Add the possibility to resize images before sending them to the API

addon/globalPlugins/openai/__init__.py

@@ -45,6 +45,13 @@
 	"conversationMode": "boolean(default=True)",
 	"saveSystem": "boolean(default=False)",
 	"advancedMode": "boolean(default=False)",
+	"images": {
+		"maxHeight": "integer(min=0, default=720)",
+		"maxWidth": "integer(min=0, default=0)",
+		"quality": "integer(min=0, max=100, default=85)",
+		"resize": "boolean(default=False)",
+		"resizeInfoDisplayed": "boolean(default=False)"
+	},
 	"renewClient": "boolean(default=False)",
 	"debug": "boolean(default=False)"
 }
@@ -106,47 +113,41 @@ def makeSettings(self, settingsSizer):
 
 		sHelper.addItem(orgSizer)
 
-		mainDialogGroup = _("Main dialog")
-		mainDialogSizer = wx.StaticBoxSizer(wx.VERTICAL, self, label=mainDialogGroup)
+		mainDialogGroupLabel = _("Main dialog")
+		mainDialogSizer = wx.StaticBoxSizer(wx.VERTICAL, self, label=mainDialogGroupLabel)
 		mainDialogBox = mainDialogSizer.GetStaticBox()
-		mainDialog = gui.guiHelper.BoxSizerHelper(self, sizer=mainDialogSizer)
-
-		if not conf["use_org"]:
-			self.org_name.Disable()
-			self.org_key.Disable()
+		mainDialogGroup = gui.guiHelper.BoxSizerHelper(self, sizer=mainDialogSizer)
 
 		label = _("Block the closing using the &escape key")
 		self.blockEscape = wx.CheckBox(
 			self,
 			label=label,
 		)
 		self.blockEscape.SetValue(conf["blockEscapeKey"])
-		mainDialog.addItem(self.blockEscape)
+		mainDialogGroup.addItem(self.blockEscape)
 
 		label = _("Remember the content of the S&ystem field between sessions")
 		self.saveSystem = wx.CheckBox(
 			self,
 			label=label,
 		)
 		self.saveSystem.SetValue(conf["saveSystem"])
-		mainDialog.addItem(self.saveSystem)
+		mainDialogGroup.addItem(self.saveSystem)
 
 		label = _("Enable &advanced settings (including temperature and probability mass)")
 		self.advancedMode = wx.CheckBox(
 			self,
 			label=label,
 		)
 		self.advancedMode.SetValue(conf["advancedMode"])
-		mainDialog.addItem(self.advancedMode)
-
-		sHelper.addItem(mainDialogSizer)
+		mainDialogGroup.addItem(self.advancedMode)
 
-		TTSGroup = _("Text To Speech")
-		TTSSizer = wx.StaticBoxSizer(wx.VERTICAL, self, label=TTSGroup)
+		TTSGroupLabel = _("Text To Speech")
+		TTSSizer = wx.StaticBoxSizer(wx.VERTICAL, self, label=TTSGroupLabel)
 		TTSBox = TTSSizer.GetStaticBox()
-		TTS = gui.guiHelper.BoxSizerHelper(self, sizer=TTSSizer)
+		TTSGroup = gui.guiHelper.BoxSizerHelper(self, sizer=TTSSizer)
 
-		self.voiceList = TTS.addLabeledControl(
+		self.voiceList = TTSGroup.addLabeledControl(
 			_("&Voice:"),
 			wx.Choice,
 			choices=TTS_VOICES,
@@ -156,7 +157,7 @@ def makeSettings(self, settingsSizer):
 			itemToSelect = TTS_VOICES.index(conf["TTSVoice"])
 		self.voiceList.SetSelection(itemToSelect)
 
-		self.modelList = TTS.addLabeledControl(
+		self.modelList = TTSGroup.addLabeledControl(
 			_("&Model:"),
 			wx.Choice,
 			choices=TTS_MODELS,
@@ -168,10 +169,67 @@ def makeSettings(self, settingsSizer):
 
 		sHelper.addItem(TTSSizer)
 
+		imageGroupLabel = _("Images")
+		imageSizer = wx.StaticBoxSizer(wx.VERTICAL, self, label=imageGroupLabel)
+		imageBox = imageSizer.GetStaticBox()
+		imageGroup = gui.guiHelper.BoxSizerHelper(self, sizer=imageSizer)
+
+		label = _("&Resize images before sending them to the API")
+		self.resize = imageGroup.addItem(
+			wx.CheckBox(
+				imageBox,
+				label=label,
+			)
+		)
+		self.resize.SetValue(conf["images"]["resize"])
+		self.resize.Bind(
+			wx.EVT_CHECKBOX,
+			self.onResize
+		)
+
+		label = _("Maximum &width (0 to resize proportionally to the height):")
+		self.maxWidth = imageGroup.addLabeledControl(
+			label,
+			wx.SpinCtrl,
+			min=0,
+			max=2000
+		)
+		self.maxWidth.SetValue(conf["images"]["maxWidth"])
+
+		label = _("Maximum &height (0 to resize proportionally to the width):")
+		self.maxHeight = imageGroup.addLabeledControl(
+			label,
+			wx.SpinCtrl,
+			min=0,
+			max=2000
+		)
+		self.maxHeight.SetValue(conf["images"]["maxHeight"])
+
+		label = _("&Quality for JPEG images (0 [worst] to 95 [best], values above 95 should be avoided):")
+		self.quality = imageGroup.addLabeledControl(
+			label,
+			wx.SpinCtrl,
+			min=1,
+			max=100
+		)
+		self.quality.SetValue(conf["images"]["quality"])
+
+		sHelper.addItem(imageSizer)
+
+		sHelper.addItem(mainDialogSizer)
+
+		self.onUseOrg(None)
+		self.onResize(None)
+
 	def onUseOrg(self, evt):
 		self.org_name.Enable(self.use_org.GetValue())
 		self.org_key.Enable(self.use_org.GetValue())
 
+	def onResize(self, evt):
+		self.maxWidth.Enable(self.resize.GetValue())
+		self.maxHeight.Enable(self.resize.GetValue())
+		self.quality.Enable(self.resize.GetValue())
+
 	def onSave(self):
 		api_key = self.APIKey.GetValue().strip()
 		api_key_manager.save_api_key(api_key)
@@ -194,9 +252,15 @@ def onSave(self):
 		conf["renewClient"] = True
 		conf["saveSystem"] = self.saveSystem.GetValue()
 		conf["advancedMode"] = self.advancedMode.GetValue()
+
 		conf["TTSVoice"] = self.voiceList.GetString(self.voiceList.GetSelection())
 		conf["TTSModel"] = self.modelList.GetString(self.modelList.GetSelection())
 
+		conf["images"]["resize"] = self.resize.GetValue()
+		conf["images"]["maxWidth"] = self.maxWidth.GetValue()
+		conf["images"]["maxHeight"] = self.maxHeight.GetValue()
+		conf["images"]["quality"] = self.quality.GetValue()
+
 
 class GlobalPlugin(globalPluginHandler.GlobalPlugin):
 

addon/globalPlugins/openai/imagehelper.py

@@ -7,11 +7,44 @@
 additionalLibsPath = os.path.join(ADDON_DIR, "lib")
 sys.path.insert(0, additionalLibsPath)
 from openai import OpenAI
+from PIL import Image
 sys.path.remove(additionalLibsPath)
 
-# Borrowed from <https://platform.openai.com/docs/guides/vision>
+def resize_image(
+	src: str,
+	max_width: int = 0,
+	max_height: int = 0,
+	quality: int = 85,
+	target: str = "Compressed.PNG"
+):
+	"""
+	Compress an image and save it to a specified file by resizing according to
+	given maximum dimensions and adjusting the quality.
+
+	@param src: path to the source image.
+	@param max_width: Maximum width for the compressed image. If 0, only `max_height` is used to calculate the ratio.
+	@param max_height: Maximum height for the compressed image. If 0, only `max_width` is used to calculate the ratio.
+	@param quality: the quality of the compressed image
+	@param target: output path for the compressed image
+	@return: True if the image was successfully compressed and saved, False otherwise
+	"""
+	if max_width <= 0 and max_height <= 0:
+		return False
+	image = Image.open(src)
+	orig_width, orig_height = image.size
+	if max_width > 0 and max_height > 0:
+		ratio = min(max_width / orig_width, max_height / orig_height)
+	elif max_width > 0:
+		ratio = max_width / orig_width
+	else:
+		ratio = max_height / orig_height
+	new_width = int(orig_width * ratio)
+	new_height = int(orig_height * ratio)
+	resized_image = image.resize((new_width, new_height), Image.ANTIALIAS)
+	resized_image.save(target, optimize=True, quality=quality)
+	return True
+
 
-# Function to encode the image
 def encode_image(image_path):
 	with open(image_path, "rb") as image_file:
 		return base64.b64encode(image_file.read()).decode('utf-8')

addon/globalPlugins/openai/maindialog.py

@@ -16,7 +16,7 @@
 import ui
 from logHandler import log
 from .consts import ADDON_DIR, DATA_DIR
-from .imagehelper import describeFromImageFileList, encode_image
+from .imagehelper import resize_image, describeFromImageFileList, encode_image
 additionalLibsPath = os.path.join(ADDON_DIR, "lib")
 sys.path.insert(0, additionalLibsPath)
 import openai
@@ -767,6 +767,18 @@ def onOk(self, evt):
 				wx.OK|wx.ICON_ERROR
 			)
 			return
+		if (
+			model.name == MODEL_VISION
+			and not self.conf["images"]["resize"]
+			and not self.conf["images"]["resizeInfoDisplayed"]
+		):
+			msg = _("Be aware that the add-on may auto-resize images before API submission to lower request sizes and costs. Adjust this feature in the Open AI settings if needed. This message won't show again.")
+			gui.messageBox(
+				msg,
+				_("Open AI"),
+				wx.OK|wx.ICON_INFORMATION
+			)
+			self.conf["images"]["resizeInfoDisplayed"] = True
 		system = self.systemText.GetValue().strip()
 		if self.conf["saveSystem"] and system != self._lastSystem and system:
 			self.data["system"] = system
@@ -930,6 +942,7 @@ def getImages(
 		self,
 		pathList: list = None
 	) -> list:
+		conf = self.conf
 		if not pathList:
 			pathList = self.pathList
 		images = []
@@ -938,6 +951,16 @@ def getImages(
 			if url_re.match(path):
 				images.append({"type": "image_url", "image_url": {"url": path}})
 			elif os.path.isfile(path):
+				if conf["images"]["resize"]:
+					path_ = os.path.join(DATA_DIR, "last_resized.jpg")
+					resize_image(
+						path,
+						max_width=conf["images"]["maxWidth"],
+						max_height=conf["images"]["maxHeight"],
+						quality=conf["images"]["quality"],
+						target=path_
+					)
+					path = path_
 				base64_image = encode_image(path)
 				format = path.split(".")[-1]
 				mime_type = f"image/{format}"