pygpt-net 2.5.88__py3-none-any.whl → 2.5.90__py3-none-any.whl

pygpt_net/CHANGELOG.txt

@@ -1,3 +1,14 @@
+2.5.90 (2025-08-07)
+
+- Fix: Initialize context summary if a conversation starts with a tool call.
+- Fix: Store splitter positions even if the object is deleted from memory.
+- Update: CSS improvements.
+
+2.5.89 (2025-08-07)
+
+- Added audio output device selection in Config -> Audio - issue #117
+- Added audio input and output backend selections in Config -> Audio.
+
 2.5.88 (2025-08-06)
 
 - Optimized the process of unloading tabs from memory.

pygpt_net/__init__.py

@@ -6,15 +6,15 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.06 00:00:00                  #
+# Updated Date: 2025.08.07 00:00:00                  #
 # ================================================== #
 
 __author__ = "Marcin Szczygliński"
 __copyright__ = "Copyright 2025, Marcin Szczygliński"
 __credits__ = ["Marcin Szczygliński"]
 __license__ = "MIT"
-__version__ = "2.5.88"
-__build__ = "2025-08-06"
+__version__ = "2.5.90"
+__build__ = "2025-08-07"
 __maintainer__ = "Marcin Szczygliński"
 __github__ = "https://github.com/szczyglis-dev/py-gpt"
 __report__ = "https://github.com/szczyglis-dev/py-gpt/issues"

pygpt_net/app.py

@@ -6,17 +6,25 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.02 03:00:00                  #
+# Updated Date: 2025.08.07 03:00:00                  #
 # ================================================== #
 
 import os
 import builtins
 import io
+import platform
 
 # disable warnings
 os.environ["TRANSFORMERS_NO_ADVISORY_WARNINGS"] = "1"
 os.environ["QT_LOGGING_RULES"] = "qt.multimedia.ffmpeg=false"
 
+if platform.system() == 'Windows':
+    # fix ffmpeg bug: [SWR] Output channel layout "" is invalid or unsupported.
+    os.environ['QT_MEDIA_BACKEND'] = 'windows'
+
+# enable debug logging
+# os.environ["QT_LOGGING_RULES"] = "*.debug=true"
+
 _original_open = builtins.open
 
 def open_wrapper(file, mode='r', *args, **kwargs):

pygpt_net/controller/access/voice.py

@@ -6,14 +6,14 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.03 14:00:00                  #
+# Updated Date: 2025.08.07 03:00:00                  #
 # ================================================== #
 
 from typing import Optional, List, Dict, Any
 
 import os
 
-from PySide6.QtCore import QTimer, Slot
+from PySide6.QtCore import QTimer, Slot, QObject
 from qasync import QApplication
 
 from pygpt_net.item.ctx import CtxItem
@@ -22,7 +22,7 @@ from pygpt_net.core.events import ControlEvent, AppEvent
 from pygpt_net.utils import trans
 
 
-class Voice:
+class Voice(QObject):
 
     MIN_FRAMES = 25  # minimum frames to start transcription
     PLAY_DELAY = 500  # ms, delay before playing audio event
@@ -33,6 +33,7 @@ class Voice:
 
         :param window: Window instance
         """
+        super().__init__()
         self.window = window
         self.is_recording = False
         self.timer = None
@@ -212,6 +213,7 @@ class Voice:
 
     def start_recording(self):
         """Start recording"""
+        self.window.controller.kernel.resume()
         # display snap warning if not displayed yet
         if (not self.window.core.config.get("audio.input.snap", False)
                 or not self.window.core.config.has("audio.input.snap")):
@@ -290,8 +292,7 @@ class Voice:
                 return
 
             if self.window.core.audio.capture.has_frames():
-                frames = self.window.core.audio.capture.get_frames()
-                if len(frames) < self.MIN_FRAMES:
+                if not self.window.core.audio.capture.has_min_frames():
                     self.window.update_status(trans("status.audio.too_short"))
                     self.window.dispatch(AppEvent(AppEvent.VOICE_CONTROL_STOPPED))  # app event
                     return

pygpt_net/controller/audio/audio.py

@@ -12,6 +12,8 @@
 import os
 from typing import Optional
 
+from qasync import QApplication
+
 from pygpt_net.core.tabs.tab import Tab
 from pygpt_net.core.events import Event, BaseEvent
 from pygpt_net.item.ctx import CtxItem
@@ -351,6 +353,7 @@ class Audio:
         :param text: text to play
         """
         self.window.update_status(trans("status.audio.start"))
+        QApplication.processEvents()  # process events to update UI
 
     def on_play(self, event: str):
         """

pygpt_net/controller/chat/output.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.05 21:00:00                  #
+# Updated Date: 2025.08.07 03:00:00                  #
 # ================================================== #
 
 import gc
@@ -226,7 +226,7 @@ class Output:
                 self.window.controller.chat.common.unlock_input()  # unlock input
 
         # handle ctx name (generate title from summary if not initialized)
-        if not reply and not internal:  # don't call if reply or internal mode
+        if not ctx.meta or not ctx.meta.initialized:  # don't call if reply or internal mode
             if self.window.core.config.get('ctx.auto_summary'):
                 self.log("Calling for prepare context name...")
                 self.window.controller.ctx.prepare_name(ctx)  # async

pygpt_net/controller/config/placeholder.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.07.30 00:00:00                  #
+# Updated Date: 2025.08.07 03:00:00                  #
 # ================================================== #
 
 from typing import Dict, Any, List
@@ -136,6 +136,12 @@ class Placeholder:
             return self.get_voice_control_actions()
         elif id == "audio_input_devices":
             return self.get_audio_input_devices()
+        elif id == "audio_output_devices":
+            return self.get_audio_output_devices()
+        elif id == "audio_input_backend":
+            return self.get_audio_input_backend()
+        elif id == "audio_output_backend":
+            return self.get_audio_output_backend()
         elif id == "audio_tts_whisper_voices":
             return self.get_audio_tts_whisper_voices()
         else:
@@ -165,6 +171,42 @@ class Placeholder:
             data.append({str(device[0]): device[1]})
         return data
 
+    def get_audio_output_devices(self) -> List[Dict[str, str]]:
+        """
+        Get audio output devices list
+
+        :return: placeholders list
+        """
+        devices = self.window.core.audio.get_output_devices()
+        data = []
+        for device in devices:
+            data.append({str(device[0]): device[1]})
+        return data
+
+    def get_audio_input_backend(self) -> List[Dict[str, str]]:
+        """
+        Get audio input backends list
+
+        :return: placeholders list
+        """
+        items = self.window.core.audio.get_input_backends()
+        data = []
+        for item in items:
+            data.append({str(item[0]): item[1]})
+        return data
+
+    def get_audio_output_backend(self) -> List[Dict[str, str]]:
+        """
+        Get audio output backends list
+
+        :return: placeholders list
+        """
+        items = self.window.core.audio.get_output_backends()
+        data = []
+        for item in items:
+            data.append({str(item[0]): item[1]})
+        return data
+
     def get_langchain_providers(self) -> List[Dict[str, str]]:
         """
         Get Langchain LLM provider placeholders list

pygpt_net/controller/ctx/ctx.py

@@ -6,9 +6,9 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.06 01:00:00                  #
+# Updated Date: 2025.08.07 03:00:00                  #
 # ================================================== #
-
+import gc
 from typing import Optional, List
 
 from PySide6.QtCore import QModelIndex, QTimer

pygpt_net/controller/kernel/kernel.py

@@ -6,10 +6,9 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.05 00:00:00                  #
+# Updated Date: 2025.08.06 19:00:00                  #
 # ================================================== #
 
-import asyncio
 import threading
 import time
 from typing import Any, Dict, Optional, Union, List
@@ -289,7 +288,7 @@ class Kernel(QObject):
         """Terminate kernel"""
         self.window.dispatch(KernelEvent(KernelEvent.TERMINATE))
         self.stop(exit=True)
-        self.window.ui.hide_loading()
+        # self.window.ui.hide_loading()
         self.window.controller.plugins.destroy()
 
     def stop(self, exit: bool = False):

pygpt_net/controller/layout/layout.py

@@ -112,8 +112,11 @@ class Layout:
             # do not save main splitter state if notepad was not opened yet
             if splitter == "calendar" and not self.window.controller.notepad.opened_once:
                 continue
-            if splitter in self.window.ui.splitters:
-                data[splitter] = self.window.ui.splitters[splitter].sizes()
+            try:
+                if splitter in self.window.ui.splitters:
+                    data[splitter] = self.window.ui.splitters[splitter].sizes()
+            except Exception as e:
+                pass
         self.window.core.config.set('layout.splitters', data)
 
     def splitters_restore(self):
@@ -151,8 +154,11 @@ class Layout:
 
         # notepads
         for id in self.window.ui.notepad:
-            scroll_id = "notepad." + str(id)
-            data[scroll_id] = self.window.ui.notepad[id].textarea.verticalScrollBar().value()
+            try:
+                scroll_id = "notepad." + str(id)
+                data[scroll_id] = self.window.ui.notepad[id].textarea.verticalScrollBar().value()
+            except Exception as e:
+                pass
         self.window.core.config.set('layout.scroll', data)
 
     def scroll_restore(self):

pygpt_net/controller/model/model.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.07.12 19:00:00                  #
+# Updated Date: 2025.08.07 02:00:00                  #
 # ================================================== #
 
 from typing import Optional
@@ -54,6 +54,19 @@ class Model:
         self.window.controller.ui.update()
         self.window.dispatch(AppEvent(AppEvent.MODEL_SELECTED))  # app event
 
+        # update model in preset
+        preset = self.window.core.config.get('preset')
+        if preset and preset != "*":
+            preset_data = self.window.core.presets.get_by_id(mode, preset)
+            if preset_data:
+                preset_data.model = model
+                self.window.core.presets.save(preset)
+
+        # update model in current ctx meta
+        self.window.core.ctx.model = model
+        self.window.core.ctx.last_model = model
+        self.window.core.ctx.update_model_in_current(model)
+
     def next(self):
         """Select next model"""
         mode = self.window.core.config.get('mode')

pygpt_net/controller/settings/editor.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.07.22 22:00:00                  #
+# Updated Date: 2025.08.07 03:00:00                  #
 # ================================================== #
 
 import copy
@@ -67,6 +67,8 @@ class Editor:
         self.window.ui.add_hook("update.config.render.code_syntax", self.hook_update)
         self.window.ui.add_hook("update.config.theme.style", self.hook_update)
         self.window.ui.add_hook("update.config.llama.idx.chat.agent.render.all", self.hook_update)
+        self.window.ui.add_hook("update.config.audio.input.backend", self.hook_update)
+        self.window.ui.add_hook("update.config.audio.output.backend", self.hook_update)
         # self.window.ui.add_hook("llama.idx.storage", self.hook_update)  # vector store update
         # self.window.ui.add_hook("update.config.llama.idx.list", self.hook_update)
 
@@ -335,6 +337,38 @@ class Editor:
             self.window.core.config.set(key, value)
             self.window.controller.debug.toggle_menu()
 
+        # audio devices
+        elif key == "audio.input.backend":
+            self.window.core.config.set(key, value)
+            items = self.window.controller.config.placeholder.get_audio_input_devices()
+            option = self.options.get("audio.input.device", {})
+            default_id, _ = self.window.core.audio.get_default_input_device()
+            if default_id is None:
+                default_id = 0
+            self.window.controller.config.update_list(option, "config", "audio.input.device", items)
+            self.window.controller.config.apply_value(
+                parent_id='config',
+                key='audio.input.device',
+                option=option,
+                value=str(default_id)  # default device,
+            )
+            self.window.core.config.set("audio.input.device", str(default_id))  # reset to default device
+        elif key == "audio.output.backend":
+            self.window.core.config.set(key, value)
+            items = self.window.controller.config.placeholder.get_audio_output_devices()
+            option = self.options.get("audio.output.device", {})
+            default_id, _ = self.window.core.audio.get_default_output_device()
+            if default_id is None:
+                default_id = 0
+            self.window.controller.config.update_list(option, "config", "audio.output.device", items)
+            self.window.controller.config.apply_value(
+                parent_id='config',
+                key='audio.output.device',
+                option=option,
+                value=str(default_id)  # default device,
+            )
+            self.window.core.config.set("audio.output.device", str(default_id))  # reset to default device
+
     def toggle_collapsed(
             self,
             id: str,

pygpt_net/controller/ui/tabs.py

@@ -6,10 +6,10 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.07.20 23:00:00                  #
+# Updated Date: 2025.08.07 18:00:00                  #
 # ================================================== #
 
-from typing import Any, Optional
+from typing import Any, Optional, Tuple
 
 from PySide6.QtCore import QTimer
 
@@ -178,6 +178,7 @@ class Tabs:
 
         prev_tab = self.current
         prev_column = self.column_idx
+
         self.current = idx
         self.column_idx = column_idx
         self.window.controller.ui.mode.update()
@@ -287,6 +288,38 @@ class Tabs:
         """
         return self.window.core.tabs.get_min_idx_by_type(type, self.column_idx)
 
+    def get_prev_idx_from(self, idx: int) -> Tuple[int, bool]:
+        """
+        Get previous tab index from given index
+
+        :param idx: tab index
+        :return: tuple of previous index and boolean indicating if it exists
+        """
+        return self.window.core.tabs.get_prev_idx_from(idx, self.column_idx)
+
+    def get_next_idx_from(self, idx: int) -> Tuple[int, bool]:
+        """
+        Get next tab index from given index
+
+        :param idx: tab index
+        :return: tuple of next index and boolean indicating if it exists
+        """
+        return self.window.core.tabs.get_next_idx_from(idx, self.column_idx)
+
+    def get_after_close_idx(self, idx: int) -> int:
+        """
+        Get tab index after closing the given index
+
+        :param idx: tab index
+        :return: previous tab index if exists, otherwise None
+        """
+        prev_idx, exists = self.get_prev_idx_from(idx)
+        if exists:
+            return prev_idx
+        next_idx, exists = self.get_next_idx_from(idx)
+        if exists:
+            return next_idx
+
     def on_column_changed(self):
         """Column changed event"""
         if self.locked:
@@ -381,7 +414,21 @@ class Tabs:
         """
         if self.locked:
             return
+
+        previous_current = self.current
+        idx_after = None  # <--- next tab index after close to switch to
+        if previous_current != idx and self.column_idx == column_idx:
+            idx_after = previous_current
+            if idx_after > idx:
+                idx_after -= 1  # if current is after closed tab, idx will be shifted
+
+        if idx_after is None:
+            idx_after = self.get_after_close_idx(idx)  # find next tab index after close
+
         self.window.core.tabs.remove_tab_by_idx(idx, column_idx)
+        if idx_after is not None:
+            self.switch_tab_by_idx(idx_after, column_idx)
+
         self.on_changed()
         self.update_current()
         self.debug()
@@ -639,7 +686,10 @@ class Tabs:
 
         :param column_idx: column index
         """
-        idx = self.get_current_idx(column_idx)
+        # append at the end of column
+        idx = self.window.core.tabs.get_max_idx_by_column(column_idx)
+        if idx == -1:
+            idx = 0
         self.append(
             type=Tab.TAB_CHAT,
             tool_id=None,

pygpt_net/controller/ui/ui.py

@@ -42,6 +42,8 @@ class UI:
             7: {'label': 'label.color.violet', 'color': QColor(238, 130, 238), 'font': QColor(255, 255, 255)},
         }
         self.stop_action = None
+        self.splitter_output_size_input = None
+        self.splitter_output_size_files = None
 
     def setup(self):
         """Setup UI"""

pygpt_net/core/audio/audio.py

@@ -6,19 +6,17 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.07.14 00:00:00                  #
+# Updated Date: 2025.08.07 03:00:00                  #
 # ================================================== #
 
 import re
 from typing import Union, Optional, Tuple, List
 
-from PySide6.QtMultimedia import QMediaDevices
-from bs4 import UnicodeDammit
-
 from pygpt_net.provider.audio_input.base import BaseProvider as InputBaseProvider
 from pygpt_net.provider.audio_output.base import BaseProvider as OutputBaseProvider
 
 from .capture import Capture
+from .output import Output
 from .whisper import Whisper
 
 
@@ -31,10 +29,11 @@ class Audio:
         """
         self.window = window
         self.capture = Capture(window)
+        self.output = Output(window)
         self.whisper = Whisper(window)
         self.providers = {
-            "input": {},
-            "output": {},
+            "input": {},  # audio transcription providers
+            "output": {}, # speech synthesis providers
         }
         self.last_error = None
 
@@ -44,37 +43,55 @@ class Audio:
 
         :return devices list: [(id, name)]
         """
-        devices = QMediaDevices.audioInputs()
-        devices_list = []
-        for index, device in enumerate(devices):
-            dammit = UnicodeDammit(device.description())
-            devices_list.append((index, dammit.unicode_markup))
-        return devices_list
-
-    def is_device_compatible(self, device_index: int) -> bool:
-        """
-        Check if device is compatible
-
-        :param device_index: device index
-        :return: True if compatible
-        """
-        import pyaudio
-        rate = int(self.window.core.config.get('audio.input.rate', 44100))
-        channels = int(self.window.core.config.get('audio.input.channels', 1))
-        p = pyaudio.PyAudio()
-        info = p.get_device_info_by_index(device_index)
-        try:
-            p.is_format_supported(
-                rate=rate,
-                input_device=info['index'],
-                input_channels=channels,
-                input_format=pyaudio.paInt16)
-            supported = True
-        except ValueError as e:
-            self.last_error = str(e)
-            supported = False
-        p.terminate()
-        return supported
+        return self.capture.get_input_devices()
+
+    def get_output_devices(self) -> List[Tuple[int, str]]:
+        """
+        Get output devices
+
+        :return devices list: [(id, name)]
+        """
+        return self.output.get_output_devices()
+
+    def get_input_backends(self) -> List[Tuple[str, str]]:
+        """
+        Get input backends
+
+        :return backends list: [(id, name)]
+        """
+        choices = []
+        choices.append(("native", "Native / QtMultimedia"))
+        choices.append(("pyaudio", "PyAudio"))
+        choices.append(("pygame", "PyGame"))
+        return choices
+
+    def get_output_backends(self) -> List[Tuple[str, str]]:
+        """
+        Get output backends
+
+        :return backends list: [(id, name)]
+        """
+        choices = []
+        choices.append(("native", "Native / QtMultimedia"))
+        choices.append(("pyaudio", "PyAudio"))
+        # choices.append(("pygame", "PyGame"))
+        return choices
+
+    def get_default_input_device(self) -> Tuple[int, str]:
+        """
+        Get default input device
+
+        :return: (id, name)
+        """
+        return self.capture.get_default_input_device()
+
+    def get_default_output_device(self) -> Tuple[int, str]:
+        """
+        Get default output device
+
+        :return: (id, name)
+        """
+        return self.output.get_default_output_device()
 
     def is_registered(self, id: str, type: str = "output") -> bool:
         """