Restructure backend

.github/workflows/build.yml

@@ -31,7 +31,7 @@ jobs:
           .venv/bin/pip install ruff pytest
 
       - name: Lint Python
-        run: .venv/bin/ruff check searchy/*.py
+        run: .venv/bin/ruff check searchy/backend/*.py
 
       - name: Test Python
-        run: .venv/bin/pytest searchy/tests/ -x -q
+        run: .venv/bin/pytest searchy/backend/tests/ -x -q

Makefile

@@ -83,19 +83,19 @@ release: dmg sha ## Full release: archive → export → DMG → SHA
 setup: ## Set up Python venv with dependencies
 	python3 -m venv .venv
 	.venv/bin/pip install --upgrade pip
-	.venv/bin/pip install -r searchy/requirements.txt
+	.venv/bin/pip install -r searchy/backend/requirements.txt
 	@echo "✓ Python env ready. Activate with: source .venv/bin/activate"
 
 # ─── Lint ─────────────────────────────────────────────────
 lint: ## Run linters (ruff for Python)
 	@command -v ruff >/dev/null 2>&1 || (echo "Installing ruff..." && pip install ruff)
-	ruff check searchy/*.py
+	ruff check searchy/backend/*.py
 	@echo "✓ Lint passed"
 
 # ─── Test ─────────────────────────────────────────────────
 test: ## Run Python unit tests
 	@test -f .venv/bin/pytest || (python3 -m venv .venv && .venv/bin/pip install -q pytest)
-	.venv/bin/pytest searchy/tests/ -x -q
+	.venv/bin/pytest searchy/backend/tests/ -x -q
 
 # ─── Check (all pre-push checks) ─────────────────────────
 check: lint test ## Run lint + tests

Scripts/pre-push-checks.sh

@@ -20,19 +20,19 @@ echo "═══ Pre-push checks ═══"
 
 # 1. Syntax check
 echo -n "• py_compile... "
-for f in searchy/*.py; do
+for f in searchy/backend/*.py; do
     $PYTHON -m py_compile "$f" || FAILED=1
 done
 [ $FAILED -eq 0 ] && echo "ok" || echo "FAIL"
 
 # 2. Ruff lint
 echo -n "• ruff... "
-"$VENV/bin/ruff" check searchy/*.py --quiet || FAILED=1
+"$VENV/bin/ruff" check searchy/backend/*.py --quiet || FAILED=1
 [ $FAILED -eq 0 ] && echo "ok" || echo "FAIL"
 
 # 3. Tests
 echo -n "• pytest... "
-$PYTHON -m pytest searchy/tests/ -x -q || FAILED=1
+$PYTHON -m pytest searchy/backend/tests/ -x -q || FAILED=1
 
 # 4. Swift build (skip with SKIP_SWIFT=1)
 if [ "${SKIP_SWIFT:-0}" != "1" ]; then

pyproject.toml

@@ -6,4 +6,4 @@ line-length = 120
 select = ["F", "E9", "W6"]
 
 [tool.pytest.ini_options]
-testpaths = ["searchy/tests"]
+testpaths = ["searchy/backend/tests"]

searchy/ContentView.swift

@@ -1401,11 +1401,7 @@ class DirectoryManager: ObservableObject {
            let directories = try? JSONDecoder().decode([WatchedDirectory].self, from: data) {
             self.watchedDirectories = directories
         } else {
-            // Default directories
-            self.watchedDirectories = [
-                WatchedDirectory(path: NSString(string: "~/Downloads").expandingTildeInPath),
-                WatchedDirectory(path: NSString(string: "~/Desktop").expandingTildeInPath, filter: "Screenshot", filterType: .startsWith)
-            ]
+            self.watchedDirectories = []
         }
     }
 

searchy/backend/constants.py

@@ -0,0 +1,46 @@
+"""Centralized constants for the Searchy project."""
+
+import os
+
+# ─── Paths ────────────────────────────────────────────────
+DEFAULT_DATA_DIR = os.path.join(
+    os.path.expanduser("~/Library/Application Support"), "searchy"
+)
+
+# ─── Image Formats ────────────────────────────────────────
+IMAGE_EXTENSIONS = {'.jpg', '.jpeg', '.png', '.gif', '.bmp', '.tiff', '.webp', '.heic'}
+
+# ─── Directories to Skip During Indexing ──────────────────
+SKIP_DIRS = {
+    'site-packages', 'node_modules', 'vendor', '__pycache__',
+    'env', 'venv', '.venv', 'virtualenv',
+    'Library', 'Caches', 'cache', '.cache',
+    'build', 'dist', 'target', '.git', '.svn',
+    'DerivedData', 'xcuserdata', 'Pods',
+    '__MACOSX', '.Trash', '.Spotlight-V100', '.fseventsd',
+}
+
+# ─── Indexing Defaults ────────────────────────────────────
+DEFAULT_BATCH_SIZE = 64
+DEFAULT_MAX_DIMENSION = 384
+LARGE_BATCH_SIZE = 500  # For bulk operations (duplicates, face clustering)
+
+# ─── Search Defaults ──────────────────────────────────────
+DEFAULT_TOP_K = 20
+DEFAULT_OCR_WEIGHT = 0.3
+DEFAULT_SIMILARITY_THRESHOLD = 0.95  # For duplicate detection
+
+# ─── Model ────────────────────────────────────────────────
+DEFAULT_EMBEDDING_DIM = 512
+DEFAULT_MODEL_NAME = "openai/clip-vit-base-patch32"
+
+# ─── TTL ──────────────────────────────────────────────────
+TTL_CHECK_INTERVAL = 30  # Seconds between TTL checks
+
+# ─── OCR ──────────────────────────────────────────────────
+OCR_TEXT_PREVIEW_LENGTH = 300
+
+# ─── Image Watcher ────────────────────────────────────────
+WATCHER_DEBOUNCE_DELAY = 0.5  # Seconds
+WATCHER_NOTIFY_TIMEOUT = 5    # Seconds
+WATCHER_INDEX_TIMEOUT = 30    # Seconds

searchy/generate_embeddings.py → searchy/backend/generate_embeddings.py

@@ -14,14 +14,15 @@
 import sys
 import json
 import argparse
-import re
 import time
 from PIL import Image
 
 from atomic_write import atomic_pickle_dump
 
 # Import centralized model manager
 from clip_model import model_manager
+from constants import SKIP_DIRS, IMAGE_EXTENSIONS
+from utils import matches_filter, is_user_image
 
 # OCR support using macOS Vision framework
 try:
@@ -96,45 +97,6 @@ def resize_image_for_fast_indexing(image, max_dimension=384):
     return image.resize((new_width, new_height), Image.LANCZOS)
 
 
-def matches_filter(filename, filter_type, filter_value):
-    """Check if filename matches the filter criteria"""
-    if not filter_value or filter_type == "all":
-        return True
-
-    filename_lower = filename.lower()
-    filter_lower = filter_value.lower()
-
-    if filter_type == "starts-with":
-        return filename_lower.startswith(filter_lower)
-    elif filter_type == "ends-with":
-        return filename_lower.endswith(filter_lower)
-    elif filter_type == "contains":
-        return filter_lower in filename_lower
-    elif filter_type == "regex":
-        try:
-            return bool(re.search(filter_value, filename, re.IGNORECASE))
-        except re.error:
-            return False
-    return True
-
-
-# Directories to skip (system, packages, caches, etc.)
-SKIP_DIRS = {
-    'site-packages', 'node_modules', 'vendor', '__pycache__',
-    'env', 'venv', '.venv', 'virtualenv',
-    'Library', 'Caches', 'cache', '.cache',
-    'build', 'dist', 'target', '.git', '.svn',
-    'DerivedData', 'xcuserdata', 'Pods',
-    '__MACOSX', '.Trash', '.Spotlight-V100', '.fseventsd'
-}
-
-
-def is_user_image(path):
-    """Check if path is a user image (not system/package file)."""
-    if os.path.basename(path).startswith('.'):
-        return False
-    parts = path.split(os.sep)
-    return not any(part in SKIP_DIRS for part in parts)
 
 
 def index_images_with_clip(output_dir, incremental=False, new_files=None,
@@ -313,7 +275,7 @@ def process_images(image_dirs, output_dir, fast_indexing=True, max_dimension=384
                     if file.startswith('.'):
                         continue
 
-                    if file.lower().endswith(('.png', '.jpg', '.jpeg', '.gif', '.webp', '.bmp')):
+                    if os.path.splitext(file.lower())[1] in IMAGE_EXTENSIONS:
                         # Apply filter if specified
                         if filter_type and filter_value:
                             if not matches_filter(file, filter_type, filter_value):

searchy/image_watcher.py → searchy/backend/image_watcher.py

@@ -10,43 +10,23 @@
 import sys
 import time
 import json
-import re
 import argparse
 import urllib.request
 import urllib.error
 from pathlib import Path
 from watchdog.observers import Observer
 from watchdog.events import FileSystemEventHandler
 
-# Supported image extensions
-IMAGE_EXTENSIONS = {'.jpg', '.jpeg', '.png', '.gif', '.bmp', '.tiff', '.webp', '.heic'}
+from constants import (
+    IMAGE_EXTENSIONS, DEFAULT_BATCH_SIZE, DEFAULT_MAX_DIMENSION,
+    WATCHER_DEBOUNCE_DELAY, WATCHER_NOTIFY_TIMEOUT, WATCHER_INDEX_TIMEOUT,
+)
+from utils import matches_filter
 
 # Default server URL (can be overridden via --server-url)
 DEFAULT_SERVER_URL = "http://127.0.0.1:7860"
 
 
-def matches_filter(filename, filter_type, filter_value):
-    """Check if filename matches the filter criteria"""
-    if not filter_value or filter_type == "all":
-        return True
-
-    filename_lower = filename.lower()
-    filter_lower = filter_value.lower()
-
-    if filter_type == "starts-with":
-        return filename_lower.startswith(filter_lower)
-    elif filter_type == "ends-with":
-        return filename_lower.endswith(filter_lower)
-    elif filter_type == "contains":
-        return filter_lower in filename_lower
-    elif filter_type == "regex":
-        try:
-            return bool(re.search(filter_value, filename, re.IGNORECASE))
-        except re.error:
-            return False
-    return True
-
-
 def call_server_notify(server_url, file_path):
     """Notify the server that a new image was detected (for immediate display)."""
     url = f"{server_url}/notify-new-image"
@@ -65,7 +45,7 @@ def call_server_notify(server_url, file_path):
             method='POST'
         )
 
-        with urllib.request.urlopen(req, timeout=5) as response:
+        with urllib.request.urlopen(req, timeout=WATCHER_NOTIFY_TIMEOUT) as response:
             result = json.loads(response.read().decode('utf-8'))
             return result
 
@@ -76,7 +56,8 @@ def call_server_notify(server_url, file_path):
 
 
 def call_server_index(server_url, data_dir, files, fast_indexing=True,
-                      max_dimension=384, batch_size=64):
+                      max_dimension=DEFAULT_MAX_DIMENSION,
+                      batch_size=DEFAULT_BATCH_SIZE):
     """Call the server's /index endpoint to index files."""
     url = f"{server_url}/index"
 
@@ -97,7 +78,7 @@ def call_server_index(server_url, data_dir, files, fast_indexing=True,
             method='POST'
         )
 
-        with urllib.request.urlopen(req, timeout=30) as response:
+        with urllib.request.urlopen(req, timeout=WATCHER_INDEX_TIMEOUT) as response:
             result = json.loads(response.read().decode('utf-8'))
             return result
 
@@ -112,7 +93,8 @@ def call_server_index(server_url, data_dir, files, fast_indexing=True,
 class ImageEventHandler(FileSystemEventHandler):
     def __init__(self, data_dir, server_url=DEFAULT_SERVER_URL,
                  filter_type=None, filter_value=None,
-                 fast_indexing=True, max_dimension=384, batch_size=64):
+                 fast_indexing=True, max_dimension=DEFAULT_MAX_DIMENSION,
+                 batch_size=DEFAULT_BATCH_SIZE):
         self.data_dir = data_dir
         self.server_url = server_url
         self.filter_type = filter_type
@@ -122,7 +104,7 @@ def __init__(self, data_dir, server_url=DEFAULT_SERVER_URL,
         self.batch_size = batch_size
         self.pending_files = set()
         self.last_index_time = time.time()
-        self.debounce_delay = 0.5  # Wait 0.5 seconds before indexing (reduced for faster response)
+        self.debounce_delay = WATCHER_DEBOUNCE_DELAY
 
     def on_created(self, event):
         """Called when a file or directory is created"""
@@ -231,7 +213,8 @@ def check_and_index(self):
 
 def watch_directory(watch_path, data_dir, server_url=DEFAULT_SERVER_URL,
                     filter_type=None, filter_value=None,
-                    fast_indexing=True, max_dimension=384, batch_size=64):
+                    fast_indexing=True, max_dimension=DEFAULT_MAX_DIMENSION,
+                    batch_size=DEFAULT_BATCH_SIZE):
     """Watch directory for new images and auto-index them via server."""
 
     print(f"Watching for new images in: {watch_path}", file=sys.stderr)
@@ -280,9 +263,9 @@ def watch_directory(watch_path, data_dir, server_url=DEFAULT_SERVER_URL,
                         help="Enable fast indexing (resize images)")
     parser.add_argument("--no-fast", action="store_false", dest="fast",
                         help="Disable fast indexing")
-    parser.add_argument("--max-dimension", type=int, default=384,
+    parser.add_argument("--max-dimension", type=int, default=DEFAULT_MAX_DIMENSION,
                         help="Maximum image dimension for fast indexing")
-    parser.add_argument("--batch-size", type=int, default=64,
+    parser.add_argument("--batch-size", type=int, default=DEFAULT_BATCH_SIZE,
                         help="Batch size for processing")
 
     args = parser.parse_args()