlay')||'').toLowerCase()!=='true'){
+        visionOn=false; img.src=''; img.style.display='none'; return;
+      }
+      var row=schedRow(n)||{}, device=schedulerDevice(n)||{};
+      var preferTail=/^100\./.test(location.hostname);
+      var ip=preferTail ? (device.tailscale_ip||row.tailscale_ip||device.lan_ip||row.lan_ip) : (device.lan_ip||row.lan_ip||device.tailscale_ip||row.tailscale_ip);
+      if(!ip){ visionOn=false; img.src=''; img.style.display='none'; return; }
+      var port=row.vision_port||5000;
       visionOn=true;
       img.onerror=function(){ visionOn=false; img.style.display='none'; img.src=''; };
       img.onload=function(){ img.style.display='block'; };
@@ -5485,7 +5487,8 @@ button.act:disabled{opacity:0.5;cursor:default;}
       var device=schedulerDevice(n)||{}, inv=device.device_inventory||{};
       if(hasRealInventory(inv)) deviceSelectorsReady=true;
       var search=$('pk-device-search'), query=(search&&search.value||'').trim().toLowerCase();
-      var video=deviceList(inv,'video',['auto','none','picamera','/dev/video0','/dev/video1','/dev/video2']);
+      // Do not invent Picamera/V4L2 choices when the robot has not reported hardware.
+      var video=deviceList(inv,'video',['auto','none']);
       var input=deviceList(inv,'audio_input',['default','none']);
       var output=deviceList(inv,'audio_output',['default','none']);
       function fill(id, list, selected){
@@ -5501,7 +5504,11 @@ button.act:disabled{opacity:0.5;cursor:default;}
       fill('pk-audio-output', output, device.audio_output_device||'default');
       var greetings=$('pk-greeting-phrases');
       if(greetings && document.activeElement!==greetings) greetings.value=(device.greeting_phrases||[]).join(String.fromCharCode(10));
-      var note=$('pk-device-note'); if(note) note.textContent=(device.id?'inventory from '+(device.name||device.id):'waiting for robot heartbeat inventory');
+      var note=$('pk-device-note');
+      if(note){
+        var source=inv.source||'robot heartbeat', counts=(inv.video||[]).length+'/'+(inv.audio_input||[]).length+'/'+(inv.audio_output||[]).length;
+        note.textContent=hasRealInventory(inv) ? ('inventory from '+source+' · camera/mic/speaker '+counts) : (device.id ? 'no real hardware inventory reported yet · restart vision-agent + preview-agent on this robot' : 'waiting for robot heartbeat inventory');
+      }
     }
     function saveDeviceSelectors(n){
       var device=schedulerDevice(n), id=device&&device.id; if(!id){ showOut('device inventory is not available yet'); return; }
@@ -5908,7 +5915,13 @@ button.act:disabled{opacity:0.5;cursor:default;}
         _bindShellSection(n);
         _spkTestRenderHistory(schedId(n));
         var runSpkTest = $('pk-spk-runtest'); if(runSpkTest) runSpkTest.addEventListener('click', function(){ if(rpReadOnly){ showOut('read-only node'); return; } runSpeakerTest(n); });
-        var gl=$('pk-golive'); if(gl) gl.addEventListener('click', function(){ if(rpReadOnly){ showOut('read-only node'); return; } goLive(n); });
+        var gl=$('pk-golive');
+        if(gl){
+          gl.addEventListener('click', function(){ if(rpReadOnly){ showOut('read-only node'); return; } goLive(n); });
+          var vl=document.createElement('button'); vl.className='rp-btn'; vl.id='pk-vision-live'; vl.textContent='Open RoboVision feed';
+          vl.addEventListener('click', function(){ tryVisionOverlay(n, true); if(!visionOn) showOut('RoboVision feed unavailable: robot has no reachable LAN/Tailscale address'); });
+          gl.after(vl);
+        }
         var pm=$('pk-prodmode'); if(pm) pm.addEventListener('change', function(){ if(rpReadOnly){ pm.checked=!pm.checked; showOut('read-only node'); return; } setProductionMode(n, pm.checked); });
         var svcCopy=$('pk-svc-clicmd-copy'); if(svcCopy) svcCopy.addEventListener('click', function(){ var i=$('pk-svc-clicmd'); i.select(); try{document.execCommand('copy');}catch(e){} showOut('command copied — run it on the robot itself'); });
         var svcList=$('pk-svc-list'); if(svcList) svcList.addEventListener('click', function(ev){

package/dist/robopark/python-env.d.ts

@@ -2,6 +2,7 @@
 export declare const VISION_REQUIRED_MODULES: string[];
 export declare function findSchedulerPath(): Promise<string | null>;
 export declare function findVisionPath(): Promise<string | null>;
+export declare function findVisionAudioPath(): Promise<string | null>;
 export declare function findPython(): string;
 /**
  * Ensure a robot-side script's Python deps are importable, installing them

package/dist/robopark/python-env.js

@@ -39,6 +39,9 @@ export async function findSchedulerPath() {
 export async function findVisionPath() {
     return findPackageScript('vision/app_pi_clean.py');
 }
+export async function findVisionAudioPath() {
+    return findPackageScript('vision/audio_server_pi.py');
+}
 export function findPython() {
     const names = process.platform === 'win32'
         ? ['python', 'python3', 'py']

package/dist/robopark/stop-all.js

@@ -23,6 +23,7 @@ const PATTERNS = [
     'scheduler.{0,3}main\\.py',
     'preview_agent.py',
     'app_pi_clean.py',
+    'audio_server_pi.py',
 ];
 export async function stopAll() {
     const platform = process.platform;

package/dist/robopark/vision-agent-launcher.js

@@ -7,7 +7,7 @@
  */
 import { spawn } from 'node:child_process';
 import chalk from 'chalk';
-import { findPython, findVisionPath, prepareRobotPython, VISION_REQUIRED_MODULES } from './python-env.js';
+import { findPython, findVisionPath, findVisionAudioPath, prepareRobotPython, VISION_REQUIRED_MODULES } from './python-env.js';
 const DEFAULT_VISION_PORT = 5000;
 const DEFAULT_MOTION_WEBHOOK_PORT = 5057; // must match preview-agent's --vision-webhook-port default
 function isPiArm() {
@@ -24,7 +24,8 @@ export async function roboparkVisionAgent(opts) {
     // requirements_pi_unified.txt) — pip-installing it here would fight the
     // ARM-optimized system build. Everywhere else: plain pip install works.
     const reqFile = isPiArm() ? 'requirements_pi_unified.txt' : 'requirements-demo.txt';
-    const python = prepareRobotPython(basePython, script, VISION_REQUIRED_MODULES, reqFile);
+    const requiredModules = isPiArm() ? [...VISION_REQUIRED_MODULES, 'fastapi', 'uvicorn', 'sounddevice', 'soundfile', 'groq'] : VISION_REQUIRED_MODULES;
+    const python = prepareRobotPython(basePython, script, requiredModules, reqFile);
     if (!python)
         process.exit(1);
     const port = opts.port ?? String(DEFAULT_VISION_PORT);
@@ -33,19 +34,37 @@ export async function roboparkVisionAgent(opts) {
     const args = [script, '--port', port, '--motion-webhook-url', webhookUrl];
     if (motionActive)
         args.push('--motion-active');
+    // The original RoboVision audio selector lives in audio_server_pi.py.
+    // Start it beside the camera service on real Pi robots so /devices and
+    // /set-device remain the authoritative source for Park dropdowns.
+    let audioScript = null;
+    if (isPiArm())
+        audioScript = await findVisionAudioPath();
     console.log(chalk.bold('\n  robopark vision-agent'));
     console.log(chalk.dim('  ' + '─'.repeat(52)));
     console.log(`  port:    ${chalk.cyan(port)}`);
     console.log(`  webhook: ${chalk.cyan(webhookUrl)}`);
     console.log(`  motion:  ${motionActive ? chalk.green('armed') : chalk.dim('off')}`);
+    if (audioScript)
+        console.log(`  audio:   ${chalk.green('RoboVision audio server :8000')}`);
     console.log();
     if (opts.foreground) {
+        if (audioScript) {
+            const audio = spawn(python, [audioScript], { stdio: 'inherit', env: { ...process.env, ROBOPARK_AUDIO_SERVER: '1' } });
+            audio.unref();
+        }
         const proc = spawn(python, args, { stdio: 'inherit' });
         await new Promise((resolve) => {
             proc.on('close', () => resolve());
         });
         return;
     }
+    if (audioScript) {
+        const audio = spawn(python, [audioScript], {
+            stdio: 'ignore', detached: true, env: { ...process.env, ROBOPARK_AUDIO_SERVER: '1' },
+        });
+        audio.unref();
+    }
     const proc = spawn(python, args, {
         stdio: 'ignore',
         detached: true,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "infinicode",
-  "version": "2.8.51",
+  "version": "2.8.53",
   "description": "OpenKernel — provider-agnostic AI execution kernel. Native coding agent + mission-driven execution runtime.",
   "type": "module",
   "main": "./dist/kernel/index.js",

package/packages/robopark/vision/app_pi_clean.py

@@ -19,6 +19,7 @@ current_camera_index = 0
 camera = None
 audio_input_device = "default"
 audio_output_device = "default"
+ROBOVISION_AUDIO_URL = os.getenv("ROBOVISION_AUDIO_URL", "http://127.0.0.1:8000")
 
 def _open_camera(index):
     # On Windows, cv2.VideoCapture(index) with no explicit backend can
@@ -344,22 +345,30 @@ def switch_camera():
 
 
 def _audio_inventory():
-    inputs = [{"id": "default", "name": "System default input"}]
-    outputs = [{"id": "default", "name": "System default output"}]
+    """Adapt RoboVision's existing audio_server_pi /devices response."""
     try:
-        import pyaudio
-        pa = pyaudio.PyAudio()
-        for index in range(pa.get_device_count()):
-            info = pa.get_device_info_by_index(index)
-            item = {"id": str(index), "name": str(info.get("name", f"Audio device {index}")), "host_api": str(info.get("hostApi", ""))}
-            if info.get("maxInputChannels", 0) > 0:
+        response = requests.get(f"{ROBOVISION_AUDIO_URL.rstrip('/')}/devices", timeout=0.8)
+        response.raise_for_status()
+        payload = response.json()
+        inputs, outputs = [], []
+        for device in payload.get("devices", []):
+            item = {
+                "id": str(device["index"]),
+                "name": str(device.get("name", f"Audio device {device['index']}")),
+                "backend": "robovision_audio",
+                "sample_rate": device.get("default_samplerate"),
+            }
+            if device.get("max_input_channels", 0) > 0:
                 inputs.append(item.copy())
-            if info.get("maxOutputChannels", 0) > 0:
+            if device.get("max_output_channels", 0) > 0:
                 outputs.append(item.copy())
-        pa.terminate()
-    except Exception:
-        pass
-    return inputs, outputs
+        return inputs, outputs, {
+            "input": payload.get("bluetooth_input"),
+            "output": payload.get("bluetooth_output"),
+            "online": True,
+        }
+    except Exception as exc:
+        return [], [], {"online": False, "error": str(exc)}
 
 
 @app.route('/api/media/inventory', methods=['GET'])
@@ -370,12 +379,17 @@ def media_inventory():
         # still a valid selectable camera and is safer than hiding it.
         cameras = [{"id": path, "index": path, "name": f"Camera {path}", "backend": "v4l2"}
                    for path in sorted(glob.glob('/dev/video*'))]
-    inputs, outputs = _audio_inventory()
+    inputs, outputs, audio_state = _audio_inventory()
     return jsonify({
         "video": [{"id": "auto", "name": "Auto detect"}, {"id": "none", "name": "Disable camera"}] + cameras,
         "audio_input": inputs,
         "audio_output": outputs,
-        "selected": {"video_device": str(current_camera_index), "audio_device": audio_input_device, "audio_output_device": audio_output_device},
+        "selected": {
+            "video_device": str(current_camera_index),
+            "audio_device": str(audio_state.get("input") if audio_state.get("input") is not None else audio_input_device),
+            "audio_output_device": str(audio_state.get("output") if audio_state.get("output") is not None else audio_output_device),
+        },
+        "audio_server": audio_state,
         "source": "robovision_pi",
     })
 
@@ -397,6 +411,18 @@ def media_config():
             audio_input_device = str(data["audio_device"])
         if "audio_output_device" in data:
             audio_output_device = str(data["audio_output_device"])
+        if "audio_device" in data or "audio_output_device" in data:
+            # audio_server_pi.py is RoboVision's authoritative selector.
+            # It accepts its original sounddevice indices via input/output.
+            payload = {}
+            if "audio_device" in data:
+                payload["input"] = int(audio_input_device) if audio_input_device.isdigit() else audio_input_device
+            if "audio_output_device" in data:
+                payload["output"] = int(audio_output_device) if audio_output_device.isdigit() else audio_output_device
+            try:
+                requests.post(f"{ROBOVISION_AUDIO_URL.rstrip('/')}/set-device", json=payload, timeout=0.8).raise_for_status()
+            except Exception:
+                pass
     return jsonify({"video_device": str(current_camera_index), "audio_device": audio_input_device, "audio_output_device": audio_output_device, "source": "robovision_pi"})
 
 @app.route('/api/motion/status', methods=['GET'])

package/packages/robopark/vision/requirements_pi_unified.txt

@@ -67,7 +67,8 @@ pydantic==2.5.0
 # NETWORKING & HTTP
 # ============================================================================
 # HTTP requests
-requests==2.31.0
+requests==2.31.0
+groq>=0.11
 
 # CORS support
 python-multipart==0.0.6