Add voice capable CLI agents and raise room capacity

README.md

@@ -216,6 +216,7 @@ pairux login                     # Optional: label your agents as yours (browser
 pairux join ABC123 --name Claude # Join with the session's join code
 pairux listen --json             # Follow chat and who is here; keeps the agent present
 pairux say "Tests pass"          # Post in the chat
+pairux voice join --json         # Hear call audio; stdin speaks through C0mpute TTS
 pairux who                       # Who is in the session
 pairux leave
 ```

apps/desktop/electron.vite.config.ts

@@ -54,6 +54,7 @@ export default defineConfig({
       rollupOptions: {
         input: {
           index: resolve(__dirname, 'src/renderer/index.html'),
+          'agent-voice': resolve(__dirname, 'src/renderer/agent-voice.html'),
         },
       },
     },

apps/desktop/public/agent-voice-worklet.js

@@ -0,0 +1,15 @@
+class PairuxVoiceCaptureProcessor extends AudioWorkletProcessor {
+  process(inputs, outputs) {
+    const output = outputs[0]?.[0];
+    output?.fill(0);
+
+    const input = inputs[0]?.[0];
+    if (input?.length) {
+      const samples = input.slice();
+      this.port.postMessage(samples, [samples.buffer]);
+    }
+    return true;
+  }
+}
+
+registerProcessor('pairux-voice-capture', PairuxVoiceCaptureProcessor);

apps/desktop/src/main/agent-cli/api.ts

@@ -42,6 +42,17 @@ export interface JoinResult {
   owned: boolean;
 }
 
+export interface AgentVoiceToken {
+  token: string;
+  url: string;
+  roomName: string;
+  iceServers?: {
+    urls: string | string[];
+    username?: string;
+    credential?: string;
+  }[];
+}
+
 export class ApiError extends Error {
   constructor(
     readonly status: number,
@@ -133,6 +144,13 @@ export class PairuxClient {
     });
   }
 
+  voiceToken(participantId: string) {
+    return this.request<AgentVoiceToken>(
+      `/api/v1/agents/${encodeURIComponent(participantId)}/voice-token`,
+      { method: 'POST', body: '{}' }
+    );
+  }
+
   /** Look up a live session by its 6-character join code. */
   lookupJoinCode(joinCode: string) {
     return this.request<{ id: string; status?: string; scheduled?: boolean }>(

apps/desktop/src/main/agent-cli/commands.ts

@@ -8,6 +8,7 @@
  *   pairux join ABC123 --name "Claude"
  *   pairux listen                follow chat + roster (stays present)
  *   pairux say "Tests pass on my side"
+ *   pairux voice join --json     hear call audio and speak through C0mpute
  *   pairux who                   who is in the session
  *   pairux leave
  *   pairux room                  link to a fresh anonymous E2EE room
@@ -30,6 +31,7 @@ import {
   revoke,
 } from './auth';
 import { ApiError, PairuxClient, type ChatMessage, type RosterEntry } from './api';
+import type { AgentVoiceRunOptions } from './voice-protocol';
 import { newRoomLink } from './room';
 import { COPILOT_HELP, runCopilot } from './copilot';
 import { UsageError } from './errors';
@@ -49,6 +51,8 @@ export interface Deps {
   maxPolls?: number;
   /** Tests stop `copilot watch` after N stream connections; production reconnects. */
   maxStreams?: number;
+  /** Starts the hidden WebRTC renderer used by `pairux voice join`. */
+  startVoice?: (options: AgentVoiceRunOptions) => Promise<number>;
   login?: typeof listenForCallback;
   /** The app version, for the User-Agent and `pairux version`. */
   version?: string;
@@ -72,6 +76,8 @@ Usage:
   pairux who [--json]                List who is in the session
   pairux status [--json]             Show the agent this CLI is running as
   pairux leave                       Leave the session
+  pairux voice join [--json]         Join call audio; stdout is transcripts, stdin speaks
+                                     Each input line is spoken via C0mpute TTS. Send /quit to leave.
   pairux room                        Print a link to a fresh anonymous room: end-to-end
                                      encrypted voice/video, no account needed to join
   pairux redeem <CODE>               Redeem a credit code (adds days of a paid plan)
@@ -86,9 +92,16 @@ Environment:
   PAIRUX_AGENT_NAME     default display name for join
   PAIRUX_AGENT_CLIENT   default --client for join
   PAIRUX_CONFIG_DIR     where config.json lives (default ~/.config/pairux)
+  C0MPUTE_SPEECH_URL    speech gateway; default http://127.0.0.1:8787
+  C0MPUTE_SPEECH_TOKEN  token used by \`c0mpute speech serve\` (required for voice)
+  C0MPUTE_TTS_ENGINE    default kokoro
+  C0MPUTE_TTS_VOICE     default af_sarah
+  C0MPUTE_TTS_LOCALE    default en-US
+  C0MPUTE_STT_LANGUAGE  optional language hint, e.g. en
 
-Agents watch and chat. They cannot see or control anyone's screen, and the
-host can remove them at any time. Docs: https://pairux.com/docs/agents`;
+Agents can follow chat and call audio, and speak into the call. They cannot see
+or control anyone's screen, and the host can remove them at any time.
+Docs: https://pairux.com/docs/agents`;
 
 function defaultName(env: NodeJS.ProcessEnv): string {
   if (env.PAIRUX_AGENT_NAME) return env.PAIRUX_AGENT_NAME;
@@ -315,7 +328,7 @@ export async function run(argv: string[], deps: Deps): Promise<number> {
         deps.out(
           `Joined ${joinCode} as ${joined.displayName}${joined.owned ? ' (your agent)' : ' (anonymous agent)'}.`
         );
-        deps.out('Run `pairux listen` to follow the chat and stay present.');
+        deps.out('Run `pairux listen` for chat or `pairux voice join` to join call audio.');
         return 0;
       }
 
@@ -335,6 +348,60 @@ export async function run(argv: string[], deps: Deps): Promise<number> {
         return 0;
       }
 
+      case 'voice': {
+        const [action, ...voiceArgs] = rest;
+        if (!action || ['help', '--help', '-h'].includes(action)) {
+          deps.out(
+            'Usage: pairux voice join [--json]\nJoin the active call as this agent. Each stdin line is spoken with C0mpute TTS; transcript events go to stdout. Send /quit to leave.'
+          );
+          return 0;
+        }
+        if (action !== 'join') throw new UsageError(`Unknown voice command "${action}".`);
+        const { values, positionals } = parseArgs({
+          args: voiceArgs,
+          allowPositionals: true,
+          options: { json: { type: 'boolean', default: false } },
+        });
+        if (positionals.length > 0) throw new UsageError('Usage: pairux voice join [--json]');
+        if (!deps.startVoice) {
+          throw new UsageError(
+            'The PairUX desktop CLI voice runtime is unavailable in this build.'
+          );
+        }
+
+        const agent = requireAgent();
+        const speechToken = deps.env.C0MPUTE_SPEECH_TOKEN;
+        if (!speechToken) {
+          throw new UsageError(
+            'Set C0MPUTE_SPEECH_TOKEN to the token used by `c0mpute speech serve` before joining voice.'
+          );
+        }
+        const livekit = await client.voiceToken(agent.participantId);
+        return await deps.startVoice({
+          participantId: agent.participantId,
+          displayName: agent.displayName,
+          token: livekit.token,
+          url: livekit.url,
+          ...(livekit.iceServers ? { iceServers: livekit.iceServers } : {}),
+          speechUrl: deps.env.C0MPUTE_SPEECH_URL ?? 'http://127.0.0.1:8787',
+          speechToken,
+          ttsEngine: deps.env.C0MPUTE_TTS_ENGINE ?? 'kokoro',
+          ttsVoice: deps.env.C0MPUTE_TTS_VOICE ?? 'af_sarah',
+          ttsLocale: deps.env.C0MPUTE_TTS_LOCALE ?? 'en-US',
+          sttModel: 'base',
+          ...(deps.env.C0MPUTE_STT_LANGUAGE ? { language: deps.env.C0MPUTE_STT_LANGUAGE } : {}),
+          json: values.json,
+          out: deps.out,
+          err: deps.err,
+          heartbeat: async () => {
+            const result = await client.poll(agent.participantId, new Date().toISOString());
+            if (result.session.status === 'ended') {
+              throw new ApiError(410, 'The session has ended');
+            }
+          },
+        });
+      }
+
       case 'who':
       case 'status': {
         const { values } = parseArgs({

apps/desktop/src/main/agent-cli/index.ts

@@ -1,5 +1,5 @@
 /**
- * `pairux login|logout|whoami|join|listen|say|who|status|leave|room|redeem|invites|copilot`
+ * `pairux login|logout|whoami|join|listen|say|who|status|leave|voice|room|redeem|invites|copilot`
  *
  * The installed `pairux` launcher passes any argument it does not handle
  * itself (update, uninstall, --help, --version) straight to the app, the same
@@ -21,6 +21,7 @@ export const AGENT_COMMANDS = new Set([
   'who',
   'status',
   'leave',
+  'voice',
   'room',
   'redeem',
   'invites',
@@ -56,6 +57,10 @@ export function runAgentCommand(args: string[], version: string): Promise<number
     openBrowser,
     sleep: (ms) => new Promise((resolve) => setTimeout(resolve, ms)),
     readStdin: readAllStdin,
+    startVoice: async (options) => {
+      const { runAgentVoice } = await import('./voice-runtime');
+      return runAgentVoice(options);
+    },
     version,
   });
 }

apps/desktop/src/main/agent-cli/voice-protocol.ts

@@ -0,0 +1,24 @@
+export interface VoiceIceServer {
+  urls: string | string[];
+  username?: string;
+  credential?: string;
+}
+
+export interface AgentVoiceRunOptions {
+  participantId: string;
+  displayName: string;
+  token: string;
+  url: string;
+  iceServers?: VoiceIceServer[];
+  speechUrl: string;
+  speechToken: string;
+  ttsEngine: string;
+  ttsVoice: string;
+  ttsLocale: string;
+  sttModel: string;
+  language?: string;
+  json: boolean;
+  out: (line: string) => void;
+  err: (line: string) => void;
+  heartbeat: () => Promise<void>;
+}