+
+
diff --git a/Tools/Dashboard/js/codex_threads.js b/Tools/Dashboard/js/codex_threads.js
index 1ee76daf6..66aca526d 100644
--- a/Tools/Dashboard/js/codex_threads.js
+++ b/Tools/Dashboard/js/codex_threads.js
@@ -6,6 +6,9 @@ let codexSessions = [];
let codexLoaded = false;
let codexSelectedId = '';
let codexPollTimer = null;
+let codexRecorder = null;
+let codexVoiceChunks = [];
+let codexVoiceStream = null;
function codexEsc(value) {
const div = document.createElement('div');
@@ -24,6 +27,145 @@ function codexShortId(id) {
return id ? id.slice(0, 8) : '';
}
+function codexPreferredAudioMimeType() {
+ const candidates = [
+ 'audio/webm;codecs=opus',
+ 'audio/webm',
+ 'audio/mp4',
+ 'audio/wav',
+ ];
+ if (!window.MediaRecorder || !MediaRecorder.isTypeSupported) return '';
+ return candidates.find(type => MediaRecorder.isTypeSupported(type)) || '';
+}
+
+function codexAudioExtension(mimeType) {
+ const lower = (mimeType || '').toLowerCase();
+ if (lower.includes('mp4')) return 'm4a';
+ if (lower.includes('wav')) return 'wav';
+ if (lower.includes('mpeg') || lower.includes('mp3')) return 'mp3';
+ if (lower.includes('ogg')) return 'ogg';
+ return 'webm';
+}
+
+function codexSetVoiceStatus(text) {
+ const el = document.getElementById('codex-voice-status');
+ if (el) el.textContent = text || '';
+}
+
+function codexSetVoiceRecording(recording) {
+ const btn = document.getElementById('codex-voice-toggle');
+ if (!btn) return;
+ btn.classList.toggle('recording', recording);
+ btn.textContent = recording ? '停止录音' : '语音输入';
+}
+
+function codexCleanupVoiceStream() {
+ if (!codexVoiceStream) return;
+ codexVoiceStream.getTracks().forEach(track => track.stop());
+ codexVoiceStream = null;
+}
+
+function codexBlobToBase64(blob) {
+ return new Promise((resolve, reject) => {
+ const reader = new FileReader();
+ reader.onload = () => {
+ const value = String(reader.result || '');
+ resolve(value.includes(',') ? value.split(',', 2)[1] : value);
+ };
+ reader.onerror = () => reject(reader.error || new Error('读取录音失败'));
+ reader.readAsDataURL(blob);
+ });
+}
+
+function codexAppendPromptText(text) {
+ const promptEl = document.getElementById('codex-prompt');
+ if (!promptEl || !text) return;
+ const current = promptEl.value || '';
+ const separator = current.trim() ? (current.endsWith('\n') ? '' : '\n') : '';
+ promptEl.value = current + separator + text.trim();
+ promptEl.focus();
+ promptEl.selectionStart = promptEl.selectionEnd = promptEl.value.length;
+}
+
+async function codexStartVoiceInput() {
+ if (!navigator.mediaDevices?.getUserMedia || !window.MediaRecorder) {
+ codexSetVoiceStatus('当前浏览器不支持录音');
+ return;
+ }
+ try {
+ codexVoiceChunks = [];
+ const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
+ codexVoiceStream = stream;
+ const mimeType = codexPreferredAudioMimeType();
+ codexRecorder = new MediaRecorder(stream, mimeType ? { mimeType } : undefined);
+ codexRecorder.addEventListener('dataavailable', event => {
+ if (event.data && event.data.size > 0) codexVoiceChunks.push(event.data);
+ });
+ codexRecorder.addEventListener('stop', () => {
+ const blob = new Blob(codexVoiceChunks, { type: codexRecorder?.mimeType || mimeType || 'audio/webm' });
+ codexRecorder = null;
+ codexCleanupVoiceStream();
+ codexSetVoiceRecording(false);
+ codexFinishVoiceInput(blob);
+ });
+ codexRecorder.addEventListener('error', event => {
+ codexSetVoiceStatus(event.error?.message || '录音失败');
+ codexRecorder = null;
+ codexCleanupVoiceStream();
+ codexSetVoiceRecording(false);
+ });
+ codexRecorder.start();
+ codexSetVoiceRecording(true);
+ codexSetVoiceStatus('录音中，再点一次停止');
+ } catch (err) {
+ codexRecorder = null;
+ codexCleanupVoiceStream();
+ codexSetVoiceRecording(false);
+ codexSetVoiceStatus(err?.message || '无法启动麦克风');
+ }
+}
+
+function codexStopVoiceInput() {
+ if (codexRecorder && codexRecorder.state !== 'inactive') {
+ codexRecorder.stop();
+ }
+}
+
+async function codexToggleVoiceInput() {
+ if (codexRecorder && codexRecorder.state === 'recording') {
+ codexStopVoiceInput();
+ } else {
+ await codexStartVoiceInput();
+ }
+}
+
+async function codexFinishVoiceInput(blob) {
+ if (!blob || blob.size === 0) {
+ codexSetVoiceStatus('没有录到声音');
+ return;
+ }
+ codexSetVoiceStatus('正在转写...');
+ try {
+ const audioBase64 = await codexBlobToBase64(blob);
+ const resp = await fetch('/api/codex/transcribe', {
+ method: 'POST',
+ headers: { 'Content-Type': 'application/json' },
+ body: JSON.stringify({
+ audioBase64,
+ mimeType: blob.type || 'audio/webm',
+ filename: `codex-voice-${Date.now()}.${codexAudioExtension(blob.type)}`,
+ language: 'zh',
+ }),
+ });
+ const data = await resp.json();
+ if (!resp.ok || !data.success) throw new Error(data.error || `HTTP ${resp.status}`);
+ codexAppendPromptText(data.text || '');
+ codexSetVoiceStatus(data.model ? `已转成文字 · ${data.model}` : '已转成文字');
+ } catch (err) {
+ codexSetVoiceStatus(err?.message || '转写失败');
+ }
+}
+
async function codexLoadSessions(force = false) {
if (codexLoaded && !force) return;
const list = document.getElementById('codex-session-list');
@@ -113,7 +255,7 @@ async function codexSelectSession(id) {
const session = data.session || {};
if (title) title.textContent = session.title || session.id || 'Codex 会话';
if (meta) {
- meta.textContent = `${codexFormatDate(session.updated_at || session.created_at)} · ${session.id || ''} · ${session.archived ? '归档' : '当前'}`;
+ meta.textContent = `${codexFormatDate(session.updated_at || session.created_at)} · ${session.id || ''} · ${session.archived ? '归档' : '当前'} · 本机 CLI 会话`;
}
codexRenderMessages(data.messages || []);
} catch (err) {
@@ -237,6 +379,11 @@ function codexBind() {
refresh.dataset.bound = '1';
refresh.addEventListener('click', () => codexLoadSessions(true));
}
+ const voice = document.getElementById('codex-voice-toggle');
+ if (voice && !voice.dataset.bound) {
+ voice.dataset.bound = '1';
+ voice.addEventListener('click', codexToggleVoiceInput);
+ }
const runNew = document.getElementById('codex-run-new');
if (runNew && !runNew.dataset.bound) {
runNew.dataset.bound = '1';
@@ -261,6 +408,7 @@ function codexBind() {
if (panel) {
observer.observe(panel, { attributes: true, attributeFilter: ['class'] });
}
+ window.addEventListener('beforeunload', codexCleanupVoiceStream);
document.addEventListener('DOMContentLoaded', () => {
codexBind();
const panelNow = document.getElementById('panel-codex');
diff --git a/Tools/Dashboard/serve.py b/Tools/Dashboard/serve.py
index dce5c7fc5..053d64cea 100644
--- a/Tools/Dashboard/serve.py
+++ b/Tools/Dashboard/serve.py
@@ -54,6 +54,8 @@ Usage:
"""
import http.server
+import base64
+import binascii
import csv
import json
import os
@@ -103,6 +105,10 @@ CODEX_HOME = os.path.join(os.path.expanduser('~'), '.codex')
CODEX_SESSIONS_DIR = os.path.join(CODEX_HOME, 'sessions')
CODEX_ARCHIVED_SESSIONS_DIR = os.path.join(CODEX_HOME, 'archived_sessions')
CODEX_SESSION_INDEX = os.path.join(CODEX_HOME, 'session_index.jsonl')
+CODEX_TRANSCRIBE_MAX_BYTES = 25 * 1024 * 1024
+OPENROUTER_ENV_FILE = os.path.join(SCRIPT_DIR, 'private', 'community_monitor.env')
+OPENROUTER_TRANSCRIBE_ENDPOINT = 'https://openrouter.ai/api/v1/audio/transcriptions'
+OPENROUTER_TRANSCRIBE_DEFAULT_MODEL = 'openai/whisper-large-v3'
CODEX_JOBS = {}
CODEX_JOBS_LOCK = threading.Lock()
COMMUNITY_MONITOR_API = None
@@ -418,6 +424,30 @@ def _read_jsonl_records(path, limit=None):
return records
+def _load_dashboard_env_file(path):
+ if not os.path.exists(path):
+ return
+ try:
+ with open(path, 'r', encoding='utf-8-sig', errors='replace') as f:
+ for raw_line in f:
+ line = raw_line.strip()
+ if not line or line.startswith('#') or '=' not in line:
+ continue
+ key, value = line.split('=', 1)
+ key = key.strip()
+ value = value.strip().strip('"').strip("'")
+ if key and key not in os.environ:
+ os.environ[key] = value
+ except Exception:
+ pass
+
+
+def _openrouter_setting(name, default=''):
+ if name not in os.environ:
+ _load_dashboard_env_file(OPENROUTER_ENV_FILE)
+ return os.environ.get(name, default)
+
+
def _message_text(content):
if isinstance(content, str):
return content
@@ -432,6 +462,90 @@ def _message_text(content):
return '\n'.join(part for part in parts if part)
+def _decode_codex_audio_base64(value):
+ if not isinstance(value, str) or not value.strip():
+ raise ValueError('audioBase64 required')
+ cleaned = value.strip()
+ if ',' in cleaned and cleaned[:80].lower().startswith('data:'):
+ cleaned = cleaned.split(',', 1)[1]
+ try:
+ data = base64.b64decode(cleaned, validate=True)
+ except (binascii.Error, ValueError):
+ raise ValueError('invalid audioBase64')
+ if not data:
+ raise ValueError('audio is empty')
+ if len(data) > CODEX_TRANSCRIBE_MAX_BYTES:
+ raise ValueError('audio too large')
+ return data
+
+
+def _codex_audio_format(filename='', mime_type=''):
+ source = f'{mime_type or ""} {filename or ""}'.lower()
+ for fmt in ('webm', 'wav', 'mp3', 'flac', 'm4a', 'ogg', 'aac'):
+ if fmt in source:
+ return fmt
+ return 'webm'
+
+
+def _codex_transcribe_audio(audio_base64, filename='', mime_type='', language=''):
+ api_key = _openrouter_setting('OPENROUTER_API_KEY', '').strip()
+ if not api_key:
+ raise RuntimeError('OPENROUTER_API_KEY is not configured')
+
+ audio_bytes = _decode_codex_audio_base64(audio_base64)
+ model = _openrouter_setting('OPENROUTER_TRANSCRIBE_MODEL', OPENROUTER_TRANSCRIBE_DEFAULT_MODEL).strip()
+ if not model:
+ model = OPENROUTER_TRANSCRIBE_DEFAULT_MODEL
+
+ payload = {
+ 'model': model,
+ 'input_audio': {
+ 'data': base64.b64encode(audio_bytes).decode('ascii'),
+ 'format': _codex_audio_format(filename, mime_type),
+ },
+ }
+ if language:
+ payload['language'] = str(language)[:32]
+
+ body = json.dumps(payload, ensure_ascii=False).encode('utf-8')
+ request = urllib.request.Request(
+ OPENROUTER_TRANSCRIBE_ENDPOINT,
+ data=body,
+ headers={
+ 'Authorization': f'Bearer {api_key}',
+ 'Content-Type': 'application/json',
+ 'HTTP-Referer': _openrouter_setting('OPENROUTER_REFERER', 'http://localhost:8080'),
+ 'X-Title': _openrouter_setting('OPENROUTER_TITLE', 'TH1 Dashboard Codex Voice Input'),
+ },
+ method='POST',
+ )
+ try:
+ with urllib.request.urlopen(request, timeout=120) as resp:
+ raw = resp.read().decode('utf-8', errors='replace')
+ except urllib.error.HTTPError as e:
+ detail = e.read().decode('utf-8', errors='replace')[:1200]
+ raise RuntimeError(f'OpenRouter transcription failed: HTTP {e.code} {detail}')
+ except urllib.error.URLError as e:
+ raise RuntimeError(f'OpenRouter transcription failed: {e.reason}')
+
+ try:
+ data = json.loads(raw)
+ except Exception:
+ raise RuntimeError('OpenRouter transcription returned invalid JSON')
+ text = str(data.get('text') or '').strip()
+ if not text and isinstance(data.get('choices'), list) and data['choices']:
+ choice = data['choices'][0] or {}
+ message = choice.get('message') if isinstance(choice.get('message'), dict) else {}
+ text = str(message.get('content') or choice.get('text') or '').strip()
+ if not text:
+ raise RuntimeError('OpenRouter transcription returned empty text')
+ return {
+ 'text': text,
+ 'model': data.get('model') or model,
+ 'duration': data.get('duration'),
+ }
+
+
def _is_codex_context_message(role, text):
if role != 'user':
return False
@@ -1112,6 +1226,8 @@ class DashboardHandler(http.server.SimpleHTTPRequestHandler):
self._handle_dashboard_preferences_save()
elif self.path == '/api/codex/run':
self._handle_codex_run()
+ elif self.path == '/api/codex/transcribe':
+ self._handle_codex_transcribe()
elif self.path.startswith('/api/community-monitor/'):
self._handle_community_monitor_post()
# SNS APIs
@@ -1516,6 +1632,21 @@ class DashboardHandler(http.server.SimpleHTTPRequestHandler):
except Exception as e:
self._send_json({'success': False, 'error': str(e)}, 500)
+ def _handle_codex_transcribe(self):
+ try:
+ payload = self._read_json_body()
+ result = _codex_transcribe_audio(
+ payload.get('audioBase64'),
+ filename=str(payload.get('filename') or ''),
+ mime_type=str(payload.get('mimeType') or ''),
+ language=str(payload.get('language') or 'zh'),
+ )
+ self._send_json({'success': True, **result})
+ except ValueError as e:
+ self._send_json({'success': False, 'error': str(e)}, 400)
+ except Exception as e:
+ self._send_json({'success': False, 'error': str(e)}, 500)
+
def _handle_codex_job_get(self):
try:
parsed = urlparse(self.path)