#!/usr/bin/env python3
"""
Search all Qdrant locations for MCP codebase embeddings.
"""
from pathlib import Path
from qdrant_client import QdrantClient
import os
from mcp_server.core.path_utils import PathUtils
def find_mcp_embeddings():
"""Search for MCP codebase embeddings in all Qdrant locations."""
print("=" * 60)
print("SEARCHING FOR MCP CODEBASE EMBEDDINGS")
print("=" * 60)
# Check all possible Qdrant locations
qdrant_locations = [
"PathUtils.get_workspace_root()/.indexes/qdrant/main.qdrant",
"PathUtils.get_workspace_root()/vector_index.qdrant",
"PathUtils.get_workspace_root()/.mcp-index/vector_index.qdrant",
]
# Add any Qdrant databases in data directory
data_path = Path("PathUtils.get_workspace_root()/data")
if data_path.exists():
for root, dirs, files in os.walk(data_path):
if "qdrant" in root:
qdrant_locations.append(root)
mcp_patterns = [
"mcp_server",
"dispatcher.py",
"semantic_indexer",
"plugin_factory",
"enhanced_dispatcher",
"PathUtils.get_workspace_root()/mcp_server",
"PathUtils.get_workspace_root() / mcp_server"
]
for qdrant_path in qdrant_locations:
if not Path(qdrant_path).exists():
continue
print(f"\nChecking: {qdrant_path}")
# Remove lock file
lock_file = Path(qdrant_path) / ".lock"
if lock_file.exists():
try:
lock_file.unlink()
except:
pass
try:
client = QdrantClient(path=qdrant_path)
collections = client.get_collections()
print(f" Found {len(collections.collections)} collections")
for collection in collections.collections:
try:
info = client.get_collection(collection.name)
if info.points_count == 0:
continue
# Sample points to check content
sample = client.scroll(
collection_name=collection.name,
limit=50,
with_payload=True,
with_vectors=False
)
if not sample[0]:
continue
# Check for MCP patterns
matches = 0
matched_files = []
for point in sample[0]:
payload_str = str(point.payload).lower()
for pattern in mcp_patterns:
if pattern.lower() in payload_str:
matches += 1
# Extract file path from payload
file_path = None
for key in ["file_path", "file", "relative_path", "filepath", "path"]:
if key in point.payload:
file_path = point.payload[key]
break
if file_path and file_path not in matched_files:
matched_files.append(file_path)
break
if matches > 0:
print(f"\n ā
Collection '{collection.name}' ({info.points_count} points)")
print(f" Found {matches} MCP-related entries in sample")
print(f" Sample files:")
for i, file in enumerate(matched_files[:3]):
print(f" {i+1}. {file}")
except Exception as e:
print(f" Error checking {collection.name}: {e}")
except Exception as e:
print(f" Error connecting: {e}")
# Also check if there are any index-specific Qdrant databases
print("\n\nChecking index-specific Qdrant databases...")
indexes_path = Path("PathUtils.get_workspace_root()/.indexes")
if indexes_path.exists():
for index_dir in indexes_path.iterdir():
if index_dir.is_dir() and index_dir.name != "qdrant":
# Check if this index has its own Qdrant
qdrant_path = index_dir / "vector_index.qdrant"
if qdrant_path.exists():
print(f"\nFound index-specific Qdrant: {qdrant_path}")
# Could check this too if needed
print("\n" + "=" * 60)
print("SEARCH COMPLETE")
print("=" * 60)
print("\nš RECOMMENDATIONS:")
print("1. The current MCP codebase embeddings are missing")
print("2. The 'codebase-f7b49f5d0ae0' collection is empty and needs to be populated")
print("3. The 'code-embeddings' collection only has test repository data")
print("\nTo fix semantic search:")
print("- Option 1: Populate the codebase-f7b49f5d0ae0 collection with current codebase")
print("- Option 2: Create a new collection with proper embeddings")
print("- Option 3: Disable semantic search and use SQL-only for now")
if __name__ == "__main__":
find_mcp_embeddings()
