pygpt-net 2.5.17__py3-none-any.whl → 2.5.18__py3-none-any.whl

pygpt_net/data/config/settings.json

@@ -1027,6 +1027,11 @@
         "step": null,
         "advanced": false,
         "keys": [
+            {"auto": "[gpt-image-1] auto"},
+            {"1024x1024": "[gpt-image-1] 1024x1024"},
+            {"1536x1024": "[gpt-image-1] 1536x1024"},
+            {"1024x1536": "[gpt-image-1] 1024x1536"},
+            {"1536x1024": "[gpt-image-1] 1536x1024"},
             {"1792x1024": "[DALL-E 3] 1792x1024"},
             {"1024x1792": "[DALL-E 3] 1024x1792"},
             {"1024x1024": "[DALL-E 3] 1024x1024"},
@@ -1047,6 +1052,10 @@
         "step": null,
         "advanced": false,
         "keys": [
+            {"auto": "[gpt-image-1] auto"},
+            {"high": "[gpt-image-1] high"},
+            {"medium": "[gpt-image-1] medium"},
+            {"medium": "[gpt-image-1] low"},
             {"standard": "[DALL-E 3] standard"},
             {"hd": "[DALL-E 3] hd"},
             {"standard": "[DALL-E 2] standard"}
@@ -1211,6 +1220,19 @@
         "step": null,
         "advanced": false
     },
+    "remote_tools.image": {
+        "section": "remote_tools",
+        "type": "bool",
+        "slider": false,
+        "label": "settings.remote_tools.image",
+        "description": "settings.remote_tools.image.desc",
+        "value": true,
+        "min": null,
+        "max": null,
+        "multiplier": null,
+        "step": null,
+        "advanced": false
+    },
     "llama.idx.list": {
         "section": "llama-index",
         "type": "dict",

pygpt_net/data/locale/locale.de.ini

@@ -989,7 +989,7 @@ tip.tokens.input = Token: Benutzereingabeaufforderung + Systemaufforderung + Kon
 tip.toolbox.assistants = Die Liste der Assistenten zeigt die erstellten Assistenten, die auf dem entfernten Server arbeiten. Alle Änderungen werden mit dem entfernten Assistenten synchronisiert.
 tip.toolbox.ctx = Erstellen Sie so viele Gesprächskontexte, wie Sie benötigen; Sie können jederzeit zu ihnen zurückkehren.
 tip.toolbox.indexes = Durch das Indizieren von Gesprächen und Dateien können Sie das verfügbare Wissen mit Ihren eigenen Daten und Gesprächsverläufen erweitern.
-tip.toolbox.mode = Sie können den Arbeitsmodus und das Modell in Echtzeit ändern. Um andere Modelle als GPT zu verwenden, nutzen Sie den Modus Chat mit Dateien.
+tip.toolbox.mode = Sie können den Arbeitsmodus und das Modell in Echtzeit ändern.
 tip.toolbox.presets = Erstellen Sie Voreinstellungen mit verschiedenen Konfigurationen, um schnell zwischen verschiedenen Einstellungen wie dem Systemprompt und anderen zu wechseln.
 tip.toolbox.prompt = Die aktuelle Systemeingabeaufforderung kann in Echtzeit geändert werden. Um Werkzeuge aus Plugins zu aktivieren, aktivieren Sie die Option "+ Werkzeuge."
 toolbox.agent.auto_stop.label = Automatischer Stopp

pygpt_net/data/locale/locale.en.ini

@@ -816,7 +816,7 @@ model.llama_index.mode.desc = Available sub-modes: chat
 model.llama_index.provider = [LlamaIndex] Provider
 model.llama_index.provider.desc = LLM provider to use in "Chat with Files" mode
 model.mode = Mode(s)
-model.mode.desc = Available modes: chat, completion, img, audio, vision, assistant, langchain, llama_index, agent, agent_llama, research
+model.mode.desc = Available modes: chat (Chat), llama_index (Chat with Files), audio (Chat with Audio), research (Research), completion (Completion), img (Image), vision (Vision), assistant (Assistants), langchain (Langchain), agent_llama (Agent LlamaIndex), agent (Agent Autonomous), expert (Experts)
 model.name = Name
 models.importer.all = Show all
 models.importer.available.label = Ollama models
@@ -829,6 +829,8 @@ models.importer.error.remove.no_model = No model selected to remove
 models.importer.error.remove.not_exists = Model already exists in current list
 models.importer.loaded = Ollama models loaded successfully.
 models.importer.status.imported = Models imported successfully.
+model.openai = OpenAI API
+model.openai.desc = Supports native OpenAI API
 model.tokens = Output tokens
 model.tokens.desc = Max model output tokens
 mode.research = Research (Perplexity)
@@ -1129,6 +1131,8 @@ settings.prompt.img = DALL-E: image generation
 settings.prompt.img.desc = Prompt for generating prompts for DALL-E (if raw-mode is disabled). Image mode only.
 settings.remote_tools.web_search = Web Search
 settings.remote_tools.web_search.desc = Enable `web_search` remote tool in Chat mode / via OpenAI Responses API.
+settings.remote_tools.image = Image generation
+settings.remote_tools.image.desc = Enable `image_generation` remote tool in Chat mode / via OpenAI Responses API.
 settings.render.code_syntax = Code syntax highlight
 settings.render.engine = Rendering engine
 settings.render.open_gl = OpenGL hardware acceleration
@@ -1238,7 +1242,7 @@ tip.tokens.input = Tokens: input prompt + system prompt + context + extra + atta
 tip.toolbox.assistants = The list of assistants shows the assistants created and operating on the remote server. Any changes will be synchronized with the remote assistant.
 tip.toolbox.ctx = Create as many conversation contexts as you need; you can return to them at any time.
 tip.toolbox.indexes = By indexing conversations and files, you can expand the available knowledge with your own data and conversation history.
-tip.toolbox.mode = You can change the working mode and model in real-time. To use models other than GPT, use the Chat with Files mode.
+tip.toolbox.mode = You can change the working mode and model in real-time.
 tip.toolbox.presets = Create presets with different configurations to quickly switch between various settings, such as the system prompt and others.
 tip.toolbox.prompt = The current system prompt can be modified in real-time. To enable tools from plugins, enable the option "+ Tools."
 toolbox.agent.auto_stop.label = Auto-stop

pygpt_net/data/locale/locale.es.ini

@@ -989,7 +989,7 @@ tip.tokens.input = Fichas: indicación del usuario + indicación del sistema + c
 tip.toolbox.assistants = La lista de asistentes muestra los asistentes creados y operando en el servidor remoto. Cualquier cambio se sincronizará con el asistente remoto.
 tip.toolbox.ctx = Crea tantos contextos de conversación como necesites; puedes volver a ellos en cualquier momento.
 tip.toolbox.indexes = Al indexar conversaciones y archivos, puedes ampliar el conocimiento disponible con tus propios datos e historial de conversaciones.
-tip.toolbox.mode = Puedes cambiar el modo de trabajo y el modelo en tiempo real. Para usar modelos distintos a GPT, utiliza el modo Chat con archivos.
+tip.toolbox.mode = Puedes cambiar el modo de trabajo y el modelo en tiempo real.
 tip.toolbox.presets = Crea preajustes con diferentes configuraciones para cambiar rápidamente entre varios ajustes, como el prompt del sistema y otros.
 tip.toolbox.prompt = La solicitud del sistema actual se puede modificar en tiempo real. Para habilitar herramientas desde complementos, habilite la opción "+ Herramientas."
 toolbox.agent.auto_stop.label = Auto-parada

pygpt_net/data/locale/locale.fr.ini

@@ -989,7 +989,7 @@ tip.tokens.input = Jetons: invite de l'utilisateur + invite système + contexte
 tip.toolbox.assistants = La liste des assistants montre les assistants créés et opérant sur le serveur distant. Tout changement sera synchronisé avec l'assistant distant.
 tip.toolbox.ctx = Créez autant de contextes de conversation que vous en avez besoin ; vous pouvez y revenir à tout moment.
 tip.toolbox.indexes = En indexant des conversations et des fichiers, vous pouvez étendre les connaissances disponibles avec vos propres données et historique de conversation.
-tip.toolbox.mode = Vous pouvez changer le mode de travail et le modèle en temps réel. Pour utiliser des modèles autres que GPT, utilisez le mode Chat avec fichiers.
+tip.toolbox.mode = Vous pouvez changer le mode de travail et le modèle en temps réel.
 tip.toolbox.presets = Créez des préréglages avec différentes configurations pour basculer rapidement entre divers réglages, tels que l'invite système et d'autres.
 tip.toolbox.prompt = L'invite système actuelle peut être modifiée en temps réel. Pour activer les outils à partir des plugins, activez l'option "+ Outils."
 toolbox.agent.auto_stop.label = Arrêt automatique

pygpt_net/data/locale/locale.pl.ini

@@ -990,7 +990,7 @@ tip.tokens.input = Tokeny: prompt użytkownika + systemowy prompt + kontekst + d
 tip.toolbox.assistants = Lista asystentów pokazuje asystentów stworzonych i działających na zdalnym serwerze. Wszelkie zmiany zostaną zsynchronizowane ze zdalnym asystentem.
 tip.toolbox.ctx = Twórz tyle kontekstów rozmów, ile potrzebujesz; możesz do nich wrócić w dowolnym momencie.
 tip.toolbox.indexes = Indeksując rozmowy i pliki, możesz rozszerzyć dostępną wiedzę o własne dane i historię rozmów.
-tip.toolbox.mode = Możesz zmienić tryb pracy i model w czasie rzeczywistym. Aby użyć modeli innych niż GPT, użyj trybu Czat z plikami.
+tip.toolbox.mode = Możesz zmienić tryb pracy i model w czasie rzeczywistym.
 tip.toolbox.presets = Twórz presety z różnymi konfiguracjami, aby szybko przełączać się między różnymi ustawieniami, takimi jak prompt systemowy i inne.
 tip.toolbox.prompt = Aktualna podpowiedź systemu może być modyfikowana w czasie rzeczywistym. Aby włączyć narzędzia z wtyczek, włącz opcję "+ Narzędzia."
 toolbox.agent.auto_stop.label = Auto-stop

pygpt_net/data/locale/locale.uk.ini

@@ -989,7 +989,7 @@ tip.tokens.input = Токени: запит користувача + систе
 tip.toolbox.assistants = Список асистентів показує асистентів, створених і що працюють на віддаленому сервері. Будь-які зміни будуть синхронізовані з віддаленим асистентом.
 tip.toolbox.ctx = Створіть стільки контекстів розмов, як вам потрібно; ви можете повернутися до них у будь-який час.
 tip.toolbox.indexes = Індексуючи розмови та файли, ви можете розширити доступні знання зі своїми власними даними та історією розмов.
-tip.toolbox.mode = Ви можете змінити робочий режим та модель в реальному часі. Щоб використовувати моделі, відмінні від GPT, використовуйте режим Чат з файлами.
+tip.toolbox.mode = Ви можете змінити робочий режим та модель в реальному часі.
 tip.toolbox.presets = Створіть пресети з різними конфігураціями для швидкого перемикання між різними налаштуваннями, такими як системний сповіщення та інші.
 tip.toolbox.prompt = Поточну системну підказку можна змінювати в режимі реального часу. Щоб увімкнути інструменти з плагінів, увімкніть опцію "+ Інструменти."
 toolbox.agent.auto_stop.label = Авто-стоп

pygpt_net/data/locale/locale.zh.ini

@@ -1104,7 +1104,7 @@ tip.tokens.input = 代币：用户输入提示 + 系统提示 + 上下文 + 额
 tip.toolbox.assistants = 助手列表顯示在遠程服務器上創建和運行的助手。任何更改都將與遠程助手同步。
 tip.toolbox.ctx = 創建所需數量的對話上下文；您隨時可以返回它們。
 tip.toolbox.indexes = 通過索引對話和文件，您可以用自己的數據和對話歷史擴展可用知識。
-tip.toolbox.mode = 您可以實時更換工作模式和模型。要使用非GPT模型，请使用“文件聊天模式”模式。
+tip.toolbox.mode = 您可以實時更換工作模式和模型。
 tip.toolbox.presets = 創建具有不同配置的預設，以便快速切換不同設置，例如系統提示等。
 tip.toolbox.prompt = 当前系统提示可以实时修改。要启用来自插件的工具，请启用“+ 工具”选项。
 toolbox.agent.auto_stop.label = 自動停止

pygpt_net/item/model.py

@@ -6,11 +6,13 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.06.24 16:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 import json
 
+from pygpt_net.core.types import MODE_CHAT
+
 
 class ModelItem:
     def __init__(self, id=None):
@@ -29,6 +31,7 @@ class ModelItem:
         self.tokens = 0
         self.default = False
         self.imported = False
+        self.openai = False  # OpenAI API supported model
         self.extra = {}
 
     def from_dict(self, data: dict):
@@ -54,6 +57,8 @@ class ModelItem:
             self.extra = data['extra']
         if 'imported' in data:
             self.imported = data['imported']
+        if 'openai' in data:
+            self.openai = data['openai']
 
         # multimodal
         if 'multimodal' in data:
@@ -105,6 +110,7 @@ class ModelItem:
         data['multimodal'] = ','.join(self.multimodal)
         data['extra'] = self.extra
         data['imported'] = self.imported
+        data['openai'] = self.openai
 
         data['langchain.provider'] = None
         data['langchain.mode'] = ""
@@ -178,6 +184,9 @@ class ModelItem:
         :param mode: Mode
         :return: True if supported
         """
+        if mode == MODE_CHAT and not self.is_openai():
+            # only OpenAI models are supported for chat mode
+            return False
         return mode in self.mode
 
     def is_multimodal(self) -> bool:
@@ -188,6 +197,21 @@ class ModelItem:
         """
         return len(self.multimodal) > 0
 
+    def is_openai(self) -> bool:
+        """
+        Check if model is supported by OpenAI API
+
+        :return: True if OpenAI
+        """
+        if (self.id.startswith("gpt-")
+                or self.id.startswith("chatgpt")
+                or self.id.startswith("o1")
+                or self.id.startswith("o3")
+                or self.id.startswith("o4")
+                or self.id.startswith("o5")):
+            return True
+        return False
+
     def is_ollama(self) -> bool:
         """
         Check if model is Ollama
@@ -196,6 +220,8 @@ class ModelItem:
         """
         if self.llama_index is None:
             return False
+        if self.llama_index.get("provider") is None:
+            return False
         return "ollama" in self.llama_index.get("provider", "")
 
     def get_ollama_model(self) -> str:
@@ -210,6 +236,14 @@ class ModelItem:
                     return arg["value"]
         return ""
 
+    def get_llama_provider(self) -> str:
+        """
+        Get Llama Index provider
+
+        :return: provider name
+        """
+        return self.llama_index.get("provider", "")
+
     def has_mode(self, mode: str) -> bool:
         """
         Check if model has mode

pygpt_net/provider/core/config/patch.py

@@ -1862,6 +1862,13 @@ class Patch:
                     data["remote_tools.web_search"] = True
                 updated = True
 
+            # < 2.5.18
+            if old < parse_version("2.5.18"):
+                print("Migrating config from < 2.5.18...")
+                if 'remote_tools.image' not in data:
+                    data["remote_tools.image"] = False
+                updated = True
+
         # update file
         migrated = False
         if updated:

pygpt_net/provider/core/model/json_file.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.02.02 02:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 import json
@@ -166,6 +166,7 @@ class JsonFileProvider(BaseProvider):
             'multimodal': item.multimodal,
             'extra': item.extra,
             'imported': item.imported,
+            'openai': item.openai,
         }
 
     @staticmethod
@@ -198,6 +199,8 @@ class JsonFileProvider(BaseProvider):
             item.extra = data['extra']
         if 'imported' in data:
             item.imported = data['imported']
+        if 'openai' in data:
+            item.openai = data['openai']
 
     def dump(self, item: ModelItem) -> str:
         """

pygpt_net/provider/core/model/patch.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.06.24 16:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 from packaging.version import parse as parse_version, Version
@@ -550,6 +550,22 @@ class Patch:
                         del data[name_to_replace]
                 updated = True
 
+            # < 2.5.18 <--- update openai flag
+            if old < parse_version("2.5.18"):
+                print("Migrating models from < 2.5.18...")
+                for id in data:
+                    model = data[id]
+                    if (model.id.startswith("o1")
+                            or model.id.startswith("o3")
+                            or model.id.startswith("gpt-")
+                            or model.id.startswith("chatgpt")
+                            or model.id.startswith("dall-e")):
+                        model.openai = True
+                    if model.is_supported("llama_index"):
+                        if "chat" not in model.mode:
+                            model.mode.append("chat")
+                updated = True
+
         # update file
         if updated:
             data = dict(sorted(data.items()))

pygpt_net/provider/gpt/__init__.py

@@ -8,6 +8,7 @@
 # Created By  : Marcin Szczygliński                  #
 # Updated Date: 2025.06.25 02:00:00                  #
 # ================================================== #
+import base64
 
 from httpx_socks import SyncProxyTransport
 
@@ -271,6 +272,19 @@ class Gpt:
                 response.usage.input_tokens,
                 response.usage.output_tokens,
             )
+            if mode == MODE_CHAT:
+                # if image generation call in responses API
+                image_data = [
+                    output.result
+                    for output in response.output
+                    if output.type == "image_generation_call"
+                ]
+                if image_data:
+                    img_path = self.window.core.image.gen_unique_path(ctx)
+                    image_base64 = image_data[0]
+                    with open(img_path, "wb") as f:
+                        f.write(base64.b64decode(image_base64))
+                    ctx.images = [img_path]
         return True
 
     def quick_call(self, context: BridgeContext, extra: dict = None) -> str:

pygpt_net/provider/gpt/image.py

@@ -6,9 +6,9 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2024.12.14 22:00:00                  #
+# Updated Date: 2025.06.26 18:00:00                  #
 # ================================================== #
-
+import base64
 import datetime
 import os
 from typing import Optional, Dict, Any
@@ -132,6 +132,7 @@ class ImageWorker(QObject, QRunnable):
         self.allowed_max_num = {
             "dall-e-2": 4,
             "dall-e-3": 1,
+            "gpt-image-1": 1,
         }
         self.allowed_resolutions = {
             "dall-e-2": [
@@ -144,6 +145,27 @@ class ImageWorker(QObject, QRunnable):
                 "1024x1792",
                 "1024x1024",
             ],
+            "gpt-image-1": [
+                "1536x1024",
+                "1024x1536",
+                "1024x1024",
+                "auto",
+            ],
+        }
+        self.allowed_quality = {
+            "dall-e-2": [
+                "standard",
+            ],
+            "dall-e-3": [
+                "standard",
+                "hd",
+            ],
+            "gpt-image-1": [
+                "auto",
+                "high",
+                "medium",
+                "low",
+            ],
         }
 
     @Slot()
@@ -188,6 +210,11 @@ class ImageWorker(QObject, QRunnable):
                 if resolution not in self.allowed_resolutions[self.model]:
                     resolution = self.allowed_resolutions[self.model][0]
 
+            quality = self.quality
+            if self.model in self.allowed_quality:
+                if quality not in self.allowed_quality[self.model]:
+                    quality = self.allowed_quality[self.model][0]
+
             # send to API
             response = None
             if self.model == "dall-e-2":
@@ -197,12 +224,12 @@ class ImageWorker(QObject, QRunnable):
                     n=self.num,
                     size=resolution,
                 )
-            elif self.model == "dall-e-3":
+            elif self.model == "dall-e-3" or self.model == "gpt-image-1":
                 response = self.client.images.generate(
                     model=self.model,
                     prompt=self.input_prompt,
                     n=self.num,
-                    quality=self.quality,
+                    quality=quality,
                     size=resolution,
                 )
 
@@ -215,20 +242,27 @@ class ImageWorker(QObject, QRunnable):
             for i in range(self.num):
                 if i >= len(response.data):
                     break
-                url = response.data[i].url
-                res = requests.get(url)
 
                 # generate filename
                 name = datetime.date.today().strftime(
                     "%Y-%m-%d") + "_" + datetime.datetime.now().strftime("%H-%M-%S") + "-" \
-                    + self.window.core.image.make_safe_filename(self.input_prompt) + "-" + str(i + 1) + ".png"
+                       + self.window.core.image.make_safe_filename(self.input_prompt) + "-" + str(i + 1) + ".png"
                 path = os.path.join(self.window.core.config.get_user_dir("img"), name)
 
                 msg = trans('img.status.downloading') + " (" + str(i + 1) + " / " + str(self.num) + ") -> " + str(path)
                 self.signals.status.emit(msg)
 
+                if response.data[i] is None:
+                    self.signals.error.emit("API Error: empty image data")
+                    return
+                if response.data[i].url:  # dall-e 2 and 3 returns URL
+                    res = requests.get(response.data[i].url)
+                    data = res.content
+                else:  # gpt-image-1 returns base64 encoded image
+                    data = base64.b64decode(response.data[i].b64_json)
+
                 # save image
-                if self.window.core.image.save_image(path, res.content):
+                if data and self.window.core.image.save_image(path, data):
                     paths.append(path)
                 else:
                     self.signals.error.emit("Error saving image")

pygpt_net/provider/gpt/responses.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.06.25 02:00:00                  #
+# Updated Date: 2025.06.26 18:00:00                  #
 # ================================================== #
 
 import json
@@ -17,13 +17,11 @@ from pygpt_net.core.types import (
     MODE_CHAT,
     MODE_VISION,
     MODE_AUDIO,
-    MODE_RESEARCH,
 )
 from pygpt_net.core.bridge.context import BridgeContext, MultimodalContext
 from pygpt_net.item.ctx import CtxItem
 from pygpt_net.item.model import ModelItem
 
-from .utils import sanitize_name
 from pygpt_net.item.attachment import AttachmentItem
 
 
@@ -38,6 +36,7 @@ class Responses:
         self.input_tokens = 0
         self.audio_prev_id = None
         self.audio_prev_expires_ts = None
+        self.prev_response_id = None
 
     def send(
             self,
@@ -80,6 +79,7 @@ class Responses:
             user_name=user_name,
             multimodal_ctx=multimodal_ctx,
         )
+
         msg_tokens = self.window.core.tokens.from_messages(
             messages,
             model.id,
@@ -116,9 +116,15 @@ class Responses:
             response_kwargs['reasoning']['effort'] = model.extra["reasoning_effort"]
 
         # extend tools with external tools
-        if not model.id.startswith("o1") and not model.id.startswith("o3"):
+        if (not model.id.startswith("o1")
+                and not model.id.startswith("o3")):
             if self.window.core.config.get("remote_tools.web_search", False):
                 tools.append({"type": "web_search_preview"})
+            if self.window.core.config.get("remote_tools.image", False):
+                tool = {"type": "image_generation"}
+                if stream:
+                    tool["partial_images"] = 1  # required for streaming
+                tools.append(tool)
 
         # tool calls are not supported for o1-mini and o1-preview
         if (model.id is not None
@@ -126,18 +132,9 @@ class Responses:
             if len(tools) > 0:
                 response_kwargs['tools'] = tools
 
-        # audio mode
-        if mode in [MODE_AUDIO]:
-            stream = False
-            voice_id = "alloy"
-            tmp_voice = self.window.core.plugins.get_option("audio_output", "openai_voice")
-            if tmp_voice:
-                voice_id = tmp_voice
-            response_kwargs["modalities"] = ["text", "audio"]
-            response_kwargs["audio"] = {
-                "voice": voice_id,
-                "format": "wav"
-            }
+        # attach previous response ID if available
+        if self.prev_response_id:
+            response_kwargs['previous_response_id'] = self.prev_response_id
 
         response = client.responses.create(
             input=messages,
@@ -145,6 +142,11 @@ class Responses:
             stream=stream,
             **response_kwargs,
         )
+
+        # store previous response ID
+        if not stream and response:
+            ctx.msg_id = response.id
+
         return response
 
     def build(
@@ -172,6 +174,7 @@ class Responses:
         :return: messages list
         """
         messages = []
+        self.prev_response_id = None  # reset
 
         # tokens config
         mode = MODE_CHAT
@@ -240,6 +243,9 @@ class Responses:
                                 }
                     messages.append(msg)
 
+                if item.msg_id and (item.cmds is None or len(item.cmds) == 0):  # if no cmds before
+                    self.prev_response_id = item.msg_id  # previous response ID to use in current input
+
         # use vision and audio if available in current model
         content = str(prompt)
         if MODE_VISION in model.mode:

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: 2024.12.14 22:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 from llama_index.llms.anthropic import Anthropic
@@ -47,4 +47,6 @@ class AnthropicLLM(BaseLLM):
         :return: LLM provider instance
         """
         args = self.parse_args(model.llama_index)
+        if "model" not in args:
+            args["model"] = model.id
         return Anthropic(**args)

pygpt_net/provider/llms/google.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.01.16 01:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 from typing import Optional, List, Dict
@@ -51,6 +51,8 @@ class GoogleLLM(BaseLLM):
         :return: LLM provider instance
         """
         args = self.parse_args(model.llama_index)
+        if "model" not in args:
+            args["model"] = model.id
         return Gemini(**args)
 
     def get_embeddings_model(

pygpt_net/provider/llms/hugging_face.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2024.12.14 22:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 from langchain_community.llms import HuggingFaceHub
@@ -39,6 +39,8 @@ class HuggingFaceLLM(BaseLLM):
         :return: LLM provider instance
         """
         args = self.parse_args(model.langchain)
+        if "model" not in args:
+            args["model"] = model.id
         return HuggingFaceHub(**args)
 
     def chat(

pygpt_net/provider/llms/hugging_face_api.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2024.12.14 22:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 import os
@@ -44,6 +44,8 @@ class HuggingFaceApiLLM(BaseLLM):
         :return: LLM provider instance
         """
         args = self.parse_args(model.llama_index)
+        if "model" not in args:
+            args["model"] = model.id
         return HuggingFaceInferenceAPI(**args)
 
     def get_embeddings_model(