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

pygpt_net/provider/audio_output/google_genai_tts.py

@@ -0,0 +1,229 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+# ================================================== #
+# This file is a part of PYGPT package               #
+# Website: https://pygpt.net                         #
+# GitHub:  https://github.com/szczyglis-dev/py-gpt   #
+# MIT License                                        #
+# Created By  : Marcin Szczygliński                  #
+# Updated Date: 2025.08.29 18:00:00                  #
+# ================================================== #
+
+import os
+import wave
+import base64
+
+from .base import BaseProvider
+
+
+class GoogleGenAITextToSpeech(BaseProvider):
+    def __init__(self, *args, **kwargs):
+        """
+        Google GenAI Text-to-Speech provider (Gemini TTS via API).
+
+        :param args: args
+        :param kwargs: kwargs
+        """
+        super(GoogleGenAITextToSpeech, self).__init__(*args, **kwargs)
+        self.plugin = kwargs.get("plugin")
+        self.id = "google_genai_tts"
+        self.name = "Google GenAI TTS"
+
+        # Supported preview TTS models (fallback to flash if invalid)
+        self.allowed_models = [
+            "gemini-2.5-flash-preview-tts",
+            "gemini-2.5-pro-preview-tts",
+        ]
+
+        # Prebuilt voice names exposed by Gemini TTS
+        # Keep list in sync with official docs; fallback to "Puck" if invalid.
+        self.allowed_voices = [
+            "Zephyr", "Puck", "Charon", "Kore", "Fenrir", "Leda", "Orus",
+            "Aoede", "Callirrhoe", "Autonoe", "Enceladus", "Iapetus",
+            "Umbriel", "Algieba", "Despina", "Erinome", "Algenib",
+            "Rasalgethi", "Laomedeia", "Achernar", "Alnilam", "Schedar",
+            "Gacrux", "Pulcherrima", "Achird", "Zubenelgenubi",
+            "Vindemiatrix", "Sadachbia", "Sadaltager", "Sulafat",
+        ]
+
+    def init_options(self):
+        """Initialize options"""
+        # Keep option names consistent with the app style; simple text fields are enough.
+        self.plugin.add_option(
+            "google_genai_tts_model",
+            type="text",
+            value="gemini-2.5-flash-preview-tts",
+            label="Model",
+            tab="google_genai_tts",
+            description="Specify Gemini TTS model, e.g.: gemini-2.5-flash-preview-tts or gemini-2.5-pro-preview-tts",
+        )
+        self.plugin.add_option(
+            "google_genai_tts_voice",
+            type="text",
+            value="Kore",
+            label="Voice",
+            tab="google_genai_tts",
+            description="Specify voice, e.g.: Puck, Kore, Charon, Leda, Zephyr... (case-sensitive)",
+            urls={
+                "Voices": "https://ai.google.dev/gemini-api/docs/speech-generation"
+            },
+        )
+
+    def speech(self, text: str) -> str:
+        """
+        Text to speech synthesis using Google GenAI (Gemini TTS).
+
+        :param text: text to synthesize
+        :return: path to generated audio file
+        """
+        # Get pre-configured GenAI client
+        client = self.plugin.window.core.api.google.get_client()
+
+        # Resolve path where audio should be written
+        output_file = self.plugin.output_file
+        path = os.path.join(self.plugin.window.core.config.path, output_file)
+
+        # Validate/select model
+        model = self.plugin.get_option_value("google_genai_tts_model") or "gemini-2.5-flash-preview-tts"
+        model = self._normalize_model_name(model)
+        if model not in self.allowed_models:
+            model = "gemini-2.5-flash-preview-tts"
+
+        # Validate/select voice
+        voice = self.plugin.get_option_value("google_genai_tts_voice") or "Kore"
+        # if voice not in self.allowed_voices:
+            # voice = "Kore"
+
+        # Build generation config for audio modality + voice
+        # Using explicit types for clarity and forward-compatibility
+        try:
+            from google.genai import types
+        except Exception as ex:
+            # Fail fast if SDK is missing or incompatible
+            raise RuntimeError("google.genai SDK is not available. Please install/update Google GenAI SDK.") from ex
+
+        gen_config = types.GenerateContentConfig(
+            response_modalities=["AUDIO"],
+            speech_config=types.SpeechConfig(
+                voice_config=types.VoiceConfig(
+                    prebuilt_voice_config=types.PrebuiltVoiceConfig(
+                        voice_name=voice
+                    )
+                )
+            ),
+            temperature=0.8,  # balanced default; keep configurable later if needed
+        )
+
+        # Perform TTS request
+        response = client.models.generate_content(
+            model=model,
+            contents=text,
+            config=gen_config,
+        )
+
+        # Extract PCM bytes from the first candidate/part
+        pcm = self._extract_pcm_bytes(response)
+
+        # Persist as standard WAV (PCM 16-bit, mono, 24 kHz)
+        self._save_wav(path, pcm, channels=1, rate=24000, sample_width=2)
+
+        return str(path)
+
+    def _extract_pcm_bytes(self, response) -> bytes:
+        """
+        Extract PCM bytes from generate_content response.
+
+        :param response: Google GenAI response object
+        :return: raw PCM byte data
+        """
+        # Defensive extraction to support minor SDK variations
+        data = None
+        try:
+            cand = response.candidates[0]
+            part = cand.content.parts[0]
+            if getattr(part, "inline_data", None) and getattr(part.inline_data, "data", None):
+                data = part.inline_data.data
+        except Exception:
+            pass
+
+        if data is None:
+            # Some SDK builds may return base64 str; try resolving alternative layout
+            try:
+                parts = getattr(response, "candidates", [])[0].content.parts
+                for p in parts:
+                    if getattr(p, "inline_data", None) and getattr(p.inline_data, "data", None):
+                        data = p.inline_data.data
+                        break
+            except Exception:
+                pass
+
+        if data is None:
+            raise RuntimeError("No audio data returned by Gemini TTS response.")
+
+        # Normalize to raw bytes
+        if isinstance(data, (bytes, bytearray)):
+            return bytes(data)
+        if isinstance(data, str):
+            # Fallback: treat as base64-encoded PCM
+            return base64.b64decode(data)
+
+        # Last resort: try bytes() cast
+        try:
+            return bytes(data)
+        except Exception as ex:
+            raise RuntimeError("Unsupported audio payload type returned by Gemini TTS.") from ex
+
+    def _save_wav(
+            self,
+            filename: str,
+            pcm_bytes: bytes,
+            channels: int = 1,
+            rate: int = 24000,
+            sample_width: int = 2
+    ):
+        """
+        Save raw PCM bytes to a WAV file.
+
+        :param filename: output WAV file path
+        :param pcm_bytes: raw PCM byte data
+        :param channels: number of audio channels (1=mono, 2=stereo)
+        :param rate: sample rate in Hz (e.g., 24000)
+        :param sample_width: sample width in bytes (e.g., 2 for 16-bit)
+        """
+        # Ensure parent directory exists
+        os.makedirs(os.path.dirname(filename), exist_ok=True)
+
+        # Write PCM payload as WAV
+        with wave.open(filename, "wb") as wf:
+            wf.setnchannels(channels)
+            wf.setsampwidth(sample_width)  # bytes per sample (2 -> 16-bit)
+            wf.setframerate(rate)
+            wf.writeframes(pcm_bytes)
+
+    def _normalize_model_name(self, model: str) -> str:
+        """
+        Normalize model id (strip optional 'models/' prefix).
+
+        :param model: model id
+        """
+        try:
+            return model.split("/")[-1]
+        except Exception:
+            return model
+
+    def is_configured(self) -> bool:
+        """
+        Check if provider is configured
+
+        :return: True if configured, False otherwise
+        """
+        api_key = self.plugin.window.core.config.get("api_key_google")
+        return api_key is not None and api_key != ""
+
+    def get_config_message(self) -> str:
+        """
+        Return message to display when provider is not configured
+
+        :return: message
+        """
+        return "Google GenAI API key is not set yet. Please configure it in settings."

pygpt_net/provider/audio_output/openai_tts.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.07 22:00:00                  #
+# Updated Date: 2025.08.29 18:00:00                  #
 # ================================================== #
 
 import os
@@ -51,6 +51,9 @@ class OpenAITextToSpeech(BaseProvider):
             use="audio_tts_whisper_voices",
             description="Specify voice, available voices: "
                         "alloy, echo, fable, onyx, nova, shimmer",
+            urls={
+                "Voices": "https://platform.openai.com/docs/guides/text-to-speech/voice-options"
+            },
         )
 
     def speech(self, text: str) -> str:
@@ -60,15 +63,15 @@ 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')
         allowed_voices = self.plugin.window.core.audio.whisper.get_voices()
-        if model not in self.allowed_models:
-            model = 'tts-1'
-        if voice not in allowed_voices:
-            voice = 'alloy'
+        # if model not in self.allowed_models:
+            # model = 'tts-1'
+        # if voice not in allowed_voices:
+            # voice = 'alloy'
         path = os.path.join(
             self.plugin.window.core.config.path,
             output_file,

pygpt_net/provider/core/config/patch.py

@@ -2355,6 +2355,32 @@ 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
+
+            # < 2.6.31
+            if old < parse_version("2.6.31"):
+                print("Migrating config from < 2.6.31...")
+                if "log.realtime" not in data:
+                    data["log.realtime"] = False
+                if "remote_tools.google.url_ctx" not in data:
+                    data["remote_tools.google.url_ctx"] = False
+                if "audio.input.auto_turn" not in data:
+                    data["audio.input.auto_turn"] = False
+                if "audio.input.vad.prefix" not in data:
+                    data["audio.input.vad.prefix"] = 300
+                if "audio.input.vad.silence" not in data:
+                    data["audio.input.vad.silence"] = 2000
+                updated = True
+
         # update file
         migrated = False
         if updated:

pygpt_net/provider/core/model/patch.py

@@ -763,6 +763,26 @@ 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
+
+            # <  2.6.31 <--- add realtime models
+            if old < parse_version("2.6.31"):
+                print("Migrating models from < 2.6.31...")
+                if "gemini-2.5-flash-preview-native-audio-dialog" not in data:
+                    data["gemini-2.5-flash-preview-native-audio-dialog"] = base_data["gemini-2.5-flash-preview-native-audio-dialog"]
+                if "gpt-realtime" not in data:
+                    data["gpt-realtime"] = base_data["gpt-realtime"]
+                if "gpt-4o-realtime-preview" not in data:
+                    data["gpt-4o-realtime-preview"] = base_data["gpt-4o-realtime-preview"]
+                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/google.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: 2025.08.30 06:00:00                  #
 # ================================================== #
 
 from typing import Optional, List, Dict
@@ -92,13 +92,12 @@ class GoogleLLM(BaseLLM):
         :return: list of models
         """
         items = []
-        client = self.get_client(window)
+        client = window.core.api.google.get_client()
         models_list = client.models.list()
-        if models_list.data:
-            for item in models_list.data:
-                id = item.id.replace("models/", "")
-                items.append({
-                    "id": id,
-                    "name": id,
-                })
+        for item in models_list:
+            id = item.name.replace("models/", "")
+            items.append({
+                "id": id,
+                "name": id,  # TODO: token limit get from API
+            })
         return items

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 = [
             {