pygpt-net 2.6.29__py3-none-any.whl → 2.6.31__py3-none-any.whl

pygpt_net/CHANGELOG.txt

@@ -1,3 +1,18 @@
+2.6.31 (2025-09-01)
+
+- Chat with Audio mode renamed to Realtime + audio.
+- Added support for real-time audio models from OpenAI (Realtime API) and Google (Live API), featuring real-time audio integration (beta).
+- Introduced new predefined models: gpt-realtime, gpt-4o-realtime-preview, and gemini-2.5-flash-preview-native-audio-dialog.
+- Included Google Gen AI audio input and output providers in the Audio Input/Output plugins.
+- Added URL Context remote tool support in Google Gen AI.
+
+2.6.30 (2025-08-29)
+
+- Added native Google GenAI API support (beta); live audio is not supported yet (#132).
+- Added new predefined models for image generation: Google Imagen3 and Imagen4.
+- Optimized token usage in the Responses API.
+- Removed Vision mode (it is now integrated into Chat).
+
 2.6.29 (2025-08-28)
 
 - Verbose options have been moved to the Developer section in settings.

pygpt_net/__init__.py

@@ -6,15 +6,15 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.28 00:00:00                  #
+# Updated Date: 2025.09.01 00:00:00                  #
 # ================================================== #
 
 __author__ = "Marcin Szczygliński"
 __copyright__ = "Copyright 2025, Marcin Szczygliński"
 __credits__ = ["Marcin Szczygliński"]
 __license__ = "MIT"
-__version__ = "2.6.29"
-__build__ = "2025-08-28"
+__version__ = "2.6.31"
+__build__ = "2025-09-01"
 __maintainer__ = "Marcin Szczygliński"
 __github__ = "https://github.com/szczyglis-dev/py-gpt"
 __report__ = "https://github.com/szczyglis-dev/py-gpt/issues"

pygpt_net/app.py

@@ -171,10 +171,12 @@ from pygpt_net.provider.audio_input.openai_whisper import OpenAIWhisper
 from pygpt_net.provider.audio_input.openai_whisper_local import OpenAIWhisperLocal
 from pygpt_net.provider.audio_input.google_speech_recognition import GoogleSpeechRecognition
 from pygpt_net.provider.audio_input.google_cloud_speech_recognition import GoogleCloudSpeechRecognition
+from pygpt_net.provider.audio_input.google_genai import GoogleGenAIAudioInput
 from pygpt_net.provider.audio_input.bing_speech_recognition import BingSpeechRecognition
 from pygpt_net.provider.audio_output.openai_tts import OpenAITextToSpeech
 from pygpt_net.provider.audio_output.ms_azure_tts import MSAzureTextToSpeech
 from pygpt_net.provider.audio_output.google_tts import GoogleTextToSpeech
+from pygpt_net.provider.audio_output.google_genai_tts import GoogleGenAITextToSpeech
 from pygpt_net.provider.audio_output.eleven_labs import ElevenLabsTextToSpeech
 
 # web search engine providers
@@ -318,10 +320,12 @@ def run(**kwargs):
     launcher.add_audio_input(OpenAIWhisperLocal())
     launcher.add_audio_input(GoogleSpeechRecognition())
     launcher.add_audio_input(GoogleCloudSpeechRecognition())
+    launcher.add_audio_input(GoogleGenAIAudioInput())
     launcher.add_audio_input(BingSpeechRecognition())
     launcher.add_audio_output(OpenAITextToSpeech())
     launcher.add_audio_output(MSAzureTextToSpeech())
     launcher.add_audio_output(GoogleTextToSpeech())
+    launcher.add_audio_output(GoogleGenAITextToSpeech())
     launcher.add_audio_output(ElevenLabsTextToSpeech())
 
     # register custom audio providers

pygpt_net/{container.py → app_core.py}

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.08 05:00:00                  #
+# Updated Date: 2025.08.28 09:00:00                  #
 # ================================================== #
 
 from pygpt_net.config import Config
@@ -45,21 +45,21 @@ from pygpt_net.core.tokens import Tokens
 from pygpt_net.core.updater import Updater
 from pygpt_net.core.vision import Vision
 from pygpt_net.core.web import Web
-from pygpt_net.provider.gpt import Gpt
 
+from pygpt_net.provider.api import Api
 
-class Container:
+class Core:
     def __init__(self, window=None):
         """
-        Service container
+        App core
 
         :param window: Window instance
         """
         self.window = window
 
-        # core
         self.access = Access(window)
         self.agents = Agents(window)
+        self.api = Api(window)
         self.assistants = Assistants(window)
         self.attachments = Attachments(window)
         self.audio = Audio(window)
@@ -75,7 +75,6 @@ class Container:
         self.dispatcher = Dispatcher(window)
         self.experts = Experts(window)
         self.filesystem = Filesystem(window)
-        self.gpt = Gpt(window)
         self.history = History(window)
         self.idx = Idx(window)
         self.image = Image(window)

pygpt_net/controller/__init__.py

@@ -1,4 +1,4 @@
-# !/usr/bin/env python3
+#!/usr/bin/env python3
 # -*- coding: utf-8 -*-
 # ================================================== #
 # This file is a part of PYGPT package               #
@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.24 23:00:00                  #
+# Updated Date: 2025.08.30 06:00:00                  #
 # ================================================== #
 
 from pygpt_net.controller.access import Access
@@ -34,6 +34,7 @@ from pygpt_net.controller.model import Model
 from pygpt_net.controller.notepad import Notepad
 from pygpt_net.controller.painter import Painter
 from pygpt_net.controller.plugins import Plugins
+from pygpt_net.controller.realtime import Realtime
 from pygpt_net.controller.presets import Presets
 from pygpt_net.controller.settings import Settings
 from pygpt_net.controller.theme import Theme
@@ -76,6 +77,7 @@ class Controller:
         self.painter = Painter(window)
         self.plugins = Plugins(window)
         self.presets = Presets(window)
+        self.realtime = Realtime(window)
         self.settings = Settings(window)
         self.theme = Theme(window)
         self.tools = Tools(window)
@@ -108,6 +110,7 @@ class Controller:
         self.attachment.setup()
         self.camera.setup_ui()
         self.access.setup()
+        self.realtime.setup()
 
     def post_setup(self):
         """Post-setup, after plugins are loaded"""

pygpt_net/controller/access/control.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2024.12.14 08:00:00                  #
+# Updated Date: 2025.08.28 09:00:00                  #
 # ================================================== #
 
 import re
@@ -14,16 +14,8 @@ import re
 from PySide6.QtCore import QTimer
 
 from pygpt_net.core.types import (
-    MODE_AGENT,
-    MODE_AGENT_LLAMA,
-    MODE_ASSISTANT,
     MODE_CHAT,
-    MODE_COMPLETION,
-    MODE_EXPERT,
-    MODE_LANGCHAIN,
     MODE_LLAMA_INDEX,
-    MODE_VISION, 
-    MODE_IMAGE,
 )
 from pygpt_net.core.tabs.tab import Tab
 from pygpt_net.core.events import ControlEvent

pygpt_net/controller/assistant/assistant.py

@@ -91,7 +91,7 @@ class Assistant:
 
                 # get status
                 try:
-                    status = self.window.core.gpt.assistants.run_stop(ctx)
+                    status = self.window.core.api.openai.assistants.run_stop(ctx)
                     if status == "cancelling" or status == "cancelled":
                         print("Run has been canceled.")
                         self.threads.log("Run status: {}".format(status))
@@ -227,7 +227,7 @@ class Assistant:
         assistant = self.window.core.assistants.create()
         self.editor.assign_data(assistant)
         try:
-            return self.window.core.gpt.assistants.create(assistant)
+            return self.window.core.api.openai.assistants.create(assistant)
         except Exception as e:
             self.window.core.debug.log(e)
             self.window.ui.dialogs.alert(e)
@@ -240,7 +240,7 @@ class Assistant:
         """
         self.editor.assign_data(assistant)
         try:
-            return self.window.core.gpt.assistants.update(assistant)
+            return self.window.core.api.openai.assistants.update(assistant)
         except Exception as e:
             self.window.core.debug.log(e)
             self.window.ui.dialogs.alert(e)
@@ -300,7 +300,7 @@ class Assistant:
 
                     # delete in API
                     try:
-                        self.window.core.gpt.assistants.delete(id)
+                        self.window.core.api.openai.assistants.delete(id)
                     except Exception as e:
                         self.window.ui.dialogs.alert(e)
 

pygpt_net/controller/assistant/batch.py

@@ -42,7 +42,7 @@ class Batch:
 
         # run asynchronous
         self.window.update_status("Importing assistants...please wait...")
-        self.window.core.gpt.assistants.importer.import_assistants()
+        self.window.core.api.openai.assistants.importer.import_assistants()
 
     def import_stores(self, force: bool = False):
         """
@@ -60,7 +60,7 @@ class Batch:
         # run asynchronous
         self.window.update_status("Importing vector stores...please wait...")
         self.window.core.assistants.store.truncate()  # clear all stores
-        self.window.core.gpt.assistants.importer.import_vector_stores()
+        self.window.core.api.openai.assistants.importer.import_vector_stores()
         self.window.controller.assistant.files.update()
         self.window.controller.assistant.store.update()
 
@@ -140,7 +140,7 @@ class Batch:
         # run asynchronous
         self.window.update_status("Removing files...please wait...")
         QApplication.processEvents()
-        self.window.core.gpt.assistants.importer.truncate_files()  # remove all files from API
+        self.window.core.api.openai.assistants.importer.truncate_files()  # remove all files from API
 
     def truncate_store_files_by_idx(self, idx: int, force: bool = False):
         """
@@ -173,7 +173,7 @@ class Batch:
         # run asynchronous
         self.window.update_status("Removing files...please wait...")
         QApplication.processEvents()
-        self.window.core.gpt.assistants.importer.truncate_files(store_id)  # remove all files from API
+        self.window.core.api.openai.assistants.importer.truncate_files(store_id)  # remove all files from API
 
     def clear_store_files_by_idx(
             self,
@@ -277,7 +277,7 @@ class Batch:
         self.window.update_status("Removing vector stores...please wait...")
         QApplication.processEvents()
         self.window.core.assistants.store.truncate()  # clear all stores
-        self.window.core.gpt.assistants.importer.truncate_vector_stores()
+        self.window.core.api.openai.assistants.importer.truncate_vector_stores()
         self.window.controller.assistant.files.update()
         self.window.controller.assistant.store.update()
         self.window.controller.assistant.store.current = None
@@ -298,7 +298,7 @@ class Batch:
             return
         self.window.update_status("Refreshing vector stores...please wait...")
         QApplication.processEvents()
-        self.window.core.gpt.assistants.importer.refresh_vector_stores()
+        self.window.core.api.openai.assistants.importer.refresh_vector_stores()
 
     def handle_imported_assistants(self, num: int):
         """
@@ -539,7 +539,7 @@ class Batch:
         store_id = self.window.controller.assistant.store.current
         self.window.update_status("Uploading files...please wait...")
         QApplication.processEvents()
-        self.window.core.gpt.assistants.importer.upload_files(store_id, self.files_to_upload)
+        self.window.core.api.openai.assistants.importer.upload_files(store_id, self.files_to_upload)
         self.files_to_upload = []  # clear files
 
     def handle_uploaded_files(self, num: int):

pygpt_net/controller/assistant/files.py

@@ -75,7 +75,7 @@ class Files:
         """
         # run asynchronous
         self.window.update_status("Importing files...please wait...")
-        self.window.core.gpt.assistants.importer.import_files(store_id)
+        self.window.core.api.openai.assistants.importer.import_files(store_id)
 
     def download(self, idx: int):
         """
@@ -276,7 +276,7 @@ class Files:
                     continue
 
                 # upload local attachment file and get new ID (file_id)
-                new_id = self.window.core.gpt.store.upload(
+                new_id = self.window.core.api.openai.store.upload(
                     attachment.path,
                 )
                 if new_id is not None:
@@ -287,14 +287,14 @@ class Files:
 
                     """
                     if assistant.vector_store is None or assistant.vector_store == "":
-                        assistant.vector_store = self.window.core.gpt.store.create_store(
+                        assistant.vector_store = self.window.core.api.openai.store.create_store(
                             "thread-" + thread_id,
                         )
                     """
 
                     # add to vector store if defined in assistant, otherwise file will be added to thread store
                     if assistant.vector_store:
-                        self.window.core.gpt.store.add_file(
+                        self.window.core.api.openai.store.add_file(
                             assistant.vector_store,
                             new_id,
                         )

pygpt_net/controller/assistant/threads.py

@@ -42,7 +42,7 @@ class Threads(QObject):
 
         :return: thread id
         """
-        thread_id = self.window.core.gpt.assistants.thread_create()
+        thread_id = self.window.core.api.openai.assistants.thread_create()
         self.window.core.config.set('assistant_thread', thread_id)
         self.window.core.ctx.append_thread(thread_id)
         return thread_id
@@ -245,7 +245,7 @@ class Threads(QObject):
 
         :param ctx: CtxItem
         """
-        data = self.window.core.gpt.assistants.msg_list(ctx.thread)
+        data = self.window.core.api.openai.assistants.msg_list(ctx.thread)
         for msg in data:
             if msg.role == "assistant":
                 try:
@@ -640,7 +640,7 @@ class RunWorker(QRunnable):
             while self.check \
                     and not self.window.is_closing \
                     and not self.window.controller.assistant.threads.stop:
-                run = self.window.core.gpt.assistants.run_get(self.ctx)
+                run = self.window.core.api.openai.assistants.run_get(self.ctx)
                 status = None
                 if run is not None:
                     status = run.status

pygpt_net/controller/attachment/attachment.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.23 15:00:00                  #
+# Updated Date: 2025.08.28 09:00:00                  #
 # ================================================== #
 
 import os
@@ -17,9 +17,6 @@ from urllib.parse import urlparse
 from PySide6.QtGui import QImage
 from PySide6.QtWidgets import QFileDialog, QApplication
 
-from pygpt_net.core.types import (
-    MODE_VISION,
-)
 from pygpt_net.core.events import AppEvent, KernelEvent
 from pygpt_net.item.attachment import AttachmentItem
 from pygpt_net.item.ctx import CtxItem
@@ -73,7 +70,7 @@ class Attachment:
         if not self.has(mode):
             self.window.controller.chat.vision.unavailable()
         else:
-            if mode == MODE_VISION or self.window.controller.plugins.is_type_enabled('vision'):
+            if self.window.controller.chat.vision.allowed():
                 self.window.controller.chat.vision.available()
 
         # update tokens counter (vision plugin, etc.)
@@ -480,7 +477,7 @@ class Attachment:
         """
         try:
             # get file info from assistant API
-            data = self.window.core.gpt.store.get_file(file_id)
+            data = self.window.core.api.openai.store.get_file(file_id)
             if data is None:
                 return
 
@@ -506,7 +503,7 @@ class Attachment:
                 path = self.get_download_path(filename)
 
             # download file
-            self.window.core.gpt.store.download(
+            self.window.core.api.openai.store.download(
                 file_id=file_id,
                 path=path,
             )

pygpt_net/controller/audio/audio.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.27 07:00:00                  #
+# Updated Date: 2025.08.31 23:00:00                  #
 # ================================================== #
 
 import os
@@ -43,6 +43,30 @@ class Audio:
         if self.window.core.config.get("audio.input.continuous", False):
             self.window.ui.plugin_addon['audio.input.btn'].continuous.setChecked(True)
 
+        if self.window.core.config.get("audio.input.auto_turn", False):
+            self.window.ui.nodes['audio.auto_turn'].box.setChecked(True)
+
+    def execute_input_stop(self):
+        """Execute input stop (from UI)"""
+        self.window.dispatch(Event(Event.AUDIO_INPUT_RECORD_TOGGLE, {
+            "state": False,
+            "auto": True,  # do not emit manual event
+        }))
+
+    def is_recording(self) -> bool:
+        """
+        Check if audio input is recording
+
+        :return: True if recording
+        """
+        return self.window.core.plugins.get("audio_input").is_recording()
+
+    def toggle_auto_turn(self):
+        """Toggle auto turn setting"""
+        value = self.window.ui.nodes['audio.auto_turn'].box.isChecked()
+        self.window.core.config.set("audio.input.auto_turn", value)
+        self.window.core.config.save()
+
     def toggle_input(
             self,
             state: bool,

pygpt_net/controller/audio/ui.py

@@ -183,7 +183,7 @@ class UI:
         """
         self.recording = True
         self.window.ui.nodes['input'].set_icon_state("mic", True)
-        if mode == "input":
+        if mode in ["input", "realtime"]:
             self.window.controller.chat.common.lock_input()
             return
         btn = self.get_input_btn() if mode == 'input' else self.get_input_control_btn()
@@ -198,7 +198,7 @@ class UI:
         """
         self.recording = False
         self.window.ui.nodes['input'].set_icon_state("mic", False)
-        if mode == "input":
+        if mode in ["input", "realtime"]:
             self.window.controller.chat.common.unlock_input()
             return
         btn = self.get_input_btn() if mode == 'input' else self.get_input_control_btn()

pygpt_net/controller/chat/audio.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2024.12.14 18:00:00                  #
+# Updated Date: 2025.08.30 06:00:00                  #
 # ================================================== #
 
 import base64
@@ -40,19 +40,12 @@ class Audio:
         """Update input/output audio"""
         mode = self.window.core.config.get("mode")
         if mode == MODE_AUDIO:
-            if not self.window.controller.audio.is_output_enabled():
-                self.window.controller.audio.enable_output()
-                self.tmp_output = True
-            else:
-                self.tmp_output = False
             if not self.window.controller.audio.is_input_enabled():
                 self.window.controller.audio.enable_input()
                 self.tmp_input = True
             else:
                 self.tmp_input = False
         else:
-            if self.tmp_output:
-                self.window.controller.audio.disable_output()
             if self.tmp_input:
                 self.window.controller.audio.disable_input()
 

pygpt_net/controller/chat/common.py

@@ -10,12 +10,13 @@
 # ================================================== #
 
 import os
+from typing import Any
 
 from PySide6.QtGui import QTextCursor
 from PySide6.QtWidgets import QFileDialog, QApplication
 
 from pygpt_net.core.events import Event, AppEvent, RenderEvent, KernelEvent
-from pygpt_net.core.types import MODE_ASSISTANT
+from pygpt_net.core.types import MODE_ASSISTANT, MODE_AUDIO
 from pygpt_net.item.ctx import CtxItem
 from pygpt_net.item.model import ModelItem
 from pygpt_net.utils import trans
@@ -119,6 +120,17 @@ class Common:
         else:
             self.window.ui.config['global']['img_raw'].setChecked(False)
 
+        # image resolution
+        resolution = self.window.core.config.get('img_resolution', '1024x1024')
+        self.window.controller.config.apply_value(
+            parent_id="global",
+            key="img_resolution",
+            option=self.window.core.image.get_resolution_option(),
+            value=resolution,
+        )
+        if not self.initialized:
+            self.window.ui.add_hook("update.global.img_resolution", self.hook_update)
+
         # set focus to input
         self.window.ui.nodes['input'].setFocus()
         self.initialized = True
@@ -257,7 +269,7 @@ class Common:
             self.window.controller.access.voice.stop_recording(timeout=True)
 
         if self.window.core.plugins.get("audio_input").handler_simple.is_recording:
-            self.window.core.plugins.get("audio_input").handler_simple.stop_recording(timeout=False)
+            self.window.dispatch(Event(Event.AUDIO_INPUT_RECORD_TOGGLE))
             return
 
         # stop audio output if playing
@@ -275,7 +287,8 @@ class Common:
         """
         # don't unlock input and leave stop btn if assistant mode or if agent/autonomous is enabled
         # send btn will be unlocked in agent mode on stop
-        if self.can_unlock(ctx):
+        mode = self.window.core.config.get('mode')
+        if self.can_unlock(ctx) and mode != MODE_AUDIO:
             if not self.window.controller.kernel.stopped():
                 self.unlock_input()  # unlock input
                 return True
@@ -307,7 +320,7 @@ class Common:
         controller.kernel.halt = True
         dispatch(RenderEvent(RenderEvent.TOOL_END))  # show waiting
 
-        core.gpt.stop()
+        core.api.openai.stop()
         self.unlock_input()
 
         controller.chat.input.generating = False
@@ -452,6 +465,19 @@ class Common:
         else:
             self.img_enable_raw()
 
+    def hook_update(self, key: str, value: Any, caller, *args, **kwargs):
+        """
+        Hook for updating image resolution
+
+        :param key: config key
+        :param value: new value
+        :param caller: caller object
+        """
+        if key == "img_resolution":
+            if not value:
+                return
+            self.window.core.config.set('img_resolution', value)
+
     def save_text(
             self,
             text: str,