pygpt-net 2.6.65__py3-none-any.whl → 2.6.66__py3-none-any.whl

pygpt_net/data/locale/locale.de.ini

@@ -51,6 +51,7 @@ action.save = Speichern
 action.save_as = Speichern unter...
 action.save_selection_as = Auswahl speichern unter...
 action.select_all = Alle auswählen
+action.select_unselect_all = Alle auswählen/abwählen
 action.tab.add.chat: Neuen Chat hinzufügen
 action.tab.add.chat.tooltip: Neuen Chat hinzufügen (RMT für mehr Optionen...)
 action.tab.add.notepad = Neuen Notizblock hinzufügen
@@ -902,7 +903,10 @@ mode.computer.tooltip = Computerbenutzung (Maus, Tastatur, Navigation)
 mode.expert = Experten (Kooperation)
 mode.expert.tooltip = Experten zum Hintergrund rufen
 mode.img = Bild und Video
+mode.img.image = Bild
+mode.img.music = Musik
 mode.img.tooltip = Bildgenerierung mit DALL-E
+mode.img.video = Video
 mode.langchain = Langchain
 mode.langchain.tooltip = Chat mit Modellen von Langchain
 model.ctx = Kontext-Token

pygpt_net/data/locale/locale.en.ini

@@ -51,6 +51,7 @@ action.save = Save
 action.save_as = Save as...
 action.save_selection_as = Save selection as...
 action.select_all = Select all
+action.select_unselect_all = Select/Unselect All
 action.tab.add.chat = Add a new chat
 action.tab.add.chat.tooltip = Add a new chat (RMB click to more options...)
 action.tab.add.notepad = Add a new notepad
@@ -917,7 +918,10 @@ mode.computer.tooltip = Computer use (mouse, keyboard, navigation)
 mode.expert = Experts (Co-op)
 mode.expert.tooltip = Experts to call in the background
 mode.img = Image and video
+mode.img.image = Image
+mode.img.music = Music
 mode.img.tooltip = Image generation
+mode.img.video = Video
 mode.langchain = Langchain
 mode.langchain.tooltip = Chat with models provided by Langchain
 model.ctx = Context tokens

pygpt_net/data/locale/locale.es.ini

@@ -51,6 +51,7 @@ action.save = Guardar
 action.save_as = Guardar como...
 action.save_selection_as = Guardar selección como...
 action.select_all = Seleccionar todo
+action.select_unselect_all = Seleccionar/Deseleccionar todo
 action.tab.add.chat: Añadir un nuevo chat
 action.tab.add.chat.tooltip: Añadir un nuevo chat (clic derecho para más opciones...)
 action.tab.add.notepad = Agregar un nuevo bloc de notas
@@ -903,7 +904,10 @@ mode.computer.tooltip = Uso de la computadora (ratón, teclado, navegación)
 mode.expert = Expertos (cooperación)
 mode.expert.tooltip = Llamar a expertos en segundo plano
 mode.img = Imagen y video
+mode.img.image = Imagen
+mode.img.music = Música
 mode.img.tooltip = Generar imagen usando DALL-E
+mode.img.video = Vídeo
 mode.langchain = Langchain
 mode.langchain.tooltip = Chatear con modelos proporcionados por Langchain
 model.ctx = Tokens de contexto

pygpt_net/data/locale/locale.fr.ini

@@ -51,6 +51,7 @@ action.save = Enregistrer
 action.save_as = Enregistrer sous...
 action.save_selection_as = Enregistrer la sélection sous...
 action.select_all = Sélectionner tout
+action.select_unselect_all = Sélectionner/Désélectionner tout
 action.tab.add.chat: Ajouter une nouvelle discussion
 action.tab.add.chat.tooltip: Ajouter une nouvelle discussion (Clic Droit pour plus d'options...)
 action.tab.add.notepad = Ajouter un nouveau bloc-notes
@@ -902,7 +903,10 @@ mode.computer.tooltip = Utilisation de l'ordinateur (souris, clavier, navigation
 mode.expert = Experts (co-opération)
 mode.expert.tooltip = Appeler des experts en arrière-plan
 mode.img = Image et vidéo
+mode.img.image = Image
+mode.img.music = Musique
 mode.img.tooltip = Génération d'image avec DALL-E
+mode.img.video = Vidéo
 mode.langchain = Langchain
 mode.langchain.tooltip = Discussion avec les modèles fournis par Langchain
 model.ctx = Jetons de contexte

pygpt_net/data/locale/locale.it.ini

@@ -51,6 +51,7 @@ action.save = Salva
 action.save_as = Salva con nome...
 action.save_selection_as = Salva selezione come...
 action.select_all = Seleziona tutto
+action.select_unselect_all = Seleziona/Deseleziona tutto
 action.tab.add.chat: Aggiungi una nuova chat
 action.tab.add.chat.tooltip: Aggiungi una nuova chat (Clic Destro per maggiori opzioni...)
 action.tab.add.notepad = Aggiungi un nuovo blocco note
@@ -902,7 +903,10 @@ mode.computer.tooltip = Uso del computer (mouse, tastiera, navigazione)
 mode.expert = Esperti (cooperazione)
 mode.expert.tooltip = Chiamare esperti in background
 mode.img = Immagine e video
+mode.img.image = Immagine
+mode.img.music = Musica
 mode.img.tooltip = Generazione immagini con DALL-E
+mode.img.video = Video
 mode.langchain = Langchain
 mode.langchain.tooltip = Chattare con modelli forniti da Langchain
 model.ctx = Token di contesto

pygpt_net/data/locale/locale.pl.ini

@@ -51,6 +51,7 @@ action.save = Zapisz
 action.save_as = Zapisz jako...
 action.save_selection_as = Zapisz zaznaczenie jako...
 action.select_all = Wybierz wszystko
+action.select_unselect_all = Zaznacz/Odznacz wszystko
 action.tab.add.chat: Dodaj nowy czat
 action.tab.add.chat.tooltip: Dodaj nowy czat (PPM dla więcej opcji...)
 action.tab.add.notepad = Dodaj nowy notatnik
@@ -903,7 +904,10 @@ mode.computer.tooltip = Kontrola komputera (mysz, klawiatura)
 mode.expert = Eksperci (współpraca)
 mode.expert.tooltip = Eksperci do wezwania w tle
 mode.img = Obraz i wideo
+mode.img.image = Obraz
+mode.img.music = Muzyka
 mode.img.tooltip = Generowanie obrazów przy użyciu DALL-E
+mode.img.video = Wideo
 mode.langchain = Langchain
 mode.langchain.tooltip = Czat z modelami dostarczonymi przez Langchain
 model.ctx = Tokeny kontekstu

pygpt_net/data/locale/locale.uk.ini

@@ -51,6 +51,7 @@ action.save = Зберегти
 action.save_as = Зберегти як...
 action.save_selection_as = Зберегти вибір як...
 action.select_all = Вибрати все
+action.select_unselect_all = Вибрати/Скасувати вибір всіх
 action.tab.add.chat: Додати новий чат
 action.tab.add.chat.tooltip: Додати новий чат (ПКМ для більше опцій...)
 action.tab.add.notepad = Додати новий блокнот
@@ -902,7 +903,10 @@ mode.computer.tooltip = Використання комп'ютера (миша,
 mode.expert = Експерти (співпраця)
 mode.expert.tooltip = Виклик експертів на задній план
 mode.img = Зображення та відео
+mode.img.image = Зображення
+mode.img.music = Музика
 mode.img.tooltip = Генерація зображень за допомогою DALL-E
+mode.img.video = Відео
 mode.langchain = Langchain
 mode.langchain.tooltip = Чат з моделями, наданими Langchain
 model.ctx = Токени контексту

pygpt_net/data/locale/locale.zh.ini

@@ -51,6 +51,7 @@ action.save = 保存
 action.save_as = 另存為...
 action.save_selection_as = 将选择保存为...
 action.select_all = 选择全部
+action.select_unselect_all = 全选/取消全选
 action.tab.add.chat: 添加新聊天
 action.tab.add.chat.tooltip: 添加新聊天（右键单击获取更多选项...)
 action.tab.add.notepad = 添加新记事本
@@ -902,7 +903,10 @@ mode.computer.tooltip = 使用计算机（鼠标、键盘、导航）
 mode.expert = 专家 (合作)
 mode.expert.tooltip = 背景中呼叫专家
 mode.img = 图像和视频
+mode.img.image = 图片
+mode.img.music = 音乐
 mode.img.tooltip = 使用DALL-E生成圖像
+mode.img.video = 视频
 mode.langchain = Langchain模式
 mode.langchain.tooltip = 使用Langchain提供的模型進行聊天
 model.ctx = 上下文令牌

pygpt_net/item/model.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.15 01:00:00                  #
+# Updated Date: 2025.12.25 20:00:00                  #
 # ================================================== #
 
 import json
@@ -274,9 +274,7 @@ class ModelItem:
 
         :return: True if supports image input
         """
-        if MULTIMODAL_IMAGE in self.input:
-            return True
-        return False
+        return MULTIMODAL_IMAGE in self.input
 
     def is_image_output(self) -> bool:
         """
@@ -284,9 +282,7 @@ class ModelItem:
 
         :return: True if supports image output
         """
-        if "image" in self.output or MODE_VISION in self.mode:
-            return True
-        return False
+        return "image" in self.output or MODE_VISION in self.mode
 
     def is_audio_input(self) -> bool:
         """
@@ -294,9 +290,7 @@ class ModelItem:
 
         :return: True if supports audio input
         """
-        if MULTIMODAL_AUDIO in self.input:
-            return True
-        return False
+        return MULTIMODAL_AUDIO in self.input
 
     def is_audio_output(self) -> bool:
         """
@@ -304,9 +298,7 @@ class ModelItem:
 
         :return: True if supports audio output
         """
-        if MULTIMODAL_AUDIO in self.output:
-            return True
-        return False
+        return MULTIMODAL_AUDIO in self.output
 
     def is_video_input(self) -> bool:
         """
@@ -314,9 +306,7 @@ class ModelItem:
 
         :return: True if supports video input
         """
-        if MULTIMODAL_VIDEO in self.input:
-            return True
-        return False
+        return MULTIMODAL_VIDEO in self.input
 
     def is_video_output(self) -> bool:
         """
@@ -324,9 +314,15 @@ class ModelItem:
 
         :return: True if supports video output
         """
-        if MULTIMODAL_VIDEO in self.output:
-            return True
-        return False
+        return MULTIMODAL_VIDEO in self.output
+
+    def is_music_output(self) -> bool:
+        """
+        Check if model supports music output
+
+        :return: True if supports music output
+        """
+        return MULTIMODAL_AUDIO in self.output
 
     def dump(self) -> str:
         """

pygpt_net/provider/api/google/__init__.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.14 00:00:00                  #
+# Updated Date: 2025.12.25 20:00:00                  #
 # ================================================== #
 
 import os
@@ -89,7 +89,14 @@ class ApiGoogle:
             filtered["location"] = os.environ.get("GOOGLE_CLOUD_LOCATION", "us-central1")
             # filtered["http_options"] = gtypes.HttpOptions(api_version="v1")
 
-        return genai.Client(**filtered)
+        # use previous client if args are the same
+        if self.client and self.last_client_args == filtered:
+            return self.client
+
+        self.last_client_args = filtered
+        self.client = genai.Client(**filtered)
+
+        return self.client
 
     def call(
             self,
@@ -138,13 +145,17 @@ class ApiGoogle:
 
         elif mode == MODE_IMAGE:
             # Route to video / music / image based on selected model.
-            if context.model.is_video_output():
-                return self.video.generate(context=context, extra=extra)  # veo, etc.
-            # Lyria / music models
-            if self.music.is_music_model(model.id if model else ""):
-                return self.music.generate(context=context, extra=extra)   # lyria, etc.
-            # Default: image
-            return self.image.generate(context=context, extra=extra)       # imagen, etc.
+            media_mode = self.window.controller.media.get_mode()
+            if media_mode == "video":
+                if context.model.is_video_output():
+                    return self.video.generate(context=context, extra=extra)  # veo, etc.
+            elif media_mode == "music":
+                # Lyria / music models
+                if self.music.is_music_model(model.id if model else ""):
+                    return self.music.generate(context=context, extra=extra)   # lyria, etc.
+            elif media_mode == "image":
+                # Default: image
+                return self.image.generate(context=context, extra=extra)       # imagen, etc.
 
         elif mode == MODE_ASSISTANT:
             return False  # not implemented for Google

pygpt_net/provider/api/google/image.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.14 00:00:00                  #
+# Updated Date: 2025.12.25 20:00:00                  #
 # ================================================== #
 
 import mimetypes
@@ -39,14 +39,8 @@ class Image:
     ) -> bool:
         """
         Generate or edit image(s) using Google GenAI API (Developer API or Vertex AI).
-
-        :param context: BridgeContext with prompt, model, attachments
-        :param extra: extra parameters (num, inline)
-        :param sync: run synchronously (blocking) if True
-        :return: True if started
         """
         # Music fast-path: delegate to Music flow if a music model is selected (e.g., Lyria).
-        # This keeps image flow unchanged while enabling music in the same "image" mode.
         try:
             model_id = (context.model.id if context and context.model else "") or ""
             if self.window and hasattr(self.window.core.api.google, "music"):
@@ -65,9 +59,6 @@ class Image:
         # decide sub-mode based on attachments
         sub_mode = self.MODE_GENERATE
         attachments = context.attachments
-        if attachments and len(attachments) > 0:
-            pass # TODO: implement edit!
-            # sub_mode = self.MODE_EDIT
 
         # model used to improve the prompt (not image model)
         prompt_model = self.window.core.models.from_defaults()
@@ -89,6 +80,11 @@ class Image:
         worker.num = num
         worker.inline = inline
 
+        if attachments and len(attachments) > 0:
+            mid = str(model.id).lower()
+            if "imagen" in mid:
+                worker.mode = self.MODE_EDIT
+
         if self.window.core.config.has('img_resolution'):
             worker.resolution = self.window.core.config.get('img_resolution') or "1024x1024"
 
@@ -132,13 +128,28 @@ class ImageWorker(QRunnable):
         self.inline = False
         self.raw = False
         self.num = 1
-        self.resolution = "1024x1024"  # used to derive aspect ratio for Imagen
+        self.resolution = "1024x1024"  # used to derive aspect ratio or image_size
 
         # limits
         self.imagen_max_num = 4  # Imagen returns up to 4 images
 
         # fallbacks
-        self.DEFAULT_GEMINI_IMAGE_MODEL = "gemini-2.0-flash-preview-image-generation"
+        self.DEFAULT_GEMINI_IMAGE_MODEL = "gemini-2.5-flash-image"
+
+        # Canonical 1K dimensions for Nano Banana Pro (Gemini 3 Pro Image Preview).
+        # Used to infer 2K/4K by 2x/4x multiples and to normalize UI inputs.
+        self._NB_PRO_1K = {
+            "1024x1024",  # 1:1
+            "848x1264",   # 2:3
+            "1264x848",   # 3:2
+            "896x1200",   # 3:4
+            "1200x896",   # 4:3
+            "928x1152",   # 4:5
+            "1152x928",   # 5:4
+            "768x1376",   # 9:16
+            "1376x768",   # 16:9
+            "1584x672",   # 21:9
+        }
 
     @Slot()
     def run(self):
@@ -179,7 +190,7 @@ class ImageWorker(QRunnable):
                         if p:
                             paths.append(p)
                 else:
-                    # Developer API fallback via Gemini image model; force v1 to avoid 404
+                    # Gemini Developer API via Gemini image models (Nano Banana / Nano Banana Pro)
                     resp = self._gemini_edit(self.input_prompt, self.attachments, self.num)
                     saved = 0
                     for cand in getattr(resp, "candidates", []) or []:
@@ -208,14 +219,8 @@ class ImageWorker(QRunnable):
                         if p:
                             paths.append(p)
                 else:
-                    # Gemini Developer API image generation (needs response_modalities)
-                    resp = self.client.models.generate_content(
-                        model=self.model,
-                        contents=[self.input_prompt],
-                        config=gtypes.GenerateContentConfig(
-                            response_modalities=[gtypes.Modality.TEXT, gtypes.Modality.IMAGE],
-                        ),
-                    )
+                    # Gemini Developer API image generation (Nano Banana / Nano Banana Pro) with robust sizing + optional reference images
+                    resp = self._gemini_generate_image(self.input_prompt, self.model, self.resolution)
                     saved = 0
                     for cand in getattr(resp, "candidates", []) or []:
                         parts = getattr(getattr(cand, "content", None), "parts", None) or []
@@ -316,10 +321,114 @@ class ImageWorker(QRunnable):
             config=cfg,
         )
 
+    def _is_gemini_pro_image_model(self, model_id: str) -> bool:
+        """
+        Detect Gemini 3 Pro Image (Nano Banana Pro) by id or UI alias.
+        """
+        mid = (model_id or "").lower()
+        return mid.startswith("gemini-") or mid.startswith("nano-banana") or mid.startswith("nb-")
+
+    def _infer_nb_pro_size_for_dims(self, w: int, h: int) -> Optional[str]:
+        """
+        Infer '1K' | '2K' | '4K' for Nano Banana Pro from WxH.
+        """
+        dims = f"{w}x{h}"
+        if dims in self._NB_PRO_1K:
+            return "1K"
+        if (w % 2 == 0) and (h % 2 == 0):
+            if f"{w // 2}x{h // 2}" in self._NB_PRO_1K:
+                return "2K"
+        if (w % 4 == 0) and (h % 4 == 0):
+            if f"{w // 4}x{h // 4}" in self._NB_PRO_1K:
+                return "4K"
+        mx = max(w, h)
+        if mx >= 4000:
+            return "4K"
+        if mx >= 2000:
+            return "2K"
+        return "1K"
+
+    def _build_gemini_image_config(self, model_id: str, resolution: str) -> Optional[gtypes.ImageConfig]:
+        """
+        Build ImageConfig for Gemini image models.
+        """
+        try:
+            aspect = self._aspect_from_resolution(resolution)
+            cfg = gtypes.ImageConfig()
+            if aspect:
+                cfg.aspect_ratio = aspect
+
+            # Only Pro supports image_size; detect by id/alias and set 1K/2K/4K from WxH.
+            if self._is_gemini_pro_image_model(model_id):
+                w_str, h_str = resolution.lower().replace("×", "x").split("x")
+                w, h = int(w_str.strip()), int(h_str.strip())
+                k = self._infer_nb_pro_size_for_dims(w, h)
+                if k:
+                    cfg.image_size = k
+            return cfg
+        except Exception:
+            return None
+
+    def _attachment_image_parts(self) -> List[gtypes.Part]:
+        """
+        Build image Parts from current attachments for Gemini models.
+        """
+        parts: List[gtypes.Part] = []
+        paths = self._collect_attachment_paths(self.attachments)
+        for p in paths:
+            try:
+                mime = self._guess_mime(p)
+                if not mime or not mime.startswith("image/"):
+                    continue
+                with open(p, "rb") as f:
+                    data = f.read()
+                parts.append(gtypes.Part.from_bytes(data=data, mime_type=mime))
+            except Exception:
+                continue
+        return parts
+
+    def _gemini_generate_image(self, prompt: str, model_id: str, resolution: str):
+        """
+        Call Gemini generate_content with robust fallback for image_size.
+        Supports optional reference images uploaded as attachments.
+        """
+        cfg = self._build_gemini_image_config(model_id, resolution)
+        image_parts = self._attachment_image_parts()
+
+        def _do_call(icfg: Optional[gtypes.ImageConfig]):
+            contents: List[Any] = []
+            # Always include the textual prompt (can be empty string).
+            contents.append(prompt or "")
+            # Append reference images, if any.
+            if image_parts:
+                contents.extend(image_parts)
+            return self.client.models.generate_content(
+                model=model_id or self.DEFAULT_GEMINI_IMAGE_MODEL,
+                contents=contents,
+                config=gtypes.GenerateContentConfig(
+                    response_modalities=[gtypes.Modality.TEXT, gtypes.Modality.IMAGE],
+                    image_config=icfg,
+                ),
+            )
+
+        try:
+            return _do_call(cfg)
+        except Exception as e:
+            msg = str(e)
+            if "imageSize" in msg or "image_size" in msg or "Unrecognized" in msg or "unsupported" in msg:
+                try:
+                    if cfg and getattr(cfg, "image_size", None):
+                        cfg2 = gtypes.ImageConfig()
+                        cfg2.aspect_ratio = getattr(cfg, "aspect_ratio", None)
+                        return _do_call(cfg2)
+                except Exception:
+                    pass
+            raise
+
     def _gemini_edit(self, prompt: str, attachments: Dict[str, Any], num: int):
         """
-        Gemini image-to-image editing via generate_content (Developer/Vertex depending on client).
-        The first attachment is used as the input image.
+        Gemini image-to-image editing via generate_content.
+        The first attachment is used as the input image. Honors aspect_ratio and (for Pro) image_size.
         """
         paths = self._collect_attachment_paths(attachments)
         if len(paths) == 0:
@@ -330,10 +439,27 @@ class ImageWorker(QRunnable):
             img_bytes = f.read()
         mime = self._guess_mime(img_path)
 
-        return self.client.models.generate_content(
-            model=self.model,
-            contents=[prompt, gtypes.Part.from_bytes(data=img_bytes, mime_type=mime)],
-        )
+        cfg = self._build_gemini_image_config(self.model, self.resolution)
+
+        def _do_call(icfg: Optional[gtypes.ImageConfig]):
+            return self.client.models.generate_content(
+                model=self.model or self.DEFAULT_GEMINI_IMAGE_MODEL,
+                contents=[prompt, gtypes.Part.from_bytes(data=img_bytes, mime_type=mime)],
+                config=gtypes.GenerateContentConfig(
+                    image_config=icfg,
+                ),
+            )
+
+        try:
+            return _do_call(cfg)
+        except Exception as e:
+            msg = str(e)
+            if "imageSize" in msg or "image_size" in msg or "Unrecognized" in msg or "unsupported" in msg:
+                if cfg and getattr(cfg, "image_size", None):
+                    cfg2 = gtypes.ImageConfig()
+                    cfg2.aspect_ratio = getattr(cfg, "aspect_ratio", None)
+                    return _do_call(cfg2)
+            raise
 
     def _collect_attachment_paths(self, attachments: Dict[str, Any]) -> List[str]:
         """Extract file paths from attachments dict."""
@@ -347,7 +473,7 @@ class ImageWorker(QRunnable):
         return out
 
     def _aspect_from_resolution(self, resolution: str) -> Optional[str]:
-        """Derive aspect ratio for Imagen."""
+        """Derive aspect ratio from WxH across supported set."""
         try:
             from math import gcd
             tolerance = 0.08
@@ -357,10 +483,15 @@ class ImageWorker(QRunnable):
                 return None
             supported = {
                 "1:1": 1 / 1,
+                "2:3": 2 / 3,
+                "3:2": 3 / 2,
                 "3:4": 3 / 4,
                 "4:3": 4 / 3,
+                "4:5": 4 / 5,
+                "5:4": 5 / 4,
                 "9:16": 9 / 16,
                 "16:9": 16 / 9,
+                "21:9": 21 / 9,
             }
             g = gcd(w, h)
             key = f"{w // g}:{h // g}"
@@ -424,6 +555,8 @@ class ImageWorker(QRunnable):
             return 'image/jpeg'
         if ext == '.webp':
             return 'image/webp'
+        if ext in ('.heic', '.heif'):
+            return 'image/heic'
         return 'image/png'
 
     def _cleanup(self):