pygpt-net 2.6.62__py3-none-any.whl → 2.6.64__py3-none-any.whl

pygpt_net/provider/api/google/music.py

@@ -0,0 +1,375 @@
+#!/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: 2025.09.27 09:30:00                  #
+# ================================================== #
+
+import base64
+import datetime
+import json
+import os
+import shutil
+import subprocess
+from typing import Optional, Dict, Any, List
+
+import requests
+from PySide6.QtCore import QObject, Signal, QRunnable, Slot
+from google import genai
+
+from pygpt_net.core.events import KernelEvent
+from pygpt_net.core.bridge.context import BridgeContext
+from pygpt_net.item.ctx import CtxItem
+from pygpt_net.utils import trans
+
+
+class Music:
+    """
+    Text-to-music generation for Google (Lyria) in the "image" mode.
+
+    This class encapsulates the whole music generation flow using Vertex AI Lyria.
+    It mirrors the architecture used for images and videos (worker + signals),
+    and saves outputs to disk as WAV by default, with optional ffmpeg transcoding to MP3/MP4.
+    """
+
+    def __init__(self, window=None):
+        self.window = window
+        self.worker: Optional[MusicWorker] = None
+
+    def is_music_model(self, model_id: str) -> bool:
+        """
+        Heuristic check whether selected model is a music model (Lyria family).
+
+        :param model_id: model id
+        :return: True if Lyria/music model, False otherwise
+        """
+        mid = (model_id or "").lower()
+        return "lyria" in mid or "music" in mid
+
+    def generate(
+            self,
+            context: BridgeContext,
+            extra: Optional[Dict[str, Any]] = None,
+            sync: bool = True
+    ) -> bool:
+        """
+        Generate music using Vertex AI Lyria.
+
+        :param context: BridgeContext with prompt, model, attachments (ignored)
+        :param extra: extra parameters:
+            - num: int, number of samples (maps to sample_count if no seed)
+            - seed: int, generation seed (mutually exclusive with sample_count)
+            - negative_prompt: str, negative prompt to exclude elements
+            - format: str, 'wav' (default), 'mp3', or 'mp4'
+            - inline: bool, inline mode
+        :param sync: run synchronously (blocking) if True
+        :return: True if started
+        """
+        extra = extra or {}
+        ctx = context.ctx or CtxItem()
+        model = context.model
+        prompt = context.prompt or ""
+        inline = bool(extra.get("inline", False))
+
+        worker = MusicWorker()
+        worker.window = self.window
+        worker.client = self.window.core.api.google.get_client()
+        worker.ctx = ctx
+
+        # config
+        worker.model = (model.id if model else "lyria-002")  # Lyria model id or resource
+        worker.input_prompt = prompt
+        worker.negative_prompt = extra.get("negative_prompt") or None
+
+        # sample_count vs seed (mutually exclusive)
+        worker.num = int(extra.get("num", 1))
+        seed = extra.get("seed")
+        worker.seed = int(seed) if seed not in (None, "") else None
+
+        # preferred output format
+        worker.out_format = str(extra.get("format") or self._default_format()).lower()
+
+        # optional prompt improvement
+        prompt_model = self.window.core.models.from_defaults()
+        tmp = self.window.core.config.get('music.prompt_model') or self.window.core.config.get('video.prompt_model')
+        if self.window.core.models.has(tmp):
+            prompt_model = self.window.core.models.get(tmp)
+        worker.model_prompt = prompt_model
+        worker.system_prompt = self.window.core.prompt.get('music') or self.window.core.prompt.get('video')
+        worker.raw = bool(self.window.core.config.get('img_raw'))
+
+        worker.inline = inline
+
+        self.worker = worker
+        # Reuse video handlers for UX consistency (status/messages/download area)
+        self.worker.signals.finished.connect(self.window.core.video.handle_finished)
+        self.worker.signals.finished_inline.connect(self.window.core.video.handle_finished_inline)
+        self.worker.signals.status.connect(self.window.core.video.handle_status)
+        self.worker.signals.error.connect(self.window.core.video.handle_error)
+
+        if sync or not self.window.controller.kernel.async_allowed(ctx):
+            self.worker.run()
+            return True
+
+        # Use video busy state for unified UX
+        self.window.dispatch(KernelEvent(KernelEvent.STATE_BUSY, {"id": "video"}))
+        self.window.threadpool.start(self.worker)
+        return True
+
+    def _default_format(self) -> str:
+        """
+        Determine default preferred output format for saved files.
+        """
+        # try config override
+        try:
+            fmt = self.window.core.config.get('music.format')
+            if isinstance(fmt, str) and fmt.strip():
+                return fmt.strip()
+        except Exception:
+            pass
+        # default to mp3 as most interoperable
+        return "mp3"
+
+
+class MusicSignals(QObject):
+    finished = Signal(object, list, str)         # ctx, paths, prompt
+    finished_inline = Signal(object, list, str)  # ctx, paths, prompt
+    status = Signal(object)                      # message
+    error = Signal(object)                       # exception
+
+
+class MusicWorker(QRunnable):
+    """
+    Worker that calls the Vertex AI Lyria REST API (predict) to generate audio,
+    then saves to disk. Transcodes to MP3/MP4 if ffmpeg is available and requested.
+    """
+
+    def __init__(self, *args, **kwargs):
+        super().__init__()
+        self.signals = MusicSignals()
+        self.window = None
+        self.client: Optional[genai.Client] = None
+        self.ctx: Optional[CtxItem] = None
+
+        # inputs
+        self.model: str = "lyria-002"
+        self.input_prompt: str = ""
+        self.negative_prompt: Optional[str] = None
+        self.num: int = 1
+        self.seed: Optional[int] = None
+        self.out_format: str = "mp3"  # wav | mp3 | mp4
+
+        # prompt improvement
+        self.model_prompt = None
+        self.system_prompt = ""
+        self.raw = False
+
+        # ui
+        self.inline: bool = False
+
+    @Slot()
+    def run(self):
+        try:
+            # Validate Vertex configuration first
+            if not self._using_vertex():
+                raise RuntimeError(
+                    "Vertex AI is required for music (Lyria). Enable Vertex in settings and configure credentials."
+                )
+
+            # Optional prompt enhancement via LLM
+            if not self.raw and self.input_prompt:
+                try:
+                    self.signals.status.emit(trans('vid.status.prompt.wait'))
+                    bridge_context = BridgeContext(
+                        prompt=self.input_prompt,
+                        system_prompt=self.system_prompt,
+                        model=self.model_prompt,
+                        max_tokens=200,
+                        temperature=1.0,
+                    )
+                    ev = KernelEvent(KernelEvent.CALL, {'context': bridge_context, 'extra': {}})
+                    self.window.dispatch(ev)
+                    resp = ev.data.get('response')
+                    if resp:
+                        self.input_prompt = resp
+                except Exception as e:
+                    # non-fatal
+                    self.signals.error.emit(e)
+                    self.signals.status.emit(trans('vid.status.prompt.error') + ": " + str(e))
+
+            # Build request
+            project = os.getenv("GOOGLE_CLOUD_PROJECT", "")
+            location = os.getenv("GOOGLE_CLOUD_LOCATION", "us-central1")
+            model_id = self._normalize_model_id(self.model)
+
+            url = f"https://{location}-aiplatform.googleapis.com/v1/projects/{project}/locations/{location}/publishers/google/models/{model_id}:predict"
+            token = self._get_access_token()
+
+            headers = {
+                "Authorization": f"Bearer {token}",
+                "Content-Type": "application/json",
+            }
+
+            # sample_count vs seed (mutually exclusive per Lyria API)
+            instances: Dict[str, Any] = {
+                "prompt": self.input_prompt,
+            }
+            if self.negative_prompt:
+                instances["negative_prompt"] = self.negative_prompt
+
+            params: Dict[str, Any] = {}
+            if self.seed is not None and self.num > 1:
+                # Keep API valid: if seed is set, do not set sample_count
+                self.signals.status.emit("Seed provided; generating a single seeded sample (sample_count ignored).")
+            if self.seed is not None:
+                instances["seed"] = int(self.seed)
+            elif self.num > 1:
+                params["sample_count"] = int(self.num)
+
+            body = {
+                "instances": [instances],
+                "parameters": params,
+            }
+
+            self.signals.status.emit(trans('vid.status.generating') + f": {self.input_prompt}...")
+
+            # Call REST API
+            resp = requests.post(url, headers=headers, data=json.dumps(body), timeout=120)
+            if resp.status_code != 200:
+                raise RuntimeError(f"Lyria API error: {resp.status_code} {resp.text}")
+
+            data = resp.json() or {}
+            preds = data.get("predictions") or []
+            if not preds:
+                raise RuntimeError("No audio predictions returned.")
+
+            # Decode and save each audio clip
+            paths: List[str] = []
+            for idx, p in enumerate(preds):
+                # API returns base64-encoded WAV in audioContent
+                b64 = p.get("audioContent") or ""
+                mime = p.get("mimeType") or "audio/wav"
+                if not b64:
+                    continue
+                wav_bytes = base64.b64decode(b64)
+
+                saved_path = self._save(idx, wav_bytes, mime)
+                if saved_path:
+                    paths.append(saved_path)
+
+            if self.inline:
+                self.signals.finished_inline.emit(self.ctx, paths, self.input_prompt)
+            else:
+                self.signals.finished.emit(self.ctx, paths, self.input_prompt)
+
+        except Exception as e:
+            self.signals.error.emit(e)
+        finally:
+            self._cleanup()
+
+    # ---------- helpers ----------
+
+    def _using_vertex(self) -> bool:
+        """Check if Vertex AI is active via env variable set by ApiGoogle.setup_env()."""
+        val = os.getenv("GOOGLE_GENAI_USE_VERTEXAI") or ""
+        return str(val).lower() in ("1", "true", "yes", "y")
+
+    def _normalize_model_id(self, model_id: str) -> str:
+        """
+        Normalize model id to the publisher form 'lyria-002'.
+        Accepts inputs like 'lyria-002', 'models/lyria-002', or full resource names.
+        """
+        if not model_id:
+            return "lyria-002"
+        # get trailing token after '/'
+        tail = model_id.split("/")[-1]
+        if tail.startswith("lyria"):
+            return tail
+        # fallback
+        return "lyria-002"
+
+    def _get_access_token(self) -> str:
+        """
+        Obtain OAuth2 access token for Vertex AI (cloud-platform scope) using ADC.
+        """
+        try:
+            import google.auth
+            from google.auth.transport.requests import Request as GRequest
+            creds, _ = google.auth.default(scopes=["https://www.googleapis.com/auth/cloud-platform"])
+            creds.refresh(GRequest())
+            return creds.token
+        except Exception as e:
+            # As a fallback, allow passing a pre-fetched token via env var.
+            token = os.getenv("VERTEX_AI_TOKEN") or os.getenv("GCP_ACCESS_TOKEN") or ""
+            if token:
+                return token
+            raise e
+
+    def _save(self, idx: int, wav_bytes: bytes, mime: str) -> Optional[str]:
+        """
+        Save audio to disk. Primary save as WAV; then optionally transcode to MP3/MP4 if requested.
+        """
+        # Build base filename
+        safe = self.window.core.video.make_safe_filename(self.input_prompt)
+        base_name = (
+            datetime.date.today().strftime("%Y-%m-%d") + "_" +
+            datetime.datetime.now().strftime("%H-%M-%S") + "-" +
+            safe + "-" + str(idx + 1)
+        )
+
+        # Directories: prefer 'audio' user dir if available, else fallback to 'video'
+        try:
+            out_dir = self.window.core.config.get_user_dir("audio")
+        except Exception:
+            out_dir = self.window.core.config.get_user_dir("video")
+
+        os.makedirs(out_dir, exist_ok=True)
+
+        # Always persist a WAV first (what API returns)
+        wav_path = os.path.join(out_dir, base_name + ".wav")
+        self.signals.status.emit(trans('vid.status.downloading') + f" ({idx + 1} / {max(1, self.num)}) -> {wav_path}")
+        with open(wav_path, "wb") as f:
+            f.write(wav_bytes)
+
+        # Transcode if needed
+        fmt = (self.out_format or "mp3").lower()
+        if fmt == "wav":
+            return wav_path
+
+        ffmpeg = shutil.which("ffmpeg") or shutil.which("ffmpeg.exe")
+        if not ffmpeg:
+            # No ffmpeg -> keep WAV
+            self.signals.status.emit("ffmpeg not found. Saved WAV output only.")
+            return wav_path
+
+        if fmt == "mp3":
+            out_path = os.path.join(out_dir, base_name + ".mp3")
+            cmd = [ffmpeg, "-y", "-i", wav_path, "-vn", "-codec:a", "libmp3lame", "-b:a", "192k", out_path]
+        elif fmt == "mp4":
+            out_path = os.path.join(out_dir, base_name + ".mp4")
+            # audio-only MP4 (AAC)
+            cmd = [ffmpeg, "-y", "-i", wav_path, "-vn", "-c:a", "aac", "-b:a", "192k", out_path]
+        else:
+            # Unknown format -> keep WAV
+            return wav_path
+
+        try:
+            subprocess.run(cmd, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, check=True)
+            return out_path
+        except Exception:
+            # If transcoding fails, keep WAV
+            return wav_path
+
+    def _cleanup(self):
+        """Cleanup signals object."""
+        sig = self.signals
+        self.signals = None
+        if sig is not None:
+            try:
+                sig.deleteLater()
+            except RuntimeError:
+                pass

pygpt_net/ui/widget/dialog/confirm.py

@@ -6,11 +6,11 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.26 10:00:00                  #
+# Updated Date: 2025.09.27 09:30:00                  #
 # ================================================== #
 
 import sys
-from PySide6.QtCore import Qt
+from PySide6.QtCore import Qt, QTimer
 from PySide6.QtWidgets import QDialog, QLabel, QHBoxLayout, QVBoxLayout, QPushButton
 
 from pygpt_net.utils import trans
@@ -44,11 +44,7 @@ class ConfirmDialog(QDialog):
 
         # Always make the neutral action (No/Cancel) the default/active one.
         # This ensures Enter triggers the safe option by default.
-        self.window.ui.nodes['dialog.confirm.btn.no'].setAutoDefault(True)
-        self.window.ui.nodes['dialog.confirm.btn.no'].setDefault(True)
-        self.window.ui.nodes['dialog.confirm.btn.no'].setFocus()
-        self.window.ui.nodes['dialog.confirm.btn.yes'].setAutoDefault(False)
-        self.window.ui.nodes['dialog.confirm.btn.yes'].setDefault(False)
+        self._apply_neutral_default()
 
         # Bottom button row with platform-specific ordering
         # Windows: affirmative on the left, neutral on the right
@@ -85,4 +81,34 @@ class ConfirmDialog(QDialog):
         :param event: close event
         """
         self.window.controller.dialogs.confirm.dismiss(self.type, self.id)
-        super(ConfirmDialog, self).closeEvent(event)
+        super(ConfirmDialog, self).closeEvent(event)
+
+    def showEvent(self, event):
+        """
+        Ensure neutral button is default/active on every show.
+
+        Using a single-shot timer defers focus/default restoration until
+        after the dialog becomes visible, which prevents focus being stolen
+        by the window manager.
+        """
+        super(ConfirmDialog, self).showEvent(event)
+        QTimer.singleShot(0, self._apply_neutral_default)
+
+    def _apply_neutral_default(self):
+        """
+        Set the neutral action (No/Cancel) as default and active.
+        Always called on construction and each time the dialog is shown.
+        """
+        btn_no = self.window.ui.nodes.get('dialog.confirm.btn.no')
+        btn_yes = self.window.ui.nodes.get('dialog.confirm.btn.yes')
+        if not btn_no or not btn_yes:
+            return
+
+        # Make sure affirmative button cannot become default by leftover state
+        btn_yes.setAutoDefault(False)
+        btn_yes.setDefault(False)
+
+        # Make neutral (No/Cancel) the active default and take focus
+        btn_no.setAutoDefault(True)
+        btn_no.setDefault(True)
+        btn_no.setFocus()

pygpt_net/ui/widget/option/combo.py

@@ -11,26 +11,118 @@
 
 from PySide6.QtCore import Qt
 from PySide6.QtWidgets import QHBoxLayout, QWidget, QComboBox
-from PySide6.QtGui import QFontMetrics
+from PySide6.QtGui import QFontMetrics, QStandardItem, QStandardItemModel  # keep existing imports, extend with items
 
 from pygpt_net.utils import trans
 
 class SeparatorComboBox(QComboBox):
-    """A combo box that supports adding separator items."""
+    """A combo box that supports adding separator items and prevents selecting them."""
+
+    def __init__(self, parent=None):
+        super().__init__(parent)
+        # Custom role used to mark separator rows without interfering with existing UserRole data
+        self._SEP_ROLE = Qt.UserRole + 1000
+        # Internal guard to avoid recursive index changes
+        self._block_guard = False
 
     def addSeparator(self, text):
         """
-        Adds a separator item to the combo box.
+        Adds a separator item to the combo box that cannot be selected.
+        This keeps separators visible but disabled/unselectable.
 
         :param text: The text to display for the separator.
         """
-        index = self.count()
-        self.addItem(text)
+        model = self.model()
+        if isinstance(model, QStandardItemModel):
+            item = QStandardItem(text)
+            # Disable and make the row unselectable
+            item.setFlags(item.flags() & ~Qt.ItemIsEnabled & ~Qt.ItemIsSelectable)
+            # Mark explicitly as separator using custom role
+            item.setData(True, self._SEP_ROLE)
+            model.appendRow(item)
+        else:
+            # Fallback: keep previous behavior and additionally tag item with custom role
+            index = self.count()
+            self.addItem(text)
+            try:
+                role = Qt.UserRole - 1
+                self.setItemData(index, 0, role)  # legacy approach used sometimes to indicate non-enabled
+            except Exception:
+                pass
+            # Tag as separator via custom role for later checks
+            self.setItemData(index, True, self._SEP_ROLE)
+
+    def is_separator(self, index: int) -> bool:
+        """Returns True if item at index is a separator."""
+        if index < 0 or index >= self.count():
+            return False
         try:
-            role = Qt.UserRole - 1
-            self.setItemData(index, 0, role)
-        except:
+            if self.itemData(index, self._SEP_ROLE):
+                return True
+        except Exception:
             pass
+        # Fallback: check flags (works with item models)
+        try:
+            idx = self.model().index(index, self.modelColumn(), self.rootModelIndex())
+            flags = self.model().flags(idx)
+            if not (flags & Qt.ItemIsEnabled) or not (flags & Qt.ItemIsSelectable):
+                return True
+        except Exception:
+            pass
+        return False
+
+    def first_valid_index(self) -> int:
+        """Returns the first non-separator index, or -1 if none."""
+        for i in range(self.count()):
+            if not self.is_separator(i):
+                return i
+        return -1
+
+    def _sanitize_index(self, index: int) -> int:
+        """Returns a corrected non-separator index, or -1 if none available."""
+        if index is None:
+            index = -1
+        if index < 0 or index >= self.count():
+            return self.first_valid_index()
+        if self.is_separator(index):
+            # Prefer the next valid item; if none, scan backwards; else -1
+            for i in range(index + 1, self.count()):
+                if not self.is_separator(i):
+                    return i
+            for i in range(index - 1, -1, -1):
+                if not self.is_separator(i):
+                    return i
+            return -1
+        return index
+
+    def ensure_valid_current(self) -> int:
+        """
+        Ensures the current index is not a separator.
+        Returns the final valid index (or -1) after correction.
+        """
+        current = super().currentIndex()
+        corrected = self._sanitize_index(current)
+        if corrected != current:
+            try:
+                self._block_guard = True
+                super().setCurrentIndex(corrected if corrected != -1 else -1)
+            finally:
+                self._block_guard = False
+        return corrected
+
+    def setCurrentIndex(self, index: int) -> None:
+        """
+        Prevent setting the current index to a separator from any caller.
+        """
+        if self._block_guard:
+            # When guarded, pass through without checks to avoid recursion
+            return super().setCurrentIndex(index)
+        corrected = self._sanitize_index(index)
+        try:
+            self._block_guard = True
+            super().setCurrentIndex(corrected if corrected != -1 else -1)
+        finally:
+            self._block_guard = False
 
 
 class NoScrollCombo(SeparatorComboBox):
@@ -116,7 +208,11 @@ class OptionCombo(QWidget):
                         else:
                             self.combo.addItem(value, key)
                 else:
-                    self.combo.addItem(item, item)
+                    # Support simple string keys including "separator::" entries
+                    if isinstance(item, str) and item.startswith("separator::"):
+                        self.combo.addSeparator(item.split("separator::", 1)[1])
+                    else:
+                        self.combo.addItem(item, item)
         elif type(self.keys) is dict:
             for key, value in self.keys.items():
                 if not isinstance(key, str):
@@ -126,6 +222,32 @@ class OptionCombo(QWidget):
                 else:
                     self.combo.addItem(value, key)
 
+        # Ensure a valid non-separator selection after population
+        self._apply_initial_selection()
+
+    def _apply_initial_selection(self):
+        """
+        Ensures that after building the list the combobox does not end up on a separator.
+        Prefers self.current_id if present; otherwise selects the first valid non-separator.
+        Signals are suppressed during this operation.
+        """
+        # lock on_change during initial selection
+        prev_locked = self.locked
+        self.locked = True
+        try:
+            index = -1
+            if self.current_id is not None and self.current_id != "":
+                index = self.combo.findData(self.current_id)
+            if index == -1:
+                index = self.combo.first_valid_index()
+            if index != -1:
+                self.combo.setCurrentIndex(index)
+            else:
+                # No valid items, clear selection
+                self.combo.setCurrentIndex(-1)
+        finally:
+            self.locked = prev_locked
+
     def set_value(self, value):
         """
         Set value
@@ -137,6 +259,9 @@ class OptionCombo(QWidget):
         index = self.combo.findData(value)
         if index != -1:
             self.combo.setCurrentIndex(index)
+        else:
+            # If requested value is not present, keep current selection but make sure it is valid.
+            self.combo.ensure_valid_current()
 
     def get_value(self):
         """
@@ -159,6 +284,8 @@ class OptionCombo(QWidget):
         self.option["keys"] = keys
         self.combo.clear()
         self.update()
+        # After rebuilding, guarantee a non-separator selection
+        self.combo.ensure_valid_current()
         if lock:
             self.locked = False
 
@@ -171,10 +298,21 @@ class OptionCombo(QWidget):
         """
         if self.locked:
             return
+
+        # If somehow a separator got focus, correct it immediately and do not propagate invalid IDs
+        if self.combo.is_separator(index):
+            self.locked = True
+            corrected = self.combo.ensure_valid_current()
+            self.locked = False
+            if corrected == -1:
+                # Nothing valid to select
+                self.current_id = None
+                return
+            index = corrected
+
         self.current_id = self.combo.itemData(index)
         self.window.controller.config.combo.on_update(self.parent_id, self.id, self.option, self.current_id)
 
     def fit_to_content(self):
         """Fit to content"""
-        self.combo.setSizeAdjustPolicy(QComboBox.AdjustToContents)
-
+        self.combo.setSizeAdjustPolicy(QComboBox.AdjustToContents)