Fix Image Search and Improve Desktop App

### Fix Image Search
  - Do not use XMP metadata by default for image search
    - It seems to be buggy currently. The returned results do not make sense with XMP metadata enabled

### Fix Image Search using Desktop App
  - Fix configuring Image Search via Desktop GUI
    - Set `input-directories`, instead of unused `input-files` for `content-type.image` in `khoj.yml`
  - Fix running Image Search via Desktop apps. 
    - Previously the transformers wasn't getting packaged into the app by pyinstaller
    - This is required by image search to run. So the desktop apps would fail to start when image search was enabled
    - Resolves #68
  - Append selected files, directories via "Add" button in Desktop GUI
    - This allows selecting multiple files, directories using Desktop GUI
    - Previously selecting multiple image directories had to be entered manually

### Improve Desktop App
  - Show Splash Screen to Desktop on App Initialization
    - The app takes a while to load during first run 
    - A splash screen signals that app is loading and not being unresponsive
    - Note: _Pyinstaller only supports splash screens on Windows, Linux. Not on Macs._
  - Add Khoj icon to the Windows, Linux app. Windows expects a `.ico` icon type
  - Only exclude `libtorch_{cuda, cpu, python}` on Linux machine
    - Seems those libraries are being used on Mac (and maybe Windows). 
    - Linux is where the app size benefits from removing these is maximum anyway
  - Fix PyInstaller Warnings on App Start
    - The warning show up as annoying error popups on Windows

.github/workflows/release.yml

@@ -35,7 +35,7 @@ jobs:
         shell: bash
         run: |
           if [ "$RUNNER_OS" == "Linux" ]; then
-            sudo apt install libegl1 libxcb-xinerama0 -y
+            sudo apt install libegl1 libxcb-xinerama0 python3-tk -y
           fi
           python -m pip install --upgrade pip
           pip install pyinstaller

Khoj.spec

@@ -2,8 +2,12 @@
 from os.path import join
 from platform import system
 from PyInstaller.utils.hooks import copy_metadata
+import sysconfig
 
-datas = [('src/interface/web', 'src/interface/web')]
+datas = [
+    ('src/interface/web', 'src/interface/web'),
+    (f'{sysconfig.get_paths()["purelib"]}/transformers', 'transformers')
+]
 datas += copy_metadata('tqdm')
 datas += copy_metadata('regex')
 datas += copy_metadata('requests')
@@ -12,10 +16,8 @@ datas += copy_metadata('filelock')
 datas += copy_metadata('numpy')
 datas += copy_metadata('tokenizers')
 
-
 block_cipher = None
 
-
 a = Analysis(
     ['src/main.py'],
     pathex=[],
@@ -32,17 +34,56 @@ a = Analysis(
     noarchive=False,
 )
 
-# Filter out unused, duplicate shared libs
+# Filter out unused and/or duplicate shared libs
-extension = {'Windows': '.dll', 'Darwin': '.dylib', 'Linux': '.so'}[system()]
 torch_lib_paths = {
-    join('torch', 'lib', 'libtorch_cuda' + extension),
+    join('torch', 'lib', 'libtorch_cuda.so'),
-    join('torch', 'lib', 'libtorch_cpu' + extension),
+    join('torch', 'lib', 'libtorch_cpu.so'),
-    join('torch', 'lib', 'libtorch_python' + extension)
 }
 a.datas = [entry for entry in a.datas if not entry[0] in torch_lib_paths]
 
+a.datas = [entry for entry in a.datas if not 'torch/_C.cp' in entry[0]]
+a.datas = [entry for entry in a.datas if not 'torch/_dl.cp' in entry[0]]
+
 pyz = PYZ(a.pure, a.zipped_data, cipher=block_cipher)
 
+if system() != 'Darwin':
+    # Add Splash screen to show on app launch
+    splash = Splash(
+        'src/interface/web/assets/icons/favicon-144x144.png',
+        binaries=a.binaries,
+        datas=a.datas,
+        text_pos=(10, 50),
+        text_size=12,
+        text_color='blue',
+        minify_script=True,
+        always_on_top=True
+    )
+
+    exe = EXE(
+        pyz,
+        a.scripts,
+        a.binaries,
+        a.zipfiles,
+        a.datas,
+        splash,
+        splash.binaries,
+        [],
+        name='Khoj',
+        debug=False,
+        bootloader_ignore_signals=False,
+        strip=False,
+        upx=True,
+        upx_exclude=[],
+        runtime_tmpdir=None,
+        console=False,
+        disable_windowed_traceback=False,
+        argv_emulation=False,
+        target_arch='x86_64',
+        codesign_identity=None,
+        entitlements_file=None,
+        icon='src/interface/web/assets/icons/favicon-144x144.ico',
+    )
+else:
     exe = EXE(
         pyz,
         a.scripts,
@@ -65,8 +106,6 @@ exe = EXE(
         entitlements_file=None,
         icon='src/interface/web/assets/icons/favicon.icns',
     )
-
-if system() == 'Darwin':
     app = BUNDLE(
         exe,
         name='Khoj.app',

config/khoj_sample.yml

@@ -22,7 +22,7 @@ content-type:
     input-filter:      # /path/to/images/*.jpg  REQUIRED IF input-directories IS NOT SET
     embeddings-file: "~/.khoj/content/image/image_embeddings.pt"
     batch-size: 50
-    use-xmp-metadata: true
+    use-xmp-metadata: false
 
   music:
     input-files:  # ["/path/to/music-file.org"] REQUIRED IF input-filter IS NOT SET OR

src/interface/desktop/file_browser.py

@@ -29,7 +29,7 @@ class FileBrowser(QtWidgets.QWidget):
         self.lineEdit.textChanged.connect(self.updateFieldHeight)
         layout.addWidget(self.lineEdit)
         
-        self.button = QtWidgets.QPushButton('Select')
+        self.button = QtWidgets.QPushButton('Add')
         self.button.clicked.connect(self.storeFilesSelectedInFileDialog)
         layout.addWidget(self.button)
         layout.addStretch()
@@ -47,7 +47,7 @@ class FileBrowser(QtWidgets.QWidget):
             return 'Images (*.jp[e]g)'
 
     def storeFilesSelectedInFileDialog(self):
-        filepaths = []
+        filepaths = self.getPaths()
         if self.search_type == SearchType.Image:
             filepaths.append(QtWidgets.QFileDialog.getExistingDirectory(self, caption='Choose Folder',
                                                     directory=self.dirpath))

src/interface/desktop/main_window.py

@@ -185,6 +185,9 @@ class MainWindow(QtWidgets.QMainWindow):
                         default_search_config = self.get_default_config(search_type = child.search_type)
                         self.new_config['content-type'][child.search_type.value] = merge_dicts(current_search_config, default_search_config)
                 elif isinstance(child, FileBrowser) and child.search_type in self.new_config['content-type']:
+                    if child.search_type.value == SearchType.Image:
+                        self.new_config['content-type'][child.search_type.value]['input-directories'] = child.getPaths() if child.getPaths() != [] else None
+                    else:
                         self.new_config['content-type'][child.search_type.value]['input-files'] = child.getPaths() if child.getPaths() != [] else None
 
     def update_processor_settings(self):

src/main.py

@@ -1,7 +1,8 @@
 # Standard Packages
-from platform import system
+import os
 import signal
 import sys
+from platform import system
 
 # External Packages
 import uvicorn
@@ -26,6 +27,9 @@ app.include_router(router)
 
 
 def run():
+    # Turn Tokenizers Parallelism Off. App does not support it.
+    os.environ["TOKENIZERS_PARALLELISM"] = 'false'
+
     # Load config from CLI
     state.cli_args = sys.argv[1:]
     args = cli(state.cli_args)
@@ -67,6 +71,18 @@ def run():
         # Start Application
         server.start()
         gui.aboutToQuit.connect(server.terminate)
+
+        # Close Splash Screen if still open
+        if system() != 'Darwin':
+            try:
+                import pyi_splash
+                # Update the text on the splash screen
+                pyi_splash.update_text("Khoj setup complete")
+                # Close Splash Screen
+                pyi_splash.close()
+            except:
+                pass
+
         gui.exec()
 
 

src/search_type/image_search.py

@@ -211,9 +211,9 @@ def collate_results(hits, image_names, output_directory, image_files_url, count=
         # Add the image metadata to the results
         results += [{
             "entry": f'{image_files_url}/{target_image_name}',
-            "score": f"{hit['score']:.3f}",
+            "score": f"{hit['score']:.9f}",
-            "image_score": f"{hit['image_score']:.3f}",
+            "image_score": f"{hit['image_score']:.9f}",
-            "metadata_score": f"{hit['metadata_score']:.3f}",
+            "metadata_score": f"{hit['metadata_score']:.9f}",
         }]
 
     return results

src/utils/cli.py

@@ -3,7 +3,7 @@ import argparse
 import pathlib
 
 # Internal Packages
-from src.utils.helpers import get_absolute_path, resolve_absolute_path
+from src.utils.helpers import resolve_absolute_path
 from src.utils.yaml import parse_config_from_file
 
 

src/utils/constants.py

@@ -30,7 +30,7 @@ default_config = {
             'input-filter': None,
             'embeddings-file': '~/.khoj/content/image/image_embeddings.pt',
             'batch-size': 50,
-            'use-xmp-metadata': True
+            'use-xmp-metadata': False
         },
         'music': {
             'input-files': None,

src/utils/helpers.py

@@ -1,5 +1,6 @@
 # Standard Packages
 import pathlib
+import sys
 from os.path import join
 
 
@@ -55,3 +56,8 @@ def load_model(model_name, model_dir, model_type):
             model.save(model_path)
 
     return model
+
+
+def is_pyinstaller_app():
+    "Returns true if the app is running from Native GUI created by PyInstaller"
+    return getattr(sys, 'frozen', False) and hasattr(sys, '_MEIPASS')