specmem-hardwicksoftware 3.7.9 → 3.7.11

package/embedding-sandbox/frankenstein-embeddings.py

@@ -3026,17 +3026,30 @@ class EmbeddingServer:
 
         try:
             # Get TOTAL count first for progress tracking
+            # Schema isolation already separates projects, so project_path filter
+            # is only needed when file_path values are absolute. Check if filter
+            # matches anything — if not, skip it (paths are likely relative).
             count_cursor = conn.cursor()
+            use_project_filter = False
             if project_path:
                 count_cursor.execute(
                     "SELECT COUNT(*) FROM codebase_files WHERE embedding IS NULL AND content IS NOT NULL AND file_path LIKE %s",
                     (f"{project_path}%",)
                 )
-                print(f"🎯 Filtering to project: {project_path}", file=sys.stderr)
+                filtered_count = count_cursor.fetchone()[0]
+                if filtered_count > 0:
+                    use_project_filter = True
+                    total_missing = filtered_count
+                    print(f"🎯 Filtering to project: {project_path} ({total_missing} files)", file=sys.stderr)
+                else:
+                    # Paths are relative — schema isolation handles project separation
+                    count_cursor.execute("SELECT COUNT(*) FROM codebase_files WHERE embedding IS NULL AND content IS NOT NULL")
+                    total_missing = count_cursor.fetchone()[0]
+                    print(f"📁 Schema-isolated mode (relative paths, {total_missing} files)", file=sys.stderr)
             else:
                 count_cursor.execute("SELECT COUNT(*) FROM codebase_files WHERE embedding IS NULL AND content IS NOT NULL")
+                total_missing = count_cursor.fetchone()[0]
                 print(f"⚠️ Processing ALL projects (no project_path filter)", file=sys.stderr)
-            total_missing = count_cursor.fetchone()[0]
             count_cursor.close()
 
             print(f"📊 Total files needing embeddings: {total_missing}", file=sys.stderr)
@@ -3054,7 +3067,7 @@ class EmbeddingServer:
 
                 # Fetch next batch - always gets files without embeddings
                 cursor = conn.cursor()
-                if project_path:
+                if use_project_filter:
                     cursor.execute("""
                         SELECT id, file_path, content
                         FROM codebase_files
@@ -3242,17 +3255,28 @@ class EmbeddingServer:
 
         try:
             # Get TOTAL count first for progress tracking
+            # Schema isolation handles project separation, so project_path filter
+            # only applies when file_path values are absolute paths.
             count_cursor = conn.cursor()
+            use_project_filter = False
             if project_path:
                 count_cursor.execute(
                     "SELECT COUNT(*) FROM code_definitions WHERE embedding IS NULL AND file_path LIKE %s",
                     (f"{project_path}%",)
                 )
-                print(f"🎯 Filtering to project: {project_path}", file=sys.stderr)
+                filtered_count = count_cursor.fetchone()[0]
+                if filtered_count > 0:
+                    use_project_filter = True
+                    total_missing = filtered_count
+                    print(f"🎯 Filtering to project: {project_path} ({total_missing} definitions)", file=sys.stderr)
+                else:
+                    count_cursor.execute("SELECT COUNT(*) FROM code_definitions WHERE embedding IS NULL")
+                    total_missing = count_cursor.fetchone()[0]
+                    print(f"📁 Schema-isolated mode (relative paths, {total_missing} definitions)", file=sys.stderr)
             else:
                 count_cursor.execute("SELECT COUNT(*) FROM code_definitions WHERE embedding IS NULL")
+                total_missing = count_cursor.fetchone()[0]
                 print(f"⚠️ Processing ALL projects (no project_path filter)", file=sys.stderr)
-            total_missing = count_cursor.fetchone()[0]
             count_cursor.close()
 
             print(f"📊 Total code_definitions needing embeddings: {total_missing}", file=sys.stderr)
@@ -3270,7 +3294,7 @@ class EmbeddingServer:
 
                 # Fetch next batch
                 cursor = conn.cursor()
-                if project_path:
+                if use_project_filter:
                     cursor.execute("""
                         SELECT id, definition_type, name, signature, docstring, language, file_path
                         FROM code_definitions

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "specmem-hardwicksoftware",
-  "version": "3.7.9",
+  "version": "3.7.11",
   "type": "module",
   "description": "Persistent memory system for coding sessions - semantic search with pgvector, token compression, team coordination, file watching. Needs root: installs system-wide hooks, manages docker/PostgreSQL, writes global configs, handles screen sessions. justcalljon.pro",
   "main": "dist/index.js",