Phase 6: Add Deepgram remote transcription (managed + BYOK modes)

New files:
- client/deepgram_transcription.py — DeepgramTranscriptionEngine with
  managed mode (proxy) and BYOK mode (direct Deepgram). Sends raw binary
  PCM audio over WebSocket, handles both proxy and Deepgram response formats.

Modified files:
- config/default_config.yaml — Replace remote_processing with new remote
  section (mode, server_url, auth_token, byok_api_key, deepgram_model, language)
- client/config.py — Add migration from old remote_processing config
- gui/settings_dialog_qt.py — Replace Remote Processing group with
  Transcription Mode section (Local/Managed/BYOK radio buttons, login/register
  dialogs, balance display, model selector)
- gui/main_window_qt.py — Select engine based on remote.mode config,
  add error and credits_low handlers

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Developer
2026-04-05 11:45:30 -07:00
parent bb8a8c251d
commit 9ff883e2e3
8 changed files with 1503 additions and 74 deletions

View File

@@ -4,7 +4,7 @@ from PySide6.QtWidgets import (
QDialog, QVBoxLayout, QHBoxLayout, QFormLayout,
QLabel, QLineEdit, QComboBox, QCheckBox, QSlider,
QPushButton, QMessageBox, QGroupBox, QScrollArea, QWidget,
QFileDialog, QColorDialog
QFileDialog, QColorDialog, QRadioButton
)
from PySide6.QtCore import Qt
from PySide6.QtGui import QScreen, QFontDatabase, QColor
@@ -487,46 +487,91 @@ class SettingsDialog(QDialog):
server_group.setLayout(server_layout)
content_layout.addWidget(server_group)
# Remote Processing Group
remote_group = QGroupBox("Remote Processing (GPU Offload)")
remote_layout = QFormLayout()
remote_layout.setSpacing(10)
# Transcription Mode Group
mode_group = QGroupBox("Transcription Mode")
mode_layout = QVBoxLayout()
mode_layout.setSpacing(10)
self.remote_enabled_check = QCheckBox()
self.remote_enabled_check.setToolTip(
"Enable remote transcription processing:\n"
"• Offload transcription to a GPU-equipped server\n"
"• Reduces local CPU/GPU usage\n"
"• Requires running the remote transcription service"
)
remote_layout.addRow("Enable Remote Processing:", self.remote_enabled_check)
# Radio buttons for mode selection
self.mode_local_radio = QRadioButton("Local (Whisper)")
self.mode_local_radio.setToolTip("Transcribe locally using Whisper models")
self.mode_managed_radio = QRadioButton("Remote - Managed")
self.mode_managed_radio.setToolTip("Use the transcription proxy service with prepaid credits")
self.mode_byok_radio = QRadioButton("Remote - BYOK (Bring Your Own Key)")
self.mode_byok_radio.setToolTip("Connect directly to Deepgram with your own API key")
self.remote_url_input = QLineEdit()
self.remote_url_input.setPlaceholderText("ws://your-server:8765/ws/transcribe")
self.remote_url_input.setToolTip(
"WebSocket URL of the remote transcription service:\n"
"• Format: ws://host:port/ws/transcribe\n"
"• Use wss:// for secure connections"
)
remote_layout.addRow("Server URL:", self.remote_url_input)
mode_layout.addWidget(self.mode_local_radio)
mode_layout.addWidget(self.mode_managed_radio)
mode_layout.addWidget(self.mode_byok_radio)
self.remote_api_key_input = QLineEdit()
self.remote_api_key_input.setEchoMode(QLineEdit.Password)
self.remote_api_key_input.setPlaceholderText("your-api-key")
self.remote_api_key_input.setToolTip(
"API key for authentication with the remote service"
)
remote_layout.addRow("API Key:", self.remote_api_key_input)
# Managed mode fields (shown when managed radio selected)
self.managed_widget = QWidget()
managed_layout = QFormLayout()
managed_layout.setSpacing(8)
self.remote_fallback_check = QCheckBox("Enable")
self.remote_fallback_check.setChecked(True)
self.remote_fallback_check.setToolTip(
"Fall back to local transcription if remote service is unavailable"
)
remote_layout.addRow("Fallback to Local:", self.remote_fallback_check)
self.managed_server_url = QLineEdit()
self.managed_server_url.setPlaceholderText("wss://your-proxy-server.com")
managed_layout.addRow("Server URL:", self.managed_server_url)
remote_group.setLayout(remote_layout)
content_layout.addWidget(remote_group)
# Login/Register buttons in a row
auth_widget = QWidget()
auth_layout = QHBoxLayout()
auth_layout.setContentsMargins(0, 0, 0, 0)
self.managed_login_btn = QPushButton("Login")
self.managed_login_btn.clicked.connect(self._managed_login)
self.managed_register_btn = QPushButton("Register")
self.managed_register_btn.clicked.connect(self._managed_register)
auth_layout.addWidget(self.managed_login_btn)
auth_layout.addWidget(self.managed_register_btn)
auth_layout.addStretch()
auth_widget.setLayout(auth_layout)
managed_layout.addRow("Account:", auth_widget)
self.managed_balance_label = QLabel("Not logged in")
managed_layout.addRow("Balance:", self.managed_balance_label)
self.managed_fallback_check = QCheckBox("Enable")
self.managed_fallback_check.setChecked(True)
self.managed_fallback_check.setToolTip("Fall back to local Whisper if remote fails")
managed_layout.addRow("Fallback to Local:", self.managed_fallback_check)
self.managed_widget.setLayout(managed_layout)
mode_layout.addWidget(self.managed_widget)
# BYOK mode fields (shown when BYOK radio selected)
self.byok_widget = QWidget()
byok_layout = QFormLayout()
byok_layout.setSpacing(8)
self.byok_api_key_input = QLineEdit()
self.byok_api_key_input.setEchoMode(QLineEdit.Password)
self.byok_api_key_input.setPlaceholderText("your-deepgram-api-key")
byok_layout.addRow("Deepgram API Key:", self.byok_api_key_input)
self.byok_model_combo = QComboBox()
self.byok_model_combo.addItems(["nova-2", "nova-2-general", "nova-2-meeting", "nova-2-phonecall", "whisper-large", "whisper-medium", "whisper-small"])
byok_layout.addRow("Model:", self.byok_model_combo)
self.byok_language_input = QLineEdit()
self.byok_language_input.setText("en-US")
self.byok_language_input.setPlaceholderText("en-US")
byok_layout.addRow("Language:", self.byok_language_input)
self.byok_fallback_check = QCheckBox("Enable")
self.byok_fallback_check.setChecked(True)
self.byok_fallback_check.setToolTip("Fall back to local Whisper if Deepgram fails")
byok_layout.addRow("Fallback to Local:", self.byok_fallback_check)
self.byok_widget.setLayout(byok_layout)
mode_layout.addWidget(self.byok_widget)
mode_group.setLayout(mode_layout)
content_layout.addWidget(mode_group)
# Connect radio buttons to show/hide relevant widgets
self.mode_local_radio.toggled.connect(self._on_mode_changed)
self.mode_managed_radio.toggled.connect(self._on_mode_changed)
self.mode_byok_radio.toggled.connect(self._on_mode_changed)
# Updates Group
updates_group = QGroupBox("Software Updates")
@@ -794,11 +839,28 @@ class SettingsDialog(QDialog):
self.server_room_input.setText(self.config.get('server_sync.room', 'default'))
self.server_passphrase_input.setText(self.config.get('server_sync.passphrase', ''))
# Remote processing settings
self.remote_enabled_check.setChecked(self.config.get('remote_processing.enabled', False))
self.remote_url_input.setText(self.config.get('remote_processing.server_url', ''))
self.remote_api_key_input.setText(self.config.get('remote_processing.api_key', ''))
self.remote_fallback_check.setChecked(self.config.get('remote_processing.fallback_to_local', True))
# Transcription mode settings
mode = self.config.get('remote.mode', 'local')
if mode == 'managed':
self.mode_managed_radio.setChecked(True)
elif mode == 'byok':
self.mode_byok_radio.setChecked(True)
else:
self.mode_local_radio.setChecked(True)
self.managed_server_url.setText(self.config.get('remote.server_url', ''))
self.managed_fallback_check.setChecked(self.config.get('remote.fallback_to_local', True))
self.byok_api_key_input.setText(self.config.get('remote.byok_api_key', ''))
self.byok_model_combo.setCurrentText(self.config.get('remote.deepgram_model', 'nova-2'))
self.byok_language_input.setText(self.config.get('remote.language', 'en-US'))
self.byok_fallback_check.setChecked(self.config.get('remote.fallback_to_local', True))
# Trigger visibility update
self._on_mode_changed()
# Update balance if managed mode and has token
if self.config.get('remote.auth_token'):
self._update_managed_balance()
# Update settings
self.update_auto_check.setChecked(self.config.get('updates.auto_check', True))
@@ -869,11 +931,21 @@ class SettingsDialog(QDialog):
self.config.set('server_sync.room', self.server_room_input.text())
self.config.set('server_sync.passphrase', self.server_passphrase_input.text())
# Remote processing settings
self.config.set('remote_processing.enabled', self.remote_enabled_check.isChecked())
self.config.set('remote_processing.server_url', self.remote_url_input.text())
self.config.set('remote_processing.api_key', self.remote_api_key_input.text())
self.config.set('remote_processing.fallback_to_local', self.remote_fallback_check.isChecked())
# Transcription mode settings
if self.mode_managed_radio.isChecked():
self.config.set('remote.mode', 'managed')
elif self.mode_byok_radio.isChecked():
self.config.set('remote.mode', 'byok')
else:
self.config.set('remote.mode', 'local')
self.config.set('remote.server_url', self.managed_server_url.text())
self.config.set('remote.fallback_to_local',
self.managed_fallback_check.isChecked() if self.mode_managed_radio.isChecked()
else self.byok_fallback_check.isChecked())
self.config.set('remote.byok_api_key', self.byok_api_key_input.text())
self.config.set('remote.deepgram_model', self.byok_model_combo.currentText())
self.config.set('remote.language', self.byok_language_input.text())
# Update settings
self.config.set('updates.auto_check', self.update_auto_check.isChecked())
@@ -892,6 +964,194 @@ class SettingsDialog(QDialog):
except Exception as e:
QMessageBox.critical(self, "Error", f"Failed to save settings:\n{e}")
def _on_mode_changed(self):
"""Show/hide mode-specific widgets based on selected radio button."""
self.managed_widget.setVisible(self.mode_managed_radio.isChecked())
self.byok_widget.setVisible(self.mode_byok_radio.isChecked())
def _managed_login(self):
"""Open a login dialog and authenticate with the managed proxy server."""
import json
import urllib.request
import urllib.error
dialog = QDialog(self)
dialog.setWindowTitle("Login")
dialog.setMinimumWidth(350)
layout = QFormLayout()
email_input = QLineEdit()
email_input.setPlaceholderText("you@example.com")
layout.addRow("Email:", email_input)
password_input = QLineEdit()
password_input.setEchoMode(QLineEdit.Password)
layout.addRow("Password:", password_input)
button_layout = QHBoxLayout()
cancel_btn = QPushButton("Cancel")
cancel_btn.clicked.connect(dialog.reject)
login_btn = QPushButton("Login")
login_btn.setDefault(True)
button_layout.addStretch()
button_layout.addWidget(cancel_btn)
button_layout.addWidget(login_btn)
layout.addRow("", button_layout)
dialog.setLayout(layout)
def do_login():
server_url = self.managed_server_url.text().rstrip('/')
if not server_url:
QMessageBox.warning(dialog, "Error", "Please enter a Server URL first.")
return
payload = json.dumps({
"email": email_input.text(),
"password": password_input.text()
}).encode('utf-8')
req = urllib.request.Request(
f"{server_url}/auth/login",
data=payload,
headers={"Content-Type": "application/json"},
method="POST"
)
try:
with urllib.request.urlopen(req, timeout=10) as resp:
data = json.loads(resp.read().decode('utf-8'))
token = data.get('token', '')
if token:
self.config.set('remote.auth_token', token)
self._update_managed_balance()
QMessageBox.information(dialog, "Success", "Logged in successfully.")
dialog.accept()
else:
QMessageBox.warning(dialog, "Error", "Login succeeded but no token received.")
except urllib.error.HTTPError as e:
try:
body = json.loads(e.read().decode('utf-8'))
msg = body.get('detail', body.get('message', str(e)))
except Exception:
msg = str(e)
QMessageBox.warning(dialog, "Login Failed", msg)
except Exception as e:
QMessageBox.warning(dialog, "Error", f"Could not connect to server:\n{e}")
login_btn.clicked.connect(do_login)
dialog.exec()
def _managed_register(self):
"""Open a registration dialog and create an account on the managed proxy server."""
import json
import urllib.request
import urllib.error
dialog = QDialog(self)
dialog.setWindowTitle("Register")
dialog.setMinimumWidth(350)
layout = QFormLayout()
email_input = QLineEdit()
email_input.setPlaceholderText("you@example.com")
layout.addRow("Email:", email_input)
password_input = QLineEdit()
password_input.setEchoMode(QLineEdit.Password)
layout.addRow("Password:", password_input)
confirm_input = QLineEdit()
confirm_input.setEchoMode(QLineEdit.Password)
layout.addRow("Confirm Password:", confirm_input)
button_layout = QHBoxLayout()
cancel_btn = QPushButton("Cancel")
cancel_btn.clicked.connect(dialog.reject)
register_btn = QPushButton("Register")
register_btn.setDefault(True)
button_layout.addStretch()
button_layout.addWidget(cancel_btn)
button_layout.addWidget(register_btn)
layout.addRow("", button_layout)
dialog.setLayout(layout)
def do_register():
if password_input.text() != confirm_input.text():
QMessageBox.warning(dialog, "Error", "Passwords do not match.")
return
server_url = self.managed_server_url.text().rstrip('/')
if not server_url:
QMessageBox.warning(dialog, "Error", "Please enter a Server URL first.")
return
payload = json.dumps({
"email": email_input.text(),
"password": password_input.text()
}).encode('utf-8')
req = urllib.request.Request(
f"{server_url}/auth/register",
data=payload,
headers={"Content-Type": "application/json"},
method="POST"
)
try:
with urllib.request.urlopen(req, timeout=10) as resp:
data = json.loads(resp.read().decode('utf-8'))
token = data.get('token', '')
if token:
self.config.set('remote.auth_token', token)
self._update_managed_balance()
QMessageBox.information(dialog, "Success", "Account created and logged in.")
dialog.accept()
else:
QMessageBox.information(dialog, "Success",
"Account created. Please log in.")
dialog.accept()
except urllib.error.HTTPError as e:
try:
body = json.loads(e.read().decode('utf-8'))
msg = body.get('detail', body.get('message', str(e)))
except Exception:
msg = str(e)
QMessageBox.warning(dialog, "Registration Failed", msg)
except Exception as e:
QMessageBox.warning(dialog, "Error", f"Could not connect to server:\n{e}")
register_btn.clicked.connect(do_register)
dialog.exec()
def _update_managed_balance(self):
"""Fetch and display the current account balance from the managed proxy server."""
import json
import urllib.request
import urllib.error
server_url = self.managed_server_url.text().rstrip('/')
token = self.config.get('remote.auth_token', '')
if not server_url or not token:
self.managed_balance_label.setText("Not logged in")
return
req = urllib.request.Request(
f"{server_url}/billing/balance",
headers={
"Authorization": f"Bearer {token}",
"Content-Type": "application/json"
},
method="GET"
)
try:
with urllib.request.urlopen(req, timeout=10) as resp:
data = json.loads(resp.read().decode('utf-8'))
balance = data.get('balance', data.get('credits', 'N/A'))
self.managed_balance_label.setText(str(balance))
except urllib.error.HTTPError as e:
if e.code == 401:
self.managed_balance_label.setText("Session expired - please login again")
self.config.set('remote.auth_token', '')
else:
self.managed_balance_label.setText("Error fetching balance")
except Exception:
self.managed_balance_label.setText("Could not connect to server")
def _check_for_updates_now(self):
"""Manually check for updates."""
from version import __version__