pygpt-net 2.6.28__py3-none-any.whl → 2.6.30__py3-none-any.whl

pygpt_net/provider/{gpt → api/openai}/chat.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.02 20:00:00                  #
+# Updated Date: 2025.08.28 09:00:00                  #
 # ================================================== #
 
 import json
@@ -16,7 +16,11 @@ from typing import Optional, Dict, Any, List
 from pygpt_net.core.types import (
     MODE_CHAT,
     MODE_VISION,
-    MODE_AUDIO, MULTIMODAL_IMAGE, MODE_RESEARCH, MODE_COMPLETION, OPENAI_DISABLE_TOOLS,
+    MODE_AUDIO,
+    MULTIMODAL_IMAGE,
+    MODE_RESEARCH,
+    MODE_COMPLETION,
+    OPENAI_DISABLE_TOOLS,
 )
 from pygpt_net.core.bridge.context import BridgeContext, MultimodalContext
 from pygpt_net.item.ctx import CtxItem
@@ -66,7 +70,7 @@ class Chat:
         user_name = ctx.input_name  # from ctx
         ai_name = ctx.output_name  # from ctx
 
-        client = self.window.core.gpt.get_client(mode, context.model)
+        client = self.window.core.api.openai.get_client(mode, context.model)
 
         # build chat messages
         messages = self.build(
@@ -94,7 +98,7 @@ class Chat:
         response_kwargs = {}
 
         # tools / functions
-        tools = self.window.core.gpt.tools.prepare(model, functions)
+        tools = self.window.core.api.openai.tools.prepare(model, functions)
 
         # fix: o1 compatibility
         if (model.id is not None
@@ -138,6 +142,9 @@ class Chat:
                 "format": "wav"
             }
 
+        if stream:
+            response_kwargs['stream_options'] = {"include_usage": True}
+
         response = client.chat.completions.create(
             messages=messages,
             model=model.id,
@@ -188,7 +195,7 @@ class Chat:
         max_ctx_tokens = self.window.core.config.get('max_total_tokens')  # max context window
 
         # fit to max model tokens
-        if max_ctx_tokens > model.ctx and model.ctx > 0:
+        if max_ctx_tokens > model.ctx > 0:
             max_ctx_tokens = model.ctx
 
         # input tokens: reset
@@ -319,12 +326,12 @@ class Chat:
         if not is_tool_output:  # append current prompt only if not tool output
             content = str(prompt)
             if model.is_image_input():
-                content = self.window.core.gpt.vision.build_content(
+                content = self.window.core.api.openai.vision.build_content(
                     content=content,
                     attachments=attachments,
                 )
             if model.is_audio_input():
-                content = self.window.core.gpt.audio.build_content(
+                content = self.window.core.api.openai.audio.build_content(
                     content=content,
                     multimodal_ctx=multimodal_ctx,
                 )
@@ -367,7 +374,7 @@ class Chat:
             output = response.choices[0].text.strip()
         elif mode in [
             MODE_CHAT,
-            MODE_VISION,
+            # MODE_VISION,
             MODE_RESEARCH,
         ]:
             if response.choices[0]:

pygpt_net/provider/{gpt → api/openai}/completion.py

@@ -76,7 +76,7 @@ class Completion:
         if user_name is not None and user_name != '':
             stop = [user_name + ':']
 
-        client = self.window.core.gpt.get_client()
+        client = self.window.core.api.openai.get_client()
 
         # fix for deprecated OpenAI davinci models
         if model_id.startswith('text-davinci'):

pygpt_net/provider/{gpt → api/openai}/image.py

@@ -74,7 +74,7 @@ class Image:
         # worker
         self.worker = ImageWorker()
         self.worker.window = self.window
-        self.worker.client = self.window.core.gpt.get_client()
+        self.worker.client = self.window.core.api.openai.get_client()
         self.worker.ctx = ctx
         self.worker.mode = sub_mode  # mode can be "generate" or "edit"
         self.worker.attachments = attachments  # attachments for edit mode

pygpt_net/provider/{gpt → api/openai}/remote_tools.py

@@ -113,7 +113,7 @@ class RemoteTools:
         # extend local tools with remote tools
         if enabled["computer_use"]:
             if not model.id in OPENAI_REMOTE_TOOL_DISABLE_COMPUTER_USE:
-                tools.append(self.window.core.gpt.computer.get_tool())
+                tools.append(self.window.core.api.openai.computer.get_tool())
         else:
             if not model.id in OPENAI_REMOTE_TOOL_DISABLE_WEB_SEARCH:
                 if enabled["web_search"]:

pygpt_net/provider/{gpt → api/openai}/responses.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.05 00:00:00                  #
+# Updated Date: 2025.08.28 09:00:00                  #
 # ================================================== #
 
 import base64
@@ -92,7 +92,7 @@ class Responses:
         user_name = ctx.input_name  # from ctx
         ai_name = ctx.output_name  # from ctx
 
-        client = self.window.core.gpt.get_client(mode, model)
+        client = self.window.core.api.openai.get_client(mode, model)
 
         # build chat messages
         messages = self.build(
@@ -122,7 +122,7 @@ class Responses:
         response_kwargs = {}
 
         # tools / functions
-        tools = self.window.core.gpt.tools.prepare_responses_api(model, functions)
+        tools = self.window.core.api.openai.tools.prepare_responses_api(model, functions)
 
         # extra arguments, o3 only
         if model.extra and "reasoning_effort" in model.extra:
@@ -130,7 +130,7 @@ class Responses:
             response_kwargs['reasoning']['effort'] = model.extra["reasoning_effort"]
 
         # append remote tools
-        tools = self.window.core.gpt.remote_tools.append_to_tools(
+        tools = self.window.core.api.openai.remote_tools.append_to_tools(
             mode=mode,
             model=model,
             stream=stream,
@@ -250,20 +250,31 @@ class Responses:
                 used_tokens,
                 max_ctx_tokens,
             )
+
+            has_response_id_in_last_item = False
+            if items and len(items) > 0:
+                last_item = items[-1]
+                if last_item and last_item.msg_id:
+                    has_response_id_in_last_item = True
+
             for item in items:
                 # input
                 if item.final_input is not None and item.final_input != "":
-                    messages.append({
-                        "role": "user",
-                        "content": item.final_input,
-                    })
+                    if not has_response_id_in_last_item:
+                        messages.append({
+                            "role": "user",
+                            "content": item.final_input,
+                        })
 
                 # output
                 if item.final_output is not None and item.final_output != "":
-                    msg = {
-                        "role": "assistant",
-                        "content": item.final_output,
-                    }
+                    if not has_response_id_in_last_item:
+                        msg = {
+                            "role": "assistant",
+                            "content": item.final_output,
+                        }
+                    else:
+                        msg = {}
                     # append previous audio ID
                     if MODE_AUDIO in model.mode:
                         if item.audio_id:
@@ -281,7 +292,9 @@ class Responses:
                                 msg["audio"] = {
                                     "id": self.audio_prev_id
                                 }
-                    messages.append(msg)
+
+                    if msg:
+                        messages.append(msg)
 
                     # ---- tool output ----
                     is_tool_output = False  # reset tool output flag
@@ -340,7 +353,7 @@ class Responses:
 
                                     # computer call output
                                     elif output_type == "computer_call":
-                                        base64img = self.window.core.gpt.vision.get_attachment(attachments)
+                                        base64img = self.window.core.api.openai.vision.get_attachment(attachments)
                                         if base64img and "call_id" in tool_call:
                                             if tool_call["call_id"]:
                                                 # tool output
@@ -382,13 +395,13 @@ class Responses:
             if (model.is_image_input()
                     and mode != MODE_COMPUTER
                     and not model.id.startswith("computer-use")):
-                content = self.window.core.gpt.vision.build_content(
+                content = self.window.core.api.openai.vision.build_content(
                     content=content,
                     attachments=attachments,
                     responses_api=True,
                 )
             if model.is_audio_input():
-                content = self.window.core.gpt.audio.build_content(
+                content = self.window.core.api.openai.audio.build_content(
                     content=content,
                     multimodal_ctx=multimodal_ctx,
                 )
@@ -404,6 +417,7 @@ class Responses:
             messages,
             model.id,
         )
+
         return messages
 
     def reset_tokens(self):
@@ -431,7 +445,7 @@ class Responses:
 
         if mode in [
             MODE_CHAT,
-            MODE_VISION,
+            # MODE_VISION,
             MODE_RESEARCH,
             MODE_COMPUTER,
         ]:
@@ -499,7 +513,7 @@ class Responses:
                 id = output.id
                 call_id = output.call_id
                 action = output.action
-                tool_calls, is_call = self.window.core.gpt.computer.handle_action(
+                tool_calls, is_call = self.window.core.api.openai.computer.handle_action(
                     id=id,
                     call_id=call_id,
                     action=action,
@@ -567,7 +581,7 @@ class Responses:
         if files:
             self.window.core.debug.info("[chat] Container files found, downloading...")
             try:
-                self.window.core.gpt.container.download_files(ctx, files)
+                self.window.core.api.openai.container.download_files(ctx, files)
             except Exception as e:
                 self.window.core.debug.error(f"[chat] Error downloading container files: {e}")
 
@@ -631,7 +645,7 @@ class Responses:
         if files:
             self.window.core.debug.info("[chat] Container files found, downloading...")
             try:
-                self.window.core.gpt.container.download_files(ctx, files)
+                self.window.core.api.openai.container.download_files(ctx, files)
             except Exception as e:
                 self.window.core.debug.error(f"[chat] Error downloading container files: {e}")
 

pygpt_net/provider/{gpt → api/openai}/store.py

@@ -30,7 +30,7 @@ class Store:
 
         :return: OpenAI client
         """
-        return self.window.core.gpt.get_client()
+        return self.window.core.api.openai.get_client()
 
     def log(
             self,
@@ -92,7 +92,7 @@ class Store:
         :param file_id: file ID
         :param path: path to save file
         """
-        client = self.window.core.gpt.get_client()
+        client = self.window.core.api.openai.get_client()
         content = client.files.content(file_id)
         data = content.read()
         with open(path, 'wb', ) as f:

pygpt_net/provider/{gpt → api/openai}/vision.py

@@ -54,7 +54,7 @@ class Vision:
         attachments = context.attachments
         model = context.model
         model_id = model.id
-        client = self.window.core.gpt.get_client()
+        client = self.window.core.api.openai.get_client()
 
         # extra API kwargs
         response_kwargs = {}

pygpt_net/provider/{gpt → api/openai}/worker/assistants.py

@@ -565,7 +565,7 @@ class Worker(QRunnable):
         """
         try:
             if self.stream:  # stream mode
-                run = self.window.core.gpt.assistants.run_create_stream(
+                run = self.window.core.api.openai.assistants.run_create_stream(
                     self.signals,
                     self.ctx,
                     self.thread_id,
@@ -575,7 +575,7 @@ class Worker(QRunnable):
                 )
             else:
                 # not stream mode
-                run = self.window.core.gpt.assistants.run_create(
+                run = self.window.core.api.openai.assistants.run_create(
                     self.thread_id,
                     self.assistant_id,
                     self.model,
@@ -596,7 +596,7 @@ class Worker(QRunnable):
         :return: result
         """
         try:
-            response = self.window.core.gpt.assistants.msg_send(
+            response = self.window.core.api.openai.assistants.msg_send(
                 self.thread_id,
                 self.prompt,
                 self.file_ids,
@@ -615,7 +615,7 @@ class Worker(QRunnable):
         :return: result
         """
         try:
-            run = self.window.core.gpt.assistants.run_submit_tool(self.ctx, self.tools_outputs)
+            run = self.window.core.api.openai.assistants.run_submit_tool(self.ctx, self.tools_outputs)
             if run is not None:
                 self.ctx.run_id = run.id  # update run id
                 self.signals.finished.emit(self.ctx, run, False)  # continue status check

pygpt_net/provider/{gpt → api/openai}/worker/importer.py

@@ -237,7 +237,7 @@ class ImportWorker(QRunnable):
             self.log("Importing assistants...")
             self.window.core.assistants.clear()
             items = self.window.core.assistants.get_all()
-            self.window.core.gpt.assistants.import_all(items, callback=self.callback)
+            self.window.core.api.openai.assistants.import_all(items, callback=self.callback)
             self.window.core.assistants.items = items
             self.window.core.assistants.save()
 
@@ -266,7 +266,7 @@ class ImportWorker(QRunnable):
             self.log("Importing vector stores...")
             self.window.core.assistants.store.clear()
             items = {}
-            self.window.core.gpt.store.import_stores(items, callback=self.callback)
+            self.window.core.api.openai.store.import_stores(items, callback=self.callback)
             self.window.core.assistants.store.import_items(items)
             if not silent:
                 self.signals.finished.emit("vector_stores", self.store_id, len(items))
@@ -285,7 +285,7 @@ class ImportWorker(QRunnable):
         """
         try:
             self.log("Truncating stores...")
-            num = self.window.core.gpt.store.remove_all(callback=self.callback)
+            num = self.window.core.api.openai.store.remove_all(callback=self.callback)
             self.window.core.assistants.store.items = {}
             self.window.core.assistants.store.save()
             if not silent:
@@ -336,12 +336,12 @@ class ImportWorker(QRunnable):
                 self.log("Truncating all files...")
                 self.window.core.assistants.files.truncate() # clear all files
                 # remove all files in API
-                num = self.window.core.gpt.store.remove_files(callback=self.callback)
+                num = self.window.core.api.openai.store.remove_files(callback=self.callback)
             else:
                 self.log("Truncating files for store: {}".format(self.store_id))
                 self.window.core.assistants.files.truncate(self.store_id)  # clear store files, remove from stores / DB
                 # remove store files in API
-                num = self.window.core.gpt.store.remove_store_files(
+                num = self.window.core.api.openai.store.remove_store_files(
                     self.store_id,
                     callback=self.callback,
                 )
@@ -365,14 +365,14 @@ class ImportWorker(QRunnable):
             self.log("Uploading files...")
             for file in self.files:
                 try:
-                    file_id = self.window.core.gpt.store.upload(file)
+                    file_id = self.window.core.api.openai.store.upload(file)
                     if file_id is not None:
-                        stored_file = self.window.core.gpt.store.add_file(
+                        stored_file = self.window.core.api.openai.store.add_file(
                             self.store_id,
                             file_id,
                         )
                         if stored_file is not None:
-                            data = self.window.core.gpt.store.get_file(file_id)
+                            data = self.window.core.api.openai.store.get_file(file_id)
                             self.window.core.assistants.files.insert(self.store_id, data)  # insert to DB
                             msg = "Uploaded file: {}/{}".format((num + 1), len(self.files))
                             self.signals.status.emit("upload_files", msg)
@@ -403,11 +403,11 @@ class ImportWorker(QRunnable):
             if self.store_id is None:
                 self.log("Importing all files...")
                 self.window.core.assistants.files.truncate_local()  # clear local DB (all)
-                num = self.window.core.gpt.store.import_stores_files(self.callback)  # import all files
+                num = self.window.core.api.openai.store.import_stores_files(self.callback)  # import all files
             else:
                 self.log("Importing files for store: {}".format(self.store_id))
                 self.window.core.assistants.files.truncate_local(self.store_id)  # clear local DB (all)
-                items = self.window.core.gpt.store.import_store_files(
+                items = self.window.core.api.openai.store.import_store_files(
                     self.store_id,
                     [],
                     callback=self.callback,

pygpt_net/provider/audio_input/openai_whisper.py

@@ -43,7 +43,7 @@ class OpenAIWhisper(BaseProvider):
         :param path: path to audio file to transcribe
         :return: transcribed text
         """
-        client = self.plugin.window.core.gpt.get_client()
+        client = self.plugin.window.core.api.openai.get_client()
         with open(path, "rb") as audio_file:
             return client.audio.transcriptions.create(
                 model=self.plugin.get_option_value('whisper_model'),

pygpt_net/provider/audio_output/google_tts.py

@@ -58,6 +58,18 @@ class GoogleTextToSpeech(BaseProvider):
                 "Voices": "https://cloud.google.com/text-to-speech/docs/voices"
             },
         )
+        self.plugin.add_option(
+            "google_voice_native",
+            type="text",
+            value="Kore",
+            label="Voice (Gemini API)",
+            tab="google",
+            description="Specify voice for Gemini API (supported voices may differ)",
+            tooltip="Voice name",
+            urls={
+                "Voices": "https://ai.google.dev/gemini-api/docs/speech-generation"
+            },
+        )
         self.plugin.add_option(
             "google_lang",
             type="text",

pygpt_net/provider/audio_output/openai_tts.py

@@ -60,7 +60,7 @@ class OpenAITextToSpeech(BaseProvider):
         :param text: text to speech
         :return: path to generated audio file or None if audio playback is handled here
         """
-        client = self.plugin.window.core.gpt.get_client()
+        client = self.plugin.window.core.api.openai.get_client()
         output_file = self.plugin.output_file
         voice = self.plugin.get_option_value('openai_voice')
         model = self.plugin.get_option_value('openai_model')

pygpt_net/provider/core/config/patch.py

@@ -2355,6 +2355,17 @@ class Patch:
                 self.window.core.updater.patch_css('web-chatgpt_wide.light.css', True)
                 updated = True
 
+            # < 2.6.30
+            if old < parse_version("2.6.30"):
+                print("Migrating config from < 2.6.30...")
+                if "api_native_google" not in data:
+                    data["api_native_google"] = True
+                if "remote_tools.google.web_search" not in data:
+                    data["remote_tools.google.web_search"] = True
+                if "remote_tools.google.code_interpreter" not in data:
+                    data["remote_tools.google.code_interpreter"] = False
+                updated = True
+
         # update file
         migrated = False
         if updated:

pygpt_net/provider/core/model/patch.py

@@ -763,6 +763,15 @@ class Patch:
                             model.mode.append(MODE_AGENT_OPENAI)
                 updated = True
 
+            # <  2.6.30 <--- add Google Imagen models
+            if old < parse_version("2.6.30"):
+                print("Migrating models from < 2.6.30...")
+                if "imagen-3.0-generate-002" not in data:
+                    data["imagen-3.0-generate-002"] = base_data["imagen-3.0-generate-002"]
+                if "imagen-4.0-generate-001" not in data:
+                    data["imagen-4.0-generate-001"] = base_data["imagen-4.0-generate-001"]
+                updated = True
+
         # update file
         if updated:
             data = dict(sorted(data.items()))

pygpt_net/provider/core/preset/json_file.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 json
@@ -192,7 +192,7 @@ class JsonFileProvider(BaseProvider):
             MODE_CHAT: item.chat,
             MODE_COMPLETION: item.completion,
             MODE_IMAGE: item.img,
-            MODE_VISION: item.vision,
+            # MODE_VISION: item.vision,
             # MODE_LANGCHAIN: item.langchain,
             MODE_ASSISTANT: item.assistant,
             MODE_LLAMA_INDEX: item.llama_index,
@@ -250,8 +250,6 @@ class JsonFileProvider(BaseProvider):
             item.llama_index = data[MODE_LLAMA_INDEX]
         if MODE_RESEARCH in data:
             item.research = data[MODE_RESEARCH]
-        if MODE_VISION in data:
-            item.vision = data[MODE_VISION]
 
         if 'agent_provider' in data:
             item.agent_provider = data['agent_provider']

pygpt_net/provider/llms/anthropic.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.26 19:00:00                  #
+# Updated Date: 2025.08.28 09:00:00                  #
 # ================================================== #
 
 from typing import List, Dict, Optional
@@ -93,10 +93,7 @@ class AnthropicLLM(BaseLLM):
         :param window: window instance
         :return: list of models
         """
-        import anthropic
-        client = anthropic.Anthropic(
-            api_key=window.core.config.get('api_key_anthropic', "")
-        )
+        client = window.core.api.anthropic.get_client()
         models_list = client.models.list()
         items = []
         if models_list.data:

pygpt_net/provider/llms/base.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.02 20:00:00                  #
+# Updated Date: 2025.08.28 09:00:00                  #
 # ================================================== #
 
 import os
@@ -18,7 +18,8 @@ from llama_index.core.multi_modal_llms import MultiModalLLM as LlamaMultiModalLL
 
 from pygpt_net.core.types import (
     MODE_LANGCHAIN,
-    MODE_LLAMA_INDEX, MODE_CHAT,
+    MODE_LLAMA_INDEX, 
+    MODE_CHAT,
 )
 from pygpt_net.item.model import ModelItem
 from pygpt_net.utils import parse_args
@@ -221,7 +222,7 @@ class BaseLLM:
         """
         model = ModelItem()
         model.provider = self.id
-        return window.core.gpt.get_client(
+        return window.core.api.openai.get_client(
             mode=MODE_CHAT,
             model=model,
         )

pygpt_net/provider/llms/openai.py

@@ -99,7 +99,7 @@ class OpenAILLM(BaseLLM):
 
         if window.core.config.get('api_use_responses_llama', False):
             tools = []
-            tools = window.core.gpt.remote_tools.append_to_tools(
+            tools = window.core.api.openai.remote_tools.append_to_tools(
                 mode=MODE_LLAMA_INDEX,
                 model=model,
                 stream=stream,

pygpt_net/provider/loaders/hub/image_vision/base.py

@@ -145,7 +145,7 @@ class ImageVisionLLMReader(BaseReader):
                 image = image.convert("RGB")
             image_str = img_2_b64(image)
 
-        client = self._window.core.gpt.get_client()
+        client = self._window.core.api.openai.get_client()
         encoded = self._encode_image(str(file))
         content = [
             {