added support for microphone

streaming-agent/gst_agent.py

@@ -229,6 +229,7 @@ async def run_signaling(session_id, video_track_name, audio_track_name):
             'sessionId':      session_id,
             'videoTrackName': video_track_name,
             'audioTrackName': audio_track_name,
+            'containerName':  os.environ.get('CONTAINER_NAME', 'css-test'),
         }))
         log.info('Registered with signaling server')
 

streaming-agent/signaling-server.js

@@ -2,6 +2,7 @@
  * Signaling Server
  *
  * Handles two tracks: video and game audio.
+ * Also relays browser microphone audio to container PulseAudio via pacat.
  * All Cloudflare API calls happen server-side.
  *
  * Run: CF_APP_ID=xxx CF_APP_TOKEN=xxx node signaling-server.js
@@ -14,12 +15,24 @@ const http      = require('http');
 const WebSocket = require('ws');
 const fetch     = require('node-fetch');
 const path      = require('path');
+const { spawn } = require('child_process');
 
 const PORT         = process.env.PORT         || 3000;
 const CF_APP_ID    = process.env.CF_APP_ID    || '';
 const CF_APP_TOKEN = process.env.CF_APP_TOKEN || '';
 const CF_BASE      = `https://rtc.live.cloudflare.com/v1/apps/${CF_APP_ID}`;
 
+// Container's PulseAudio sink to write mic audio into
+// pacat writes raw PCM to the null sink which virtual_mic reads from
+const PACAT_ARGS = [
+  '--playback',
+  '--device=mic_sink',
+  '--format=s16le',
+  '--rate=48000',
+  '--channels=1',
+  '--latency-msec=20',
+];
+
 if (!CF_APP_ID || !CF_APP_TOKEN) {
   console.error('[sig] CF_APP_ID and CF_APP_TOKEN must be set');
   process.exit(1);
@@ -57,12 +70,68 @@ app.get('/', (req, res) => res.sendFile(path.join(__dirname, '..', 'client.html'
 const agentSessions  = new Map();
 const browserToAgent = new Map();
 
+// Per-browser pacat process for mic audio
+// browserSessionId -> pacat child process
+const pacatProcesses = new Map();
+
+function startPacat(browserSessId) {
+  // Find the container ID for this browser session
+  // pacat runs inside the Docker container via docker exec
+  const agentId = browserToAgent.get(browserSessId);
+  const entry   = agentSessions.get(agentId);
+  if (!entry) return null;
+
+  // Get container name from agent — agent sends it on register
+  const containerName = entry.containerName || 'css-test';
+
+  const pacat = spawn('docker', [
+    'exec', '-i', containerName,
+    'pacat', ...PACAT_ARGS,
+  ]);
+
+  pacat.stderr.on('data', d => {
+    console.error(`[pacat:${browserSessId}]`, d.toString().trim());
+  });
+
+  pacat.on('close', code => {
+    console.log(`[pacat:${browserSessId}] exited with code ${code}`);
+    pacatProcesses.delete(browserSessId);
+  });
+
+  pacatProcesses.set(browserSessId, pacat);
+  console.log(`[sig] pacat started for browser session ${browserSessId}`);
+  return pacat;
+}
+
+function stopPacat(browserSessId) {
+  const pacat = pacatProcesses.get(browserSessId);
+  if (pacat) {
+    pacat.kill();
+    pacatProcesses.delete(browserSessId);
+    console.log(`[sig] pacat stopped for browser session ${browserSessId}`);
+  }
+}
+
 wss.on('connection', ws => {
   let role           = null;
   let agentSessionId = null;
   let browserSessId  = null;
 
-  ws.on('message', async raw => {
+  ws.on('message', async (raw, isBinary) => {
+    // Binary messages are raw PCM audio from browser microphone
+    if (isBinary) {
+      if (browserSessId) {
+        let pacat = pacatProcesses.get(browserSessId);
+        if (!pacat) {
+          pacat = startPacat(browserSessId);
+        }
+        if (pacat && pacat.stdin.writable) {
+          pacat.stdin.write(raw);
+        }
+      }
+      return;
+    }
+
     let msg;
     try { msg = JSON.parse(raw); } catch { return; }
 
@@ -73,9 +142,17 @@ wss.on('connection', ws => {
         agentSessionId: msg.sessionId,
         videoTrackName: msg.videoTrackName,
         audioTrackName: msg.audioTrackName,
+        containerName:  msg.containerName || 'css-test',
         agentWs:        ws,
       });
-      console.log(`[sig] agent registered — sessionId=${agentSessionId} video=${msg.videoTrackName} audio=${msg.audioTrackName}`);
+      console.log(`[sig] agent registered — sessionId=${agentSessionId} container=${msg.containerName || 'css-test'}`);
+    }
+
+    else if (msg.type === 'mic_stop') {
+      // Browser released PTT key — stop pacat
+      if (browserSessId) {
+        stopPacat(browserSessId);
+      }
     }
 
     else if (msg.type === 'browser_connect') {
@@ -147,6 +224,7 @@ wss.on('connection', ws => {
       console.log(`[sig] agent disconnected — sessionId=${agentSessionId}`);
     }
     if (role === 'browser' && browserSessId) {
+      stopPacat(browserSessId);
       browserToAgent.delete(browserSessId);
     }
   });