[{"_1":2,"_410":-5,"_411":-5},"loaderData",{"_3":4,"_22":23,"_24":25,"_289":290},"root",{"_5":6,"_7":8},"toasts",[],"visitor",{"_9":10,"_11":12},"viewportWidth",1280,"visitorSession",{"_13":14,"_16":17,"_18":-5,"_19":20,"_21":-5},"attributes",[15],"bot","id",11143072,"membership","referenceId","01989bb3-6f03-717c-8403-57f67b03b520","userAccount","routes/_public/_layout",{},"routes/_public/mcp/servers/~namespace/~slug/_pages/_layout",{"_26":27,"_82":-5,"_83":40,"_29":84,"_207":208,"_288":174},"availableLocales",[28,40,47,54,61,68,75],{"_29":30,"_33":37,"_38":39},"mcpServer",{"_31":32,"_33":34,"_35":36},"description","Bietet semantischen Speicher und dauerhaften Speicher für Claude und nutzt ChromaDB und Satztransformatoren für erweiterte Such- und Abruffunktionen.","name","MCP-Speicherdienst","readme","# MCP-Speicherdienst\n\n[![Lizenz: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![Schmiedeabzeichen](https://smithery.ai/badge/@doobidoo/mcp-memory-service)](https://smithery.ai/server/@doobidoo/mcp-memory-service) [![Verifiziert auf MseeP](https://mseep.ai/badge.svg)](https://mseep.ai/app/0513fb92-e941-4fe0-9948-2a1dbb870dcf)\n\nEin MCP-Server, der semantischen Speicher und persistente Speicherfunktionen für Claude Desktop mithilfe von ChromaDB und Satztransformatoren bereitstellt. Dieser Dienst ermöglicht Langzeitspeicherung mit semantischen Suchfunktionen und eignet sich daher ideal für die Kontextpflege über Konversationen und Instanzen hinweg.\n\n## Helfen\n\nSprechen Sie mit [TalkToGitHub](https://talktogithub.com/doobidoo/mcp-memory-service) mit dem Repo!\n\n## Merkmale\n\n* Semantische Suche mit Satztransformatoren\n* **Zeitbasiertes Erinnern in natürlicher Sprache** (z. B. „letzte Woche“, „gestern Morgen“)\n* Tag-basiertes Speicherabrufsystem\n* Persistenter Speicher mit ChromaDB\n* Automatische Datenbanksicherungen\n* Tools zur Speicheroptimierung\n* Genaue Übereinstimmungssuche\n* Debug-Modus für Ähnlichkeitsanalyse\n* Überwachung der Datenbankintegrität\n* Duplikaterkennung und -bereinigung\n* Anpassbares Einbettungsmodell\n* **Plattformübergreifende Kompatibilität** (Apple Silicon, Intel, Windows, Linux)\n* **Hardwarebewusste Optimierungen** für verschiedene Umgebungen\n* **Anmutige Fallbacks** für begrenzte Hardwareressourcen\n\n## Installation\n\n### Schnellstart (empfohlen)\n\nDas erweiterte Installationsskript erkennt Ihr System automatisch und installiert die entsprechenden Abhängigkeiten:\n\n```bash\n# Clone the repository\ngit clone https://github.com/doobidoo/mcp-memory-service.git\ncd mcp-memory-service\n\n# Create and activate a virtual environment\npython -m venv venv\nsource venv/bin/activate # On Windows: venv\\Scripts\\activate\n\n# Run the installation script\npython install.py\n```\n\nDas Skript `install.py` führt Folgendes aus:\n\n1. Ermitteln Sie Ihre Systemarchitektur und verfügbare Hardwarebeschleuniger\n2. Installieren Sie die entsprechenden Abhängigkeiten für Ihre Plattform\n3. Konfigurieren Sie die optimalen Einstellungen für Ihre Umgebung\n4. Überprüfen Sie die Installation und stellen Sie bei Bedarf eine Diagnose bereit\n\n### Docker-Installation\n\nSie können den Memory Service mit Docker ausführen:\n\n```bash\n# Using Docker Compose (recommended)\ndocker-compose up\n\n# Using Docker directly\ndocker build -t mcp-memory-service .\ndocker run -p 8000:8000 -v /path/to/data:/app/chroma_db -v /path/to/backups:/app/backups mcp-memory-service\n```\n\nWir bieten mehrere Docker Compose-Konfigurationen für verschiedene Szenarien:\n\n* `docker-compose.yml` – Standardkonfiguration mit Pip-Installation\n* `docker-compose.uv.yml` – Alternative Konfiguration mit UV-Paketmanager\n* `docker-compose.pythonpath.yml` – Konfiguration mit expliziten PYTHONPATH-Einstellungen\n\nSo verwenden Sie eine alternative Konfiguration:\n\n```bash\ndocker-compose -f docker-compose.uv.yml up\n```\n\n### Windows-Installation (Sonderfall)\n\nWindows-Benutzer können aufgrund der plattformspezifischen Verfügbarkeit von Wheels auf Probleme bei der PyTorch-Installation stoßen. Verwenden Sie unser Windows-spezifisches Installationsskript:\n\n```bash\n# After activating your virtual environment\npython scripts/install_windows.py\n```\n\nDieses Skript behandelt:\n\n1. Erkennen der CUDA-Verfügbarkeit und -Version\n2. Installieren der entsprechenden PyTorch-Version von der richtigen Index-URL\n3. Installieren anderer Abhängigkeiten ohne Konflikte mit PyTorch\n4. Überprüfen der Installation\n\n### Installation über Smithery\n\nSo installieren Sie Memory Service für Claude Desktop automatisch über [Smithery](https://smithery.ai/server/@doobidoo/mcp-memory-service) :\n\n```bash\nnpx -y @smithery/cli install @doobidoo/mcp-memory-service --client claude\n```\n\n### Detaillierte Installationsanleitung\n\nAusführliche Installationsanweisungen und Hinweise zur Fehlerbehebung finden Sie im [Installationshandbuch](docs/guides/installation.md) .\n\n## Claude MCP-Konfiguration\n\n### Standardkonfiguration\n\nFügen Sie Ihrer Datei `claude_desktop_config.json` Folgendes hinzu:\n\n```json\n{\n \"memory\": {\n \"command\": \"uv\",\n \"args\": [\n \"--directory\",\n \"your_mcp_memory_service_directory\", // e.g., \"C:\\\\REPOSITORIES\\\\mcp-memory-service\"\n \"run\",\n \"memory\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"your_chroma_db_path\", // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\"\n \"MCP_MEMORY_BACKUPS_PATH\": \"your_backups_path\" // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n### Windows-spezifische Konfiguration (empfohlen)\n\nFür Windows-Benutzer empfehlen wir die Verwendung des Wrapper-Skripts, um sicherzustellen, dass PyTorch ordnungsgemäß installiert ist:\n\n```json\n{\n \"memory\": {\n \"command\": \"python\",\n \"args\": [\n \"C:\\\\path\\\\to\\\\mcp-memory-service\\\\memory_wrapper.py\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\",\n \"MCP_MEMORY_BACKUPS_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\nDas Wrapper-Skript wird:\n\n1. Überprüfen Sie, ob PyTorch installiert und richtig konfiguriert ist\n2. Installieren Sie PyTorch bei Bedarf mit der richtigen Index-URL\n3. Führen Sie den Speicherserver mit der entsprechenden Konfiguration aus\n\n## Benutzerhandbuch\n\nAusführliche Anweisungen zur Interaktion mit dem Speicherdienst in Claude Desktop:\n\n* [Aufrufhandbuch](docs/guides/invocation_guide.md) - Lernen Sie die spezifischen Schlüsselwörter und Ausdrücke, die Speicheroperationen in Claude auslösen\n* [Installationshandbuch](docs/guides/installation.md) - Detaillierte Einrichtungsanweisungen\n\nDer Speicherdienst wird in Ihren Gesprächen mit Claude über natürliche Sprachbefehle aufgerufen. Beispiel:\n\n* Zum Speichern: „Bitte denken Sie daran, dass der Abgabetermin für mein Projekt der 15. Mai ist.“\n* Zum Abrufen: „Erinnern Sie sich, was ich Ihnen über die Deadline meines Projekts gesagt habe?“\n* Zum Löschen: „Bitte vergessen Sie, was ich Ihnen über meine Adresse gesagt habe.“\n\nEine vollständige Liste der Befehle und ausführliche Anwendungsbeispiele finden Sie im [Aufrufhandbuch](docs/guides/invocation_guide.md) .\n\n## Speicheroperationen\n\nDer Speicherdienst stellt über den MCP-Server die folgenden Vorgänge bereit:\n\n### Kernspeichervorgänge\n\n1. `store_memory` - Neue Informationen mit optionalen Tags speichern\n2. `retrieve_memory` - Semantische Suche nach relevanten Erinnerungen durchführen\n3. `recall_memory` - Erinnerungen mithilfe natürlicher Sprachzeitausdrücke abrufen\n4. `search_by_tag` – Finden Sie Erinnerungen mithilfe bestimmter Tags\n5. `exact_match_retrieve` - Finde Erinnerungen mit exakter Inhaltsübereinstimmung\n6. `debug_retrieve` - Erinnerungen mit Ähnlichkeitsbewertungen abrufen\n\n### Datenbankverwaltung\n\n7. `create_backup` - Datenbanksicherung erstellen\n8. `get_stats` - Speicherstatistiken abrufen\n9. `optimize_db` - Datenbankleistung optimieren\n10. `check_database_health` - Datenbank-Integritätsmetriken abrufen\n11. `check_embedding_model` - Modellstatus überprüfen\n\n### Speicherverwaltung\n\n12. `delete_memory` - Löscht bestimmten Speicher nach Hash\n13. `delete_by_tag` - Löscht alle Erinnerungen mit einem bestimmten Tag\n14. `cleanup_duplicates` - Doppelte Einträge entfernen\n\n## Konfigurationsoptionen\n\nKonfigurieren Sie über Umgebungsvariablen:\n\n```\nCHROMA_DB_PATH: Path to ChromaDB storage\nBACKUP_PATH: Path for backups\nAUTO_BACKUP_INTERVAL: Backup interval in hours (default: 24)\nMAX_MEMORIES_BEFORE_OPTIMIZE: Threshold for auto-optimization (default: 10000)\nSIMILARITY_THRESHOLD: Default similarity threshold (default: 0.7)\nMAX_RESULTS_PER_QUERY: Maximum results per query (default: 10)\nBACKUP_RETENTION_DAYS: Number of days to keep backups (default: 7)\nLOG_LEVEL: Logging level (default: INFO)\n\n# Hardware-specific environment variables\nPYTORCH_ENABLE_MPS_FALLBACK: Enable MPS fallback for Apple Silicon (default: 1)\nMCP_MEMORY_USE_ONNX: Use ONNX Runtime for CPU-only deployments (default: 0)\nMCP_MEMORY_USE_DIRECTML: Use DirectML for Windows acceleration (default: 0)\nMCP_MEMORY_MODEL_NAME: Override the default embedding model\nMCP_MEMORY_BATCH_SIZE: Override the default batch size\n```\n\n## Hardwarekompatibilität\n\n| Plattform | Architektur | Beschleuniger | Status |\n| --------- | ----------------------------- | ------------- | --------------------------- |\n| macOS | Apple Silicon (M1/M2/M3) | MPS | ✅ Vollständig unterstützt |\n| macOS | Apple Silicon unter Rosetta 2 | CPU | ✅ Unterstützt mit Fallbacks |\n| macOS | Intel | CPU | ✅ Vollständig unterstützt |\n| Windows | x86\\_64 | CUDA | ✅ Vollständig unterstützt |\n| Windows | x86\\_64 | DirectML | ✅ Unterstützt |\n| Windows | x86\\_64 | CPU | ✅ Unterstützt mit Fallbacks |\n| Linux | x86\\_64 | CUDA | ✅ Vollständig unterstützt |\n| Linux | x86\\_64 | ROCm | ✅ Unterstützt |\n| Linux | x86\\_64 | CPU | ✅ Unterstützt mit Fallbacks |\n| Linux | ARM64 | CPU | ✅ Unterstützt mit Fallbacks |\n\n## Testen\n\n```bash\n# Install test dependencies\npip install pytest pytest-asyncio\n\n# Run all tests\npytest tests/\n\n# Run specific test categories\npytest tests/test_memory_ops.py\npytest tests/test_semantic_search.py\npytest tests/test_database.py\n\n# Verify environment compatibility\npython scripts/verify_environment_enhanced.py\n\n# Verify PyTorch installation on Windows\npython scripts/verify_pytorch_windows.py\n\n# Perform comprehensive installation verification\npython scripts/test_installation.py\n```\n\n## Fehlerbehebung\n\nAusführliche Schritte zur Fehlerbehebung finden Sie in der [Installationsanleitung](docs/guides/installation.md#troubleshooting-common-installation-issues) .\n\n### Tipps zur schnellen Fehlerbehebung\n\n* **Windows PyTorch-Fehler** : Verwenden Sie `python scripts/install_windows.py`\n* **macOS Intel-Abhängigkeitskonflikte** : Verwenden Sie `python install.py --force-compatible-deps`\n* **Rekursionsfehler** : Führen Sie `python scripts/fix_sitecustomize.py` aus\n* **Umgebungsüberprüfung** : Führen Sie `python scripts/verify_environment_enhanced.py` aus\n* **Speicherprobleme** : Setzen Sie `MCP_MEMORY_BATCH_SIZE=4` und versuchen Sie es mit einem kleineren Modell\n* **Apple Silicon** : Stellen Sie sicher, dass Python 3.10+ für ARM64 erstellt wurde, und setzen Sie `PYTORCH_ENABLE_MPS_FALLBACK=1`\n* **Installationstest** : Führen Sie `python scripts/test_installation.py` aus\n\n## Projektstruktur\n\n```\nmcp-memory-service/\n├── src/mcp_memory_service/ # Core package code\n│ ├── __init__.py\n│ ├── config.py # Configuration utilities\n│ ├── models/ # Data models\n│ ├── storage/ # Storage implementations\n│ ├── utils/ # Utility functions\n│ └── server.py # Main MCP server\n├── scripts/ # Helper scripts\n├── memory_wrapper.py # Windows wrapper script\n├── install.py # Enhanced installation script\n└── tests/ # Test suite\n```\n\n## Entwicklungsrichtlinien\n\n* Python 3.10+ mit Typhinweisen\n* Verwenden Sie Datenklassen für Modelle\n* Dreifach zitierte Docstrings für Module und Funktionen\n* Async/Await-Muster für alle E/A-Vorgänge\n* Befolgen Sie die PEP 8-Stilrichtlinien\n* Schließen Sie Tests für neue Funktionen ein\n\n## Lizenz\n\nMIT-Lizenz – Einzelheiten finden Sie in der Datei „LICENSE“\n\n## Danksagung\n\n* ChromaDB-Team für die Vektordatenbank\n* Sentence Transformers-Projekt zum Einbetten von Modellen\n* MCP-Projekt zur Protokollspezifikation\n\n## Kontakt\n\n[Telegramm](t.me/doobeedoo)\n\n## Integrationen\n\nDer MCP Memory Service kann mit verschiedenen Tools und Dienstprogrammen erweitert werden. Eine Liste der verfügbaren Optionen finden Sie unter [Integrationen](docs/integrations.md) , darunter:\n\n* [MCP Memory Dashboard](https://github.com/doobidoo/mcp-memory-dashboard) – Web-Benutzeroberfläche zum Durchsuchen und Verwalten von Erinnerungen\n* [Claude-Speicherkontext](https://github.com/doobidoo/claude-memory-context) – Speicherkontext in Claude-Projektanweisungen einfügen","German","tag","de-DE",{"_29":41,"_33":45,"_38":46},{"_31":42,"_33":43,"_35":44},"Provides semantic memory and persistent storage for Claude, leveraging ChromaDB and sentence transformers for enhanced search and retrieval capabilities.","MCP Memory Service","# MCP Memory Service\n\n[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)\n[![smithery badge](https://smithery.ai/badge/@doobidoo/mcp-memory-service)](https://smithery.ai/server/@doobidoo/mcp-memory-service)\n[![Verified on MseeP](https://mseep.ai/badge.svg)](https://mseep.ai/app/0513fb92-e941-4fe0-9948-2a1dbb870dcf)\n\nAn MCP server providing semantic memory and persistent storage capabilities for Claude Desktop using ChromaDB and sentence transformers. This service enables long-term memory storage with semantic search capabilities, making it ideal for maintaining context across conversations and instances.\n\n $\"grafik\"$ \n $\"Memory$ \n\n## Help\nTalk to the Repo with [TalkToGitHub](https://talktogithub.com/doobidoo/mcp-memory-service)!\n## Features\n\n- Semantic search using sentence transformers\n- **Natural language time-based recall** (e.g., \"last week\", \"yesterday morning\")\n- Tag-based memory retrieval system\n- Persistent storage using ChromaDB\n- Automatic database backups\n- Memory optimization tools\n- Exact match retrieval\n- Debug mode for similarity analysis\n- Database health monitoring\n- Duplicate detection and cleanup\n- Customizable embedding model\n- **Cross-platform compatibility** (Apple Silicon, Intel, Windows, Linux)\n- **Hardware-aware optimizations** for different environments\n- **Graceful fallbacks** for limited hardware resources\n\n## Installation\n\n### Quick Start (Recommended)\n\nThe enhanced installation script automatically detects your system and installs the appropriate dependencies:\n\n```bash\n# Clone the repository\ngit clone https://github.com/doobidoo/mcp-memory-service.git\ncd mcp-memory-service\n\n# Create and activate a virtual environment\npython -m venv venv\nsource venv/bin/activate # On Windows: venv\\Scripts\\activate\n\n# Run the installation script\npython install.py\n```\n\nThe `install.py` script will:\n1. Detect your system architecture and available hardware accelerators\n2. Install the appropriate dependencies for your platform\n3. Configure the optimal settings for your environment\n4. Verify the installation and provide diagnostics if needed\n\n### Docker Installation\n\nYou can run the Memory Service using Docker:\n\n```bash\n# Using Docker Compose (recommended)\ndocker-compose up\n\n# Using Docker directly\ndocker build -t mcp-memory-service .\ndocker run -p 8000:8000 -v /path/to/data:/app/chroma_db -v /path/to/backups:/app/backups mcp-memory-service\n```\n\nWe provide multiple Docker Compose configurations for different scenarios:\n- `docker-compose.yml` - Standard configuration using pip install\n- `docker-compose.uv.yml` - Alternative configuration using UV package manager\n- `docker-compose.pythonpath.yml` - Configuration with explicit PYTHONPATH settings\n\nTo use an alternative configuration:\n```bash\ndocker-compose -f docker-compose.uv.yml up\n```\n\n### Windows Installation (Special Case)\n\nWindows users may encounter PyTorch installation issues due to platform-specific wheel availability. Use our Windows-specific installation script:\n\n```bash\n# After activating your virtual environment\npython scripts/install_windows.py\n```\n\nThis script handles:\n1. Detecting CUDA availability and version\n2. Installing the appropriate PyTorch version from the correct index URL\n3. Installing other dependencies without conflicting with PyTorch\n4. Verifying the installation\n\n### Installing via Smithery\n\nTo install Memory Service for Claude Desktop automatically via [Smithery](https://smithery.ai/server/@doobidoo/mcp-memory-service):\n\n```bash\nnpx -y @smithery/cli install @doobidoo/mcp-memory-service --client claude\n```\n\n### Detailed Installation Guide\n\nFor comprehensive installation instructions and troubleshooting, see the [Installation Guide](docs/guides/installation.md).\n\n## Claude MCP Configuration\n\n### Standard Configuration\n\nAdd the following to your `claude_desktop_config.json` file:\n\n```json\n{\n \"memory\": {\n \"command\": \"uv\",\n \"args\": [\n \"--directory\",\n \"your_mcp_memory_service_directory\", // e.g., \"C:\\\\REPOSITORIES\\\\mcp-memory-service\"\n \"run\",\n \"memory\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"your_chroma_db_path\", // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\"\n \"MCP_MEMORY_BACKUPS_PATH\": \"your_backups_path\" // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n### Windows-Specific Configuration (Recommended)\n\nFor Windows users, we recommend using the wrapper script to ensure PyTorch is properly installed:\n\n```json\n{\n \"memory\": {\n \"command\": \"python\",\n \"args\": [\n \"C:\\\\path\\\\to\\\\mcp-memory-service\\\\memory_wrapper.py\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\",\n \"MCP_MEMORY_BACKUPS_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\nThe wrapper script will:\n1. Check if PyTorch is installed and properly configured\n2. Install PyTorch with the correct index URL if needed\n3. Run the memory server with the appropriate configuration\n\n## Usage Guide\n\nFor detailed instructions on how to interact with the memory service in Claude Desktop:\n\n- [Invocation Guide](docs/guides/invocation_guide.md) - Learn the specific keywords and phrases that trigger memory operations in Claude\n- [Installation Guide](docs/guides/installation.md) - Detailed setup instructions\n\nThe memory service is invoked through natural language commands in your conversations with Claude. For example:\n- To store: \"Please remember that my project deadline is May 15th.\"\n- To retrieve: \"Do you remember what I told you about my project deadline?\"\n- To delete: \"Please forget what I told you about my address.\"\n\nSee the [Invocation Guide](docs/guides/invocation_guide.md) for a complete list of commands and detailed usage examples.\n\n## Memory Operations\n\nThe memory service provides the following operations through the MCP server:\n\n### Core Memory Operations\n\n1. `store_memory` - Store new information with optional tags\n2. `retrieve_memory` - Perform semantic search for relevant memories\n3. `recall_memory` - Retrieve memories using natural language time expressions \n4. `search_by_tag` - Find memories using specific tags\n5. `exact_match_retrieve` - Find memories with exact content match\n6. `debug_retrieve` - Retrieve memories with similarity scores\n\n### Database Management\n\n7. `create_backup` - Create database backup\n8. `get_stats` - Get memory statistics\n9. `optimize_db` - Optimize database performance\n10. `check_database_health` - Get database health metrics\n11. `check_embedding_model` - Verify model status\n\n### Memory Management\n\n12. `delete_memory` - Delete specific memory by hash\n13. `delete_by_tag` - Delete all memories with specific tag\n14. `cleanup_duplicates` - Remove duplicate entries\n\n## Configuration Options\n\nConfigure through environment variables:\n\n```\nCHROMA_DB_PATH: Path to ChromaDB storage\nBACKUP_PATH: Path for backups\nAUTO_BACKUP_INTERVAL: Backup interval in hours (default: 24)\nMAX_MEMORIES_BEFORE_OPTIMIZE: Threshold for auto-optimization (default: 10000)\nSIMILARITY_THRESHOLD: Default similarity threshold (default: 0.7)\nMAX_RESULTS_PER_QUERY: Maximum results per query (default: 10)\nBACKUP_RETENTION_DAYS: Number of days to keep backups (default: 7)\nLOG_LEVEL: Logging level (default: INFO)\n\n# Hardware-specific environment variables\nPYTORCH_ENABLE_MPS_FALLBACK: Enable MPS fallback for Apple Silicon (default: 1)\nMCP_MEMORY_USE_ONNX: Use ONNX Runtime for CPU-only deployments (default: 0)\nMCP_MEMORY_USE_DIRECTML: Use DirectML for Windows acceleration (default: 0)\nMCP_MEMORY_MODEL_NAME: Override the default embedding model\nMCP_MEMORY_BATCH_SIZE: Override the default batch size\n```\n\n## Hardware Compatibility\n\n| Platform | Architecture | Accelerator | Status |\n|----------|--------------|-------------|--------|\n| macOS | Apple Silicon (M1/M2/M3) | MPS | ✅ Fully supported |\n| macOS | Apple Silicon under Rosetta 2 | CPU | ✅ Supported with fallbacks |\n| macOS | Intel | CPU | ✅ Fully supported |\n| Windows | x86_64 | CUDA | ✅ Fully supported |\n| Windows | x86_64 | DirectML | ✅ Supported |\n| Windows | x86_64 | CPU | ✅ Supported with fallbacks |\n| Linux | x86_64 | CUDA | ✅ Fully supported |\n| Linux | x86_64 | ROCm | ✅ Supported |\n| Linux | x86_64 | CPU | ✅ Supported with fallbacks |\n| Linux | ARM64 | CPU | ✅ Supported with fallbacks |\n\n## Testing\n\n```bash\n# Install test dependencies\npip install pytest pytest-asyncio\n\n# Run all tests\npytest tests/\n\n# Run specific test categories\npytest tests/test_memory_ops.py\npytest tests/test_semantic_search.py\npytest tests/test_database.py\n\n# Verify environment compatibility\npython scripts/verify_environment_enhanced.py\n\n# Verify PyTorch installation on Windows\npython scripts/verify_pytorch_windows.py\n\n# Perform comprehensive installation verification\npython scripts/test_installation.py\n```\n\n## Troubleshooting\n\nSee the [Installation Guide](docs/guides/installation.md#troubleshooting-common-installation-issues) for detailed troubleshooting steps.\n\n### Quick Troubleshooting Tips\n\n- **Windows PyTorch errors**: Use `python scripts/install_windows.py`\n- **macOS Intel dependency conflicts**: Use `python install.py --force-compatible-deps`\n- **Recursion errors**: Run `python scripts/fix_sitecustomize.py` \n- **Environment verification**: Run `python scripts/verify_environment_enhanced.py`\n- **Memory issues**: Set `MCP_MEMORY_BATCH_SIZE=4` and try a smaller model\n- **Apple Silicon**: Ensure Python 3.10+ built for ARM64, set `PYTORCH_ENABLE_MPS_FALLBACK=1`\n- **Installation testing**: Run `python scripts/test_installation.py`\n\n## Project Structure\n\n```\nmcp-memory-service/\n├── src/mcp_memory_service/ # Core package code\n│ ├── __init__.py\n│ ├── config.py # Configuration utilities\n│ ├── models/ # Data models\n│ ├── storage/ # Storage implementations\n│ ├── utils/ # Utility functions\n│ └── server.py # Main MCP server\n├── scripts/ # Helper scripts\n├── memory_wrapper.py # Windows wrapper script\n├── install.py # Enhanced installation script\n└── tests/ # Test suite\n```\n\n## Development Guidelines\n\n- Python 3.10+ with type hints\n- Use dataclasses for models\n- Triple-quoted docstrings for modules and functions\n- Async/await pattern for all I/O operations\n- Follow PEP 8 style guidelines\n- Include tests for new features\n\n## License\n\nMIT License - See LICENSE file for details\n\n## Acknowledgments\n\n- ChromaDB team for the vector database\n- Sentence Transformers project for embedding models\n- MCP project for the protocol specification\n\n## Contact\n\n[Telegram](t.me/doobeedoo)\n\n## Integrations\n\nThe MCP Memory Service can be extended with various tools and utilities. See [Integrations](docs/integrations.md) for a list of available options, including:\n\n- [MCP Memory Dashboard](https://github.com/doobidoo/mcp-memory-dashboard) - Web UI for browsing and managing memories\n- [Claude Memory Context](https://github.com/doobidoo/claude-memory-context) - Inject memory context into Claude project instructions\n","English","en-US",{"_29":48,"_33":52,"_38":53},{"_31":49,"_33":50,"_35":51},"Proporciona memoria semántica y almacenamiento persistente para Claude, aprovechando ChromaDB y transformadores de oraciones para mejorar las capacidades de búsqueda y recuperación.","Servicio de memoria MCP","# Servicio de memoria MCP\n\n[![Licencia: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![insignia de herrería](https://smithery.ai/badge/@doobidoo/mcp-memory-service)](https://smithery.ai/server/@doobidoo/mcp-memory-service)\n\nUn servidor MCP que proporciona memoria semántica y capacidades de almacenamiento persistente para Claude Desktop mediante ChromaDB y transformadores de oraciones. Este servicio permite el almacenamiento en memoria a largo plazo con funciones de búsqueda semántica, lo que lo hace ideal para mantener el contexto entre conversaciones e instancias.\n\n## Ayuda\n\n¡Hable con el repositorio con [TalkToGitHub](https://talktogithub.com/doobidoo/mcp-memory-service) !\n\n## Características\n\n* Búsqueda semántica mediante transformadores de oraciones\n* **Recuerdo basado en el tiempo del lenguaje natural** (por ejemplo, \"la semana pasada\", \"ayer por la mañana\")\n* Sistema de recuperación de memoria basado en etiquetas\n* Almacenamiento persistente mediante ChromaDB\n* Copias de seguridad automáticas de bases de datos\n* Herramientas de optimización de memoria\n* Recuperación de coincidencias exactas\n* Modo de depuración para análisis de similitud\n* Monitoreo del estado de la base de datos\n* Detección y limpieza de duplicados\n* Modelo de incrustación personalizable\n* **Compatibilidad entre plataformas** (Apple Silicon, Intel, Windows, Linux)\n* **Optimizaciones basadas en hardware** para diferentes entornos\n* **Respaldos elegantes** para recursos de hardware limitados\n\n## Instalación\n\n### Inicio rápido (recomendado)\n\nEl script de instalación mejorado detecta automáticamente su sistema e instala las dependencias adecuadas:\n\n```bash\n# Clone the repository\ngit clone https://github.com/doobidoo/mcp-memory-service.git\ncd mcp-memory-service\n\n# Create and activate a virtual environment\npython -m venv venv\nsource venv/bin/activate # On Windows: venv\\Scripts\\activate\n\n# Run the installation script\npython install.py\n```\n\nEl script `install.py` hará lo siguiente:\n\n1. Detecte la arquitectura de su sistema y los aceleradores de hardware disponibles\n2. Instale las dependencias adecuadas para su plataforma\n3. Configure los ajustes óptimos para su entorno\n4. Verificar la instalación y proporcionar diagnósticos si es necesario\n\n### Instalación de Docker\n\nPuede ejecutar el servicio de memoria usando Docker:\n\n```bash\n# Using Docker Compose (recommended)\ndocker-compose up\n\n# Using Docker directly\ndocker build -t mcp-memory-service .\ndocker run -p 8000:8000 -v /path/to/data:/app/chroma_db -v /path/to/backups:/app/backups mcp-memory-service\n```\n\nProporcionamos múltiples configuraciones de Docker Compose para diferentes escenarios:\n\n* `docker-compose.yml` - Configuración estándar mediante pip install\n* `docker-compose.uv.yml` : configuración alternativa mediante el administrador de paquetes UV\n* `docker-compose.pythonpath.yml` : configuración con ajustes PYTHONPATH explícitos\n\nPara utilizar una configuración alternativa:\n\n```bash\ndocker-compose -f docker-compose.uv.yml up\n```\n\n### Instalación de Windows (caso especial)\n\nLos usuarios de Windows podrían experimentar problemas de instalación de PyTorch debido a la disponibilidad de ruedas específicas de cada plataforma. Utilice nuestro script de instalación específico para Windows:\n\n```bash\n# After activating your virtual environment\npython scripts/install_windows.py\n```\n\nEste script maneja:\n\n1. Detección de la disponibilidad y versión de CUDA\n2. Instalar la versión adecuada de PyTorch desde la URL de índice correcta\n3. Instalar otras dependencias sin entrar en conflicto con PyTorch\n4. Verificando la instalación\n\n### Instalación mediante herrería\n\nPara instalar el Servicio de memoria para Claude Desktop automáticamente a través de [Smithery](https://smithery.ai/server/@doobidoo/mcp-memory-service) :\n\n```bash\nnpx -y @smithery/cli install @doobidoo/mcp-memory-service --client claude\n```\n\n### Guía de instalación detallada\n\nPara obtener instrucciones completas de instalación y solución de problemas, consulte la [Guía de instalación](docs/guides/installation.md) .\n\n## Configuración de Claude MCP\n\n### Configuración estándar\n\nAgregue lo siguiente a su archivo `claude_desktop_config.json` :\n\n```json\n{\n \"memory\": {\n \"command\": \"uv\",\n \"args\": [\n \"--directory\",\n \"your_mcp_memory_service_directory\", // e.g., \"C:\\\\REPOSITORIES\\\\mcp-memory-service\"\n \"run\",\n \"memory\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"your_chroma_db_path\", // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\"\n \"MCP_MEMORY_BACKUPS_PATH\": \"your_backups_path\" // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n### Configuración específica de Windows (recomendada)\n\nPara los usuarios de Windows, recomendamos utilizar el script contenedor para garantizar que PyTorch esté instalado correctamente:\n\n```json\n{\n \"memory\": {\n \"command\": \"python\",\n \"args\": [\n \"C:\\\\path\\\\to\\\\mcp-memory-service\\\\memory_wrapper.py\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\",\n \"MCP_MEMORY_BACKUPS_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\nEl script contenedor hará lo siguiente:\n\n1. Compruebe si PyTorch está instalado y configurado correctamente\n2. Instale PyTorch con la URL de índice correcta si es necesario\n3. Ejecute el servidor de memoria con la configuración adecuada\n\n## Guía de uso\n\nPara obtener instrucciones detalladas sobre cómo interactuar con el servicio de memoria en Claude Desktop:\n\n* [Guía de invocación](docs/guides/invocation_guide.md) : aprenda las palabras clave y frases específicas que activan las operaciones de memoria en Claude\n* [Guía de instalación](docs/guides/installation.md) : instrucciones de configuración detalladas\n\nEl servicio de memoria se invoca mediante comandos de lenguaje natural en tus conversaciones con Claude. Por ejemplo:\n\n* Para guardar: \"Por favor, recuerde que la fecha límite de mi proyecto es el 15 de mayo\".\n* Para recuperar: \"¿Recuerdas lo que te dije sobre la fecha límite de mi proyecto?\"\n* Para borrar: \"Por favor olvida lo que te dije sobre mi dirección\".\n\nConsulte la [Guía de invocación](docs/guides/invocation_guide.md) para obtener una lista completa de comandos y ejemplos de uso detallados.\n\n## Operaciones de memoria\n\nEl servicio de memoria proporciona las siguientes operaciones a través del servidor MCP:\n\n### Operaciones de memoria central\n\n1. `store_memory` - Almacena nueva información con etiquetas opcionales\n2. `retrieve_memory` - Realizar una búsqueda semántica de recuerdos relevantes\n3. `recall_memory` - Recupera recuerdos usando expresiones de tiempo en lenguaje natural\n4. `search_by_tag` - Encuentra recuerdos usando etiquetas específicas\n5. `exact_match_retrieve` - Encuentra recuerdos con coincidencia exacta de contenido\n6. `debug_retrieve` - Recupera memorias con puntuaciones de similitud\n\n### Gestión de bases de datos\n\n7. `create_backup` - Crear copia de seguridad de la base de datos\n8. `get_stats` - Obtener estadísticas de memoria\n9. `optimize_db` - Optimizar el rendimiento de la base de datos\n10. `check_database_health` - Obtener métricas de salud de la base de datos\n11. `check_embedding_model` - Verificar el estado del modelo\n\n### Gestión de la memoria\n\n12. `delete_memory` - Eliminar memoria específica por hash\n13. `delete_by_tag` - Elimina todos los recuerdos con una etiqueta específica\n14. `cleanup_duplicates` - Eliminar entradas duplicadas\n\n## Opciones de configuración\n\nConfigurar a través de variables de entorno:\n\n```\nCHROMA_DB_PATH: Path to ChromaDB storage\nBACKUP_PATH: Path for backups\nAUTO_BACKUP_INTERVAL: Backup interval in hours (default: 24)\nMAX_MEMORIES_BEFORE_OPTIMIZE: Threshold for auto-optimization (default: 10000)\nSIMILARITY_THRESHOLD: Default similarity threshold (default: 0.7)\nMAX_RESULTS_PER_QUERY: Maximum results per query (default: 10)\nBACKUP_RETENTION_DAYS: Number of days to keep backups (default: 7)\nLOG_LEVEL: Logging level (default: INFO)\n\n# Hardware-specific environment variables\nPYTORCH_ENABLE_MPS_FALLBACK: Enable MPS fallback for Apple Silicon (default: 1)\nMCP_MEMORY_USE_ONNX: Use ONNX Runtime for CPU-only deployments (default: 0)\nMCP_MEMORY_USE_DIRECTML: Use DirectML for Windows acceleration (default: 0)\nMCP_MEMORY_MODEL_NAME: Override the default embedding model\nMCP_MEMORY_BATCH_SIZE: Override the default batch size\n```\n\n## Compatibilidad de hardware\n\n| Plataforma | Arquitectura | Acelerador | Estado |\n| ---------- | ---------------------------- | ---------- | ----------------------------- |\n| macOS | Silicona de Apple (M1/M2/M3) | MPS | ✅ Totalmente compatible |\n| macOS | Apple Silicon bajo Rosetta 2 | UPC | ✅ Compatible con alternativas |\n| macOS | Intel | UPC | ✅ Totalmente compatible |\n| Ventanas | x86\\_64 | CUDA | ✅ Totalmente compatible |\n| Ventanas | x86\\_64 | DirectML | ✅ Compatible |\n| Ventanas | x86\\_64 | UPC | ✅ Compatible con alternativas |\n| Linux | x86\\_64 | CUDA | ✅ Totalmente compatible |\n| Linux | x86\\_64 | ROCm | ✅ Compatible |\n| Linux | x86\\_64 | UPC | ✅ Compatible con alternativas |\n| Linux | ARM64 | UPC | ✅ Compatible con alternativas |\n\n## Pruebas\n\n```bash\n# Install test dependencies\npip install pytest pytest-asyncio\n\n# Run all tests\npytest tests/\n\n# Run specific test categories\npytest tests/test_memory_ops.py\npytest tests/test_semantic_search.py\npytest tests/test_database.py\n\n# Verify environment compatibility\npython scripts/verify_environment_enhanced.py\n\n# Verify PyTorch installation on Windows\npython scripts/verify_pytorch_windows.py\n\n# Perform comprehensive installation verification\npython scripts/test_installation.py\n```\n\n## Solución de problemas\n\nConsulte la [Guía de instalación](docs/guides/installation.md#troubleshooting-common-installation-issues) para obtener pasos detallados para la solución de problemas.\n\n### Consejos rápidos para la solución de problemas\n\n* **Errores de PyTorch de Windows** : utilice `python scripts/install_windows.py`\n* **Conflictos de dependencia de Intel en macOS** : use `python install.py --force-compatible-deps`\n* **Errores de recursión** : Ejecute `python scripts/fix_sitecustomize.py`\n* **Verificación del entorno** : ejecutar `python scripts/verify_environment_enhanced.py`\n* **Problemas de memoria** : establezca `MCP_MEMORY_BATCH_SIZE=4` y pruebe un modelo más pequeño\n* **Apple Silicon** : asegúrese de que Python 3.10+ esté creado para ARM64, configure `PYTORCH_ENABLE_MPS_FALLBACK=1`\n* **Prueba de instalación** : ejecute `python scripts/test_installation.py`\n\n## Estructura del proyecto\n\n```\nmcp-memory-service/\n├── src/mcp_memory_service/ # Core package code\n│ ├── __init__.py\n│ ├── config.py # Configuration utilities\n│ ├── models/ # Data models\n│ ├── storage/ # Storage implementations\n│ ├── utils/ # Utility functions\n│ └── server.py # Main MCP server\n├── scripts/ # Helper scripts\n├── memory_wrapper.py # Windows wrapper script\n├── install.py # Enhanced installation script\n└── tests/ # Test suite\n```\n\n## Directrices de desarrollo\n\n* Python 3.10+ con sugerencias de tipos\n* Utilice clases de datos para modelos\n* Documentstrings entre comillas triples para módulos y funciones\n* Patrón asíncrono/en espera para todas las operaciones de E/S\n* Siga las pautas de estilo de PEP 8\n* Incluir pruebas para nuevas funciones\n\n## Licencia\n\nLicencia MIT: consulte el archivo de LICENCIA para obtener más detalles\n\n## Expresiones de gratitud\n\n* Equipo de ChromaDB para la base de datos vectorial\n* Proyecto Transformadores de Sentencias para incrustar modelos\n* Proyecto MCP para la especificación del protocolo\n\n## Contacto\n\n[Telegrama](t.me/doobeedoo)\n\n## Integraciones\n\nEl servicio de memoria MCP se puede ampliar con diversas herramientas y utilidades. Consulte [Integraciones](docs/integrations.md) para ver una lista de las opciones disponibles, entre ellas:\n\n* [Panel de memoria MCP](https://github.com/doobidoo/mcp-memory-dashboard) : interfaz web para explorar y administrar memorias\n* [Contexto de memoria de Claude](https://github.com/doobidoo/claude-memory-context) : Inyectar contexto de memoria en las instrucciones del proyecto Claude","Spanish","es-ES",{"_29":55,"_33":59,"_38":60},{"_31":56,"_33":57,"_35":58},"強化された検索および取得機能のために ChromaDB とセンテンストランスフォーマーを活用し、Claude にセマンティックメモリと永続ストレージを提供します。","MCPメモリサービス","# MCPメモリサービス\n\n[![ライセンス: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![鍛冶屋のバッジ](https://smithery.ai/badge/@doobidoo/mcp-memory-service)](https://smithery.ai/server/@doobidoo/mcp-memory-service)\n\nChromaDBとセンテンストランスフォーマーを活用し、Claude Desktopにセマンティックメモリと永続ストレージ機能を提供するMCPサーバー。このサービスは、セマンティック検索機能を備えた長期メモリストレージを可能にし、会話やインスタンス間でのコンテキスト維持に最適です。\n\n## ヘルプ\n\n[TalkToGitHub](https://talktogithub.com/doobidoo/mcp-memory-service)でリポジトリと話しましょう!\n\n## 特徴\n\n* 文変換を用いた意味検索\n* **自然言語による時間ベースの想起**（例：「先週」、「昨日の朝」）\n* タグベースの記憶検索システム\n* ChromaDB を使用した永続ストレージ\n* 自動データベースバックアップ\n* メモリ最適化ツール\n* 完全一致検索\n* 類似性分析のデバッグモード\n* データベースの健全性監視\n* 重複検出とクリーンアップ\n* カスタマイズ可能な埋め込みモデル\n* **クロスプラットフォーム互換性**（Apple Silicon、Intel、Windows、Linux）\n* さまざまな環境に合わせた**ハードウェア対応の最適化**\n* 限られたハードウェアリソースに対する**適切なフォールバック**\n\n## インストール\n\n### クイックスタート（推奨）\n\n拡張インストールスクリプトはシステムを自動的に検出し、適切な依存関係をインストールします。\n\n```bash\n# Clone the repository\ngit clone https://github.com/doobidoo/mcp-memory-service.git\ncd mcp-memory-service\n\n# Create and activate a virtual environment\npython -m venv venv\nsource venv/bin/activate # On Windows: venv\\Scripts\\activate\n\n# Run the installation script\npython install.py\n```\n\n`install.py`スクリプトは次の処理を実行します。\n\n1. システムアーキテクチャと利用可能なハードウェアアクセラレータを検出します\n2. プラットフォームに適した依存関係をインストールします\n3. 環境に最適な設定を構成する\n4. インストールを確認し、必要に応じて診断を提供します\n\n### Dockerのインストール\n\nDocker を使用してメモリサービスを実行できます。\n\n```bash\n# Using Docker Compose (recommended)\ndocker-compose up\n\n# Using Docker directly\ndocker build -t mcp-memory-service .\ndocker run -p 8000:8000 -v /path/to/data:/app/chroma_db -v /path/to/backups:/app/backups mcp-memory-service\n```\n\nさまざまなシナリオに合わせて複数の Docker Compose 構成を提供します。\n\n* `docker-compose.yml` - pip install を使用した標準設定\n* `docker-compose.uv.yml` - UV パッケージマネージャーを使用した代替構成\n* `docker-compose.pythonpath.yml` - 明示的な PYTHONPATH 設定による構成\n\n代替構成を使用するには:\n\n```bash\ndocker-compose -f docker-compose.uv.yml up\n```\n\n### Windows インストール (特殊なケース)\n\nWindows ユーザーは、プラットフォーム固有のホイールの可用性により、PyTorch のインストールで問題が発生する可能性があります。Windows 専用のインストールスクリプトをご利用ください。\n\n```bash\n# After activating your virtual environment\npython scripts/install_windows.py\n```\n\nこのスクリプトは次の処理を行います。\n\n1. CUDAの可用性とバージョンの検出\n2. 正しいインデックス URL から適切な PyTorch バージョンをインストールする\n3. PyTorchと競合せずに他の依存関係をインストールする\n4. インストールの確認\n\n### Smithery経由でインストール\n\n[Smithery](https://smithery.ai/server/@doobidoo/mcp-memory-service)経由で Claude Desktop のメモリサービスを自動的にインストールするには:\n\n```bash\nnpx -y @smithery/cli install @doobidoo/mcp-memory-service --client claude\n```\n\n### 詳細なインストールガイド\n\n包括的なインストール手順とトラブルシューティングについては、『[インストールガイド』](docs/guides/installation.md)を参照してください。\n\n## クロードMCP構成\n\n### 標準構成\n\n`claude_desktop_config.json`ファイルに次の内容を追加します。\n\n```json\n{\n \"memory\": {\n \"command\": \"uv\",\n \"args\": [\n \"--directory\",\n \"your_mcp_memory_service_directory\", // e.g., \"C:\\\\REPOSITORIES\\\\mcp-memory-service\"\n \"run\",\n \"memory\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"your_chroma_db_path\", // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\"\n \"MCP_MEMORY_BACKUPS_PATH\": \"your_backups_path\" // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n### Windows 固有の構成 (推奨)\n\nWindows ユーザーの場合は、PyTorch が適切にインストールされていることを確認するために、ラッパースクリプトを使用することをお勧めします。\n\n```json\n{\n \"memory\": {\n \"command\": \"python\",\n \"args\": [\n \"C:\\\\path\\\\to\\\\mcp-memory-service\\\\memory_wrapper.py\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\",\n \"MCP_MEMORY_BACKUPS_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\nラッパースクリプトは次の処理を実行します。\n\n1. PyTorchがインストールされ、適切に設定されているか確認する\n2. 必要に応じて正しいインデックス URL で PyTorch をインストールします\n3. 適切な構成でメモリサーバーを実行する\n\n## 使用ガイド\n\nClaude Desktop のメモリサービスと対話する方法の詳細な手順については、以下を参照してください。\n\n* [呼び出しガイド](docs/guides/invocation_guide.md)- クロードでメモリ操作をトリガーする特定のキーワードとフレーズを学びます\n* [インストールガイド](docs/guides/installation.md)- 詳細なセットアップ手順\n\nメモリサービスは、クロードとの会話の中で自然言語コマンドを通じて呼び出されます。例えば：\n\n* 店舗へ: 「私のプロジェクトの締め切りは 5 月 15 日です。覚えておいてください。」\n* 取得するには: 「プロジェクトの締め切りについて私が言ったことを覚えていますか?」\n* 削除するには: 「私の住所についてお伝えしたことは忘れてください。」\n\nコマンドの完全なリストと詳細な使用例については、[呼び出しガイド](docs/guides/invocation_guide.md)を参照してください。\n\n## メモリ操作\n\nメモリサービスは、MCP サーバーを通じて次の操作を提供します。\n\n### コアメモリ操作\n\n1. `store_memory` - オプションのタグを使用して新しい情報を保存する\n2. `retrieve_memory` - 関連する記憶の意味的検索を実行する\n3. `recall_memory` - 自然言語の時間表現を使って記憶を思い出す\n4. `search_by_tag` - 特定のタグを使って思い出を探す\n5. `exact_match_retrieve` - コンテンツが完全に一致する思い出を検索\n6. `debug_retrieve` - 類似度スコアで記憶を取得する\n\n### データベース管理\n\n7. `create_backup` - データベースのバックアップを作成する\n8. `get_stats` - メモリ統計を取得する\n9. `optimize_db` - データベースのパフォーマンスを最適化する\n10. `check_database_health` - データベースの健全性メトリクスを取得する\n11. `check_embedding_model` - モデルの状態を確認する\n\n### メモリ管理\n\n12. `delete_memory` - ハッシュで特定のメモリを削除する\n13. `delete_by_tag` - 特定のタグが付いたすべての思い出を削除する\n14. `cleanup_duplicates` - 重複したエントリを削除する\n\n## 設定オプション\n\n環境変数を使用して設定します。\n\n```\nCHROMA_DB_PATH: Path to ChromaDB storage\nBACKUP_PATH: Path for backups\nAUTO_BACKUP_INTERVAL: Backup interval in hours (default: 24)\nMAX_MEMORIES_BEFORE_OPTIMIZE: Threshold for auto-optimization (default: 10000)\nSIMILARITY_THRESHOLD: Default similarity threshold (default: 0.7)\nMAX_RESULTS_PER_QUERY: Maximum results per query (default: 10)\nBACKUP_RETENTION_DAYS: Number of days to keep backups (default: 7)\nLOG_LEVEL: Logging level (default: INFO)\n\n# Hardware-specific environment variables\nPYTORCH_ENABLE_MPS_FALLBACK: Enable MPS fallback for Apple Silicon (default: 1)\nMCP_MEMORY_USE_ONNX: Use ONNX Runtime for CPU-only deployments (default: 0)\nMCP_MEMORY_USE_DIRECTML: Use DirectML for Windows acceleration (default: 0)\nMCP_MEMORY_MODEL_NAME: Override the default embedding model\nMCP_MEMORY_BATCH_SIZE: Override the default batch size\n```\n\n## ハードウェアの互換性\n\n| プラットフォーム | 建築 | アクセル | 状態 |\n| -------- | --------------------------- | ------- | -------------------- |\n| macOS | Appleシリコン（M1/M2/M3） | MPS | ✅ 完全にサポートされています |\n| macOS | Rosetta 2 搭載の Apple Silicon | CPU | ✅ フォールバックでサポートされています |\n| macOS | インテル | CPU | ✅ 完全にサポートされています |\n| ウィンドウズ | x86\\_64 | CUDA | ✅ 完全にサポートされています |\n| ウィンドウズ | x86\\_64 | ダイレクトML | ✅ サポートされています |\n| ウィンドウズ | x86\\_64 | CPU | ✅ フォールバックでサポートされています |\n| リナックス | x86\\_64 | CUDA | ✅ 完全にサポートされています |\n| リナックス | x86\\_64 | ROCm | ✅ サポートされています |\n| リナックス | x86\\_64 | CPU | ✅ フォールバックでサポートされています |\n| リナックス | ARM64 | CPU | ✅ フォールバックでサポートされています |\n\n## テスト\n\n```bash\n# Install test dependencies\npip install pytest pytest-asyncio\n\n# Run all tests\npytest tests/\n\n# Run specific test categories\npytest tests/test_memory_ops.py\npytest tests/test_semantic_search.py\npytest tests/test_database.py\n\n# Verify environment compatibility\npython scripts/verify_environment_enhanced.py\n\n# Verify PyTorch installation on Windows\npython scripts/verify_pytorch_windows.py\n\n# Perform comprehensive installation verification\npython scripts/test_installation.py\n```\n\n## トラブルシューティング\n\n詳細なトラブルシューティング手順については、[インストールガイド](docs/guides/installation.md#troubleshooting-common-installation-issues)を参照してください。\n\n### 簡単なトラブルシューティングのヒント\n\n* **Windows PyTorch エラー**: `python scripts/install_windows.py`を使用する\n* **macOS Intel 依存関係の競合**: `python install.py --force-compatible-deps`を使用する\n* **再帰エラー**: `python scripts/fix_sitecustomize.py`を実行します\n* **環境検証**: `python scripts/verify_environment_enhanced.py`を実行します。\n* **メモリの問題**: `MCP_MEMORY_BATCH_SIZE=4`に設定し、より小さいモデルを試してください\n* **Apple Silicon** : Python 3.10+ が ARM64 用にビルドされていることを確認し、 `PYTORCH_ENABLE_MPS_FALLBACK=1`設定します\n* **インストールテスト**: `python scripts/test_installation.py`を実行します。\n\n## プロジェクト構造\n\n```\nmcp-memory-service/\n├── src/mcp_memory_service/ # Core package code\n│ ├── __init__.py\n│ ├── config.py # Configuration utilities\n│ ├── models/ # Data models\n│ ├── storage/ # Storage implementations\n│ ├── utils/ # Utility functions\n│ └── server.py # Main MCP server\n├── scripts/ # Helper scripts\n├── memory_wrapper.py # Windows wrapper script\n├── install.py # Enhanced installation script\n└── tests/ # Test suite\n```\n\n## 開発ガイドライン\n\n* 型ヒント付きのPython 3.10+\n* モデルにデータクラスを使用する\n* モジュールと関数の三重引用符で囲まれたドキュメント文字列\n* すべてのI/O操作の非同期/待機パターン\n* PEP 8スタイルガイドラインに従う\n* 新機能のテストを含める\n\n## ライセンス\n\nMITライセンス - 詳細はLICENSEファイルを参照\n\n## 謝辞\n\n* ベクターデータベースのChromaDBチーム\n* 埋め込みモデルのためのSentence Transformersプロジェクト\n* プロトコル仕様のMCPプロジェクト\n\n## 接触\n\n[電報](t.me/doobeedoo)\n\n## 統合\n\nMCPメモリサービスは、様々なツールやユーティリティで拡張できます。利用可能なオプションの一覧については、 [「統合」を](docs/integrations.md)ご覧ください。\n\n* [MCP メモリダッシュボード](https://github.com/doobidoo/mcp-memory-dashboard)- メモリの閲覧と管理のための Web UI\n* [Claude メモリコンテキスト](https://github.com/doobidoo/claude-memory-context)- Claude プロジェクトの指示にメモリコンテキストを挿入します","Japanese","ja-JP",{"_29":62,"_33":66,"_38":67},{"_31":63,"_33":64,"_35":65},"클로드에게 의미적 메모리와 영구 저장소를 제공하고, ChromaDB와 문장 변환기를 활용하여 향상된 검색 및 조회 기능을 제공합니다.","MCP 메모리 서비스","# MCP 메모리 서비스\n\n[![라이센스: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![대장간 배지](https://smithery.ai/badge/@doobidoo/mcp-memory-service)](https://smithery.ai/server/@doobidoo/mcp-memory-service)\n\nChromaDB와 문장 변환기를 사용하여 Claude Desktop에 의미 메모리와 영구 저장 기능을 제공하는 MCP 서버입니다. 이 서비스는 의미 검색 기능을 갖춘 장기 메모리 저장 기능을 제공하여 대화와 인스턴스 전반의 맥락을 유지하는 데 이상적입니다.\n\n## 돕다\n\n[TalkToGitHub](https://talktogithub.com/doobidoo/mcp-memory-service) 에서 저장소와 대화해보세요!\n\n## 특징\n\n* 문장 변환기를 사용한 의미 검색\n* **자연어 시간 기반 리콜** (예: \"지난주\", \"어제 아침\")\n* 태그 기반 메모리 검색 시스템\n* ChromaDB를 사용한 영구 저장소\n* 자동 데이터베이스 백업\n* 메모리 최적화 도구\n* 정확한 일치 검색\n* 유사성 분석을 위한 디버그 모드\n* 데이터베이스 상태 모니터링\n* 중복 감지 및 정리\n* 사용자 정의 가능한 임베딩 모델\n* **크로스 플랫폼 호환성** (Apple Silicon, Intel, Windows, Linux)\n* 다양한 환경에 대한 **하드웨어 인식 최적화**\n* 제한된 하드웨어 리소스에 대한 **우아한 대체**\n\n## 설치\n\n### 빠른 시작(권장)\n\n향상된 설치 스크립트는 자동으로 시스템을 감지하고 적절한 종속성을 설치합니다.\n\n지엑스피1\n\n`install.py` 스크립트는 다음을 수행합니다.\n\n1. 시스템 아키텍처와 사용 가능한 하드웨어 가속기를 감지합니다.\n2. 플랫폼에 적합한 종속성을 설치하세요\n3. 사용자 환경에 맞는 최적의 설정을 구성하세요\n4. 설치를 확인하고 필요한 경우 진단을 제공합니다.\n\n### 도커 설치\n\nDocker를 사용하여 메모리 서비스를 실행할 수 있습니다.\n\n```bash\n# Using Docker Compose (recommended)\ndocker-compose up\n\n# Using Docker directly\ndocker build -t mcp-memory-service .\ndocker run -p 8000:8000 -v /path/to/data:/app/chroma_db -v /path/to/backups:/app/backups mcp-memory-service\n```\n\n다양한 시나리오에 맞춰 여러 가지 Docker Compose 구성을 제공합니다.\n\n* `docker-compose.yml` - pip install을 사용한 표준 구성\n* `docker-compose.uv.yml` - UV 패키지 관리자를 사용한 대체 구성\n* `docker-compose.pythonpath.yml` - 명시적 PYTHONPATH 설정을 사용한 구성\n\n대체 구성을 사용하려면:\n\n```bash\ndocker-compose -f docker-compose.uv.yml up\n```\n\n### Windows 설치(특수한 경우)\n\nWindows 사용자는 플랫폼별 휠 가용성으로 인해 PyTorch 설치 문제가 발생할 수 있습니다. Windows 전용 설치 스크립트를 사용하세요.\n\n```bash\n# After activating your virtual environment\npython scripts/install_windows.py\n```\n\n이 스크립트는 다음을 처리합니다.\n\n1. CUDA 가용성 및 버전 감지\n2. 올바른 인덱스 URL에서 적절한 PyTorch 버전 설치\n3. PyTorch와 충돌하지 않고 다른 종속성 설치\n4. 설치 확인\n\n### Smithery를 통해 설치\n\n[Smithery를](https://smithery.ai/server/@doobidoo/mcp-memory-service) 통해 Claude Desktop용 메모리 서비스를 자동으로 설치하려면:\n\n```bash\nnpx -y @smithery/cli install @doobidoo/mcp-memory-service --client claude\n```\n\n### 자세한 설치 가이드\n\n포괄적인 설치 지침과 문제 해결에 대해서는 [설치 가이드를](docs/guides/installation.md) 참조하세요.\n\n## 클로드 MCP 구성\n\n### 표준 구성\n\n`claude_desktop_config.json` 파일에 다음을 추가하세요.\n\n```json\n{\n \"memory\": {\n \"command\": \"uv\",\n \"args\": [\n \"--directory\",\n \"your_mcp_memory_service_directory\", // e.g., \"C:\\\\REPOSITORIES\\\\mcp-memory-service\"\n \"run\",\n \"memory\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"your_chroma_db_path\", // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\"\n \"MCP_MEMORY_BACKUPS_PATH\": \"your_backups_path\" // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n### Windows 특정 구성(권장)\n\nWindows 사용자의 경우 PyTorch가 제대로 설치되었는지 확인하기 위해 래퍼 스크립트를 사용하는 것이 좋습니다.\n\n```json\n{\n \"memory\": {\n \"command\": \"python\",\n \"args\": [\n \"C:\\\\path\\\\to\\\\mcp-memory-service\\\\memory_wrapper.py\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\",\n \"MCP_MEMORY_BACKUPS_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n래퍼 스크립트는 다음을 수행합니다.\n\n1. PyTorch가 설치되고 올바르게 구성되었는지 확인하세요.\n2. 필요한 경우 올바른 인덱스 URL로 PyTorch를 설치하세요.\n3. 적절한 구성으로 메모리 서버를 실행하세요\n\n## 사용 가이드\n\nClaude Desktop에서 메모리 서비스와 상호 작용하는 방법에 대한 자세한 지침:\n\n* [호출 가이드](docs/guides/invocation_guide.md) - Claude에서 메모리 작업을 트리거하는 특정 키워드와 구문을 알아보세요\n* [설치 가이드](docs/guides/installation.md) - 자세한 설정 지침\n\n메모리 서비스는 클로드와의 대화에서 자연어 명령을 통해 호출됩니다. 예:\n\n* 저장 방법: \"제 프로젝트 마감일은 5월 15일임을 기억해 주세요.\"\n* 검색: \"프로젝트 마감일과 관련해서 제가 말씀드린 걸 기억하시나요?\"\n* 삭제 방법: \"제 주소에 대해 제가 말한 내용은 잊어주세요.\"\n\n전체 명령어 목록과 자세한 사용 예는 [호출 가이드를](docs/guides/invocation_guide.md) 참조하세요.\n\n## 메모리 작업\n\n메모리 서비스는 MCP 서버를 통해 다음과 같은 작업을 제공합니다.\n\n### 코어 메모리 작업\n\n1. `store_memory` - 선택적 태그를 사용하여 새 정보를 저장합니다.\n2. `retrieve_memory` - 관련 메모리에 대한 의미 검색을 수행합니다.\n3. `recall_memory` - 자연어 시간 표현을 사용하여 메모리 검색\n4. `search_by_tag` - 특정 태그를 사용하여 추억 찾기\n5. `exact_match_retrieve` - 정확한 콘텐츠 일치로 메모리 찾기\n6. `debug_retrieve` - 유사도 점수로 메모리 검색\n\n### 데이터베이스 관리\n\n7. `create_backup` - 데이터베이스 백업 생성\n8. `get_stats` - 메모리 통계 가져오기\n9. `optimize_db` - 데이터베이스 성능 최적화\n10. `check_database_health` - 데이터베이스 상태 지표 가져오기\n11. `check_embedding_model` - 모델 상태 확인\n\n### 메모리 관리\n\n12. `delete_memory` - 해시로 특정 메모리 삭제\n13. `delete_by_tag` - 특정 태그가 있는 모든 메모리 삭제\n14. `cleanup_duplicates` - 중복 항목 제거\n\n## 구성 옵션\n\n환경 변수를 통해 구성:\n\n```\nCHROMA_DB_PATH: Path to ChromaDB storage\nBACKUP_PATH: Path for backups\nAUTO_BACKUP_INTERVAL: Backup interval in hours (default: 24)\nMAX_MEMORIES_BEFORE_OPTIMIZE: Threshold for auto-optimization (default: 10000)\nSIMILARITY_THRESHOLD: Default similarity threshold (default: 0.7)\nMAX_RESULTS_PER_QUERY: Maximum results per query (default: 10)\nBACKUP_RETENTION_DAYS: Number of days to keep backups (default: 7)\nLOG_LEVEL: Logging level (default: INFO)\n\n# Hardware-specific environment variables\nPYTORCH_ENABLE_MPS_FALLBACK: Enable MPS fallback for Apple Silicon (default: 1)\nMCP_MEMORY_USE_ONNX: Use ONNX Runtime for CPU-only deployments (default: 0)\nMCP_MEMORY_USE_DIRECTML: Use DirectML for Windows acceleration (default: 0)\nMCP_MEMORY_MODEL_NAME: Override the default embedding model\nMCP_MEMORY_BATCH_SIZE: Override the default batch size\n```\n\n## 하드웨어 호환성\n\n| 플랫폼 | 건축학 | 촉진 신경 | 상태 |\n| --- | --------------------------- | ------ | --------- |\n| 맥OS | 애플 실리콘(M1/M2/M3) | MPS | ✅ 완벽히 지원됨 |\n| 맥OS | Rosetta 2 아래의 Apple Silicon | CPU | ✅ 폴백 지원 |\n| 맥OS | 인텔 | CPU | ✅ 완벽히 지원됨 |\n| 윈도우 | x86\\_64 | 쿠다 | ✅ 완벽히 지원됨 |\n| 윈도우 | x86\\_64 | 다이렉트ML | ✅ 지원됨 |\n| 윈도우 | x86\\_64 | CPU | ✅ 폴백 지원 |\n| 리눅스 | x86\\_64 | 쿠다 | ✅ 완벽히 지원됨 |\n| 리눅스 | x86\\_64 | ROCm | ✅ 지원됨 |\n| 리눅스 | x86\\_64 | CPU | ✅ 폴백 지원 |\n| 리눅스 | ARM64 | CPU | ✅ 폴백 지원 |\n\n## 테스트\n\n```bash\n# Install test dependencies\npip install pytest pytest-asyncio\n\n# Run all tests\npytest tests/\n\n# Run specific test categories\npytest tests/test_memory_ops.py\npytest tests/test_semantic_search.py\npytest tests/test_database.py\n\n# Verify environment compatibility\npython scripts/verify_environment_enhanced.py\n\n# Verify PyTorch installation on Windows\npython scripts/verify_pytorch_windows.py\n\n# Perform comprehensive installation verification\npython scripts/test_installation.py\n```\n\n## 문제 해결\n\n자세한 문제 해결 단계는 [설치 가이드를](docs/guides/installation.md#troubleshooting-common-installation-issues) 참조하세요.\n\n### 빠른 문제 해결 팁\n\n* **Windows PyTorch 오류** : `python scripts/install_windows.py` 사용하세요\n* **macOS Intel 종속성 충돌** : `python install.py --force-compatible-deps` 사용하세요\n* **재귀 오류** : `python scripts/fix_sitecustomize.py` 실행하세요\n* **환경 검증** : `python scripts/verify_environment_enhanced.py` 실행합니다.\n* **메모리 문제** : `MCP_MEMORY_BATCH_SIZE=4` 설정하고 더 작은 모델을 시도해 보세요.\n* **Apple Silicon** : ARM64용으로 빌드된 Python 3.10+를 확인하고 `PYTORCH_ENABLE_MPS_FALLBACK=1` 설정합니다.\n* **설치 테스트** : `python scripts/test_installation.py` 실행합니다.\n\n## 프로젝트 구조\n\n```\nmcp-memory-service/\n├── src/mcp_memory_service/ # Core package code\n│ ├── __init__.py\n│ ├── config.py # Configuration utilities\n│ ├── models/ # Data models\n│ ├── storage/ # Storage implementations\n│ ├── utils/ # Utility functions\n│ └── server.py # Main MCP server\n├── scripts/ # Helper scripts\n├── memory_wrapper.py # Windows wrapper script\n├── install.py # Enhanced installation script\n└── tests/ # Test suite\n```\n\n## 개발 지침\n\n* 유형 힌트가 포함된 Python 3.10+\n* 모델에 데이터 클래스를 사용하세요\n* 모듈 및 함수에 대한 삼중 따옴표로 묶인 문서 문자열\n* 모든 I/O 작업에 대한 비동기/대기 패턴\n* PEP 8 스타일 가이드라인을 따르세요\n* 새로운 기능에 대한 테스트를 포함합니다\n\n## 특허\n\nMIT 라이선스 - 자세한 내용은 라이선스 파일을 참조하세요.\n\n## 감사의 말\n\n* 벡터 데이터베이스를 위한 ChromaDB 팀\n* 모델 임베딩을 위한 문장 변환기 프로젝트\n* 프로토콜 사양을 위한 MCP 프로젝트\n\n## 연락하다\n\n[전보](t.me/doobeedoo)\n\n## 통합\n\nMCP 메모리 서비스는 다양한 도구와 유틸리티를 통해 확장할 수 있습니다. 사용 가능한 옵션 목록은 [통합을](docs/integrations.md) 참조하세요.\n\n* [MCP 메모리 대시보드](https://github.com/doobidoo/mcp-memory-dashboard) - 메모리 탐색 및 관리를 위한 웹 UI\n* [Claude 메모리 컨텍스트](https://github.com/doobidoo/claude-memory-context) - Claude 프로젝트 지침에 메모리 컨텍스트 주입","Korean","ko-KR",{"_29":69,"_33":73,"_38":74},{"_31":70,"_33":71,"_35":72},"Предоставляет семантическую память и постоянное хранилище для Клода, используя ChromaDB и преобразователи предложений для расширенных возможностей поиска и извлечения.","Служба памяти MCP","# Служба памяти MCP\n\n[![Лицензия: Массачусетский технологический институт](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![значок кузнеца](https://smithery.ai/badge/@doobidoo/mcp-memory-service)](https://smithery.ai/server/@doobidoo/mcp-memory-service) [![Проверено на MseeP](https://mseep.ai/badge.svg)](https://mseep.ai/app/0513fb92-e941-4fe0-9948-2a1dbb870dcf)\n\nMCP-сервер, предоставляющий возможности семантической памяти и постоянного хранения для Claude Desktop с использованием ChromaDB и преобразователей предложений. Этот сервис обеспечивает долгосрочное хранение памяти с возможностями семантического поиска, что делает его идеальным для поддержания контекста в разговорах и случаях.\n\n## Помощь\n\nПообщайтесь с репозиторием с помощью [TalkToGitHub](https://talktogithub.com/doobidoo/mcp-memory-service) !\n\n## Функции\n\n* Семантический поиск с использованием преобразователей предложений\n* **Припоминание естественного языка с учетом времени** (например, «на прошлой неделе», «вчера утром»)\n* Система извлечения памяти на основе тегов\n* Постоянное хранилище с использованием ChromaDB\n* Автоматическое резервное копирование баз данных\n* Инструменты оптимизации памяти\n* Поиск точного соответствия\n* Режим отладки для анализа сходства\n* Мониторинг работоспособности базы данных\n* Обнаружение и очистка дубликатов\n* Настраиваемая модель встраивания\n* **Кроссплатформенная совместимость** (Apple Silicon, Intel, Windows, Linux)\n* **Аппаратно-ориентированная оптимизация** для различных сред\n* **Изящные откаты** при ограниченных аппаратных ресурсах\n\n## Установка\n\n### Быстрый старт (рекомендуется)\n\nРасширенный скрипт установки автоматически обнаруживает вашу систему и устанавливает соответствующие зависимости:\n\n```bash\n# Clone the repository\ngit clone https://github.com/doobidoo/mcp-memory-service.git\ncd mcp-memory-service\n\n# Create and activate a virtual environment\npython -m venv venv\nsource venv/bin/activate # On Windows: venv\\Scripts\\activate\n\n# Run the installation script\npython install.py\n```\n\nСкрипт `install.py` выполнит следующие действия:\n\n1. Определите архитектуру вашей системы и доступные аппаратные ускорители\n2. Установите соответствующие зависимости для вашей платформы.\n3. Настройте оптимальные параметры для вашей среды\n4. Проверьте установку и при необходимости проведите диагностику.\n\n### Установка докера\n\nВы можете запустить службу памяти с помощью Docker:\n\n```bash\n# Using Docker Compose (recommended)\ndocker-compose up\n\n# Using Docker directly\ndocker build -t mcp-memory-service .\ndocker run -p 8000:8000 -v /path/to/data:/app/chroma_db -v /path/to/backups:/app/backups mcp-memory-service\n```\n\nМы предоставляем несколько конфигураций Docker Compose для различных сценариев:\n\n* `docker-compose.yml` — стандартная конфигурация с использованием pip install\n* `docker-compose.uv.yml` — Альтернативная конфигурация с использованием менеджера пакетов UV\n* `docker-compose.pythonpath.yml` — Конфигурация с явными настройками PYTHONPATH\n\nЧтобы использовать альтернативную конфигурацию:\n\n```bash\ndocker-compose -f docker-compose.uv.yml up\n```\n\n### Установка Windows (особый случай)\n\nПользователи Windows могут столкнуться с проблемами установки PyTorch из-за доступности колеса для определенной платформы. Используйте наш скрипт установки для Windows:\n\n```bash\n# After activating your virtual environment\npython scripts/install_windows.py\n```\n\nЭтот скрипт обрабатывает:\n\n1. Определение доступности и версии CUDA\n2. Установка соответствующей версии PyTorch с правильного индексного URL-адреса\n3. Установка других зависимостей без конфликта с PyTorch\n4. Проверка установки\n\n### Установка через Smithery\n\nЧтобы автоматически установить Memory Service для Claude Desktop через [Smithery](https://smithery.ai/server/@doobidoo/mcp-memory-service) :\n\n```bash\nnpx -y @smithery/cli install @doobidoo/mcp-memory-service --client claude\n```\n\n### Подробное руководство по установке\n\nПодробные инструкции по установке и устранению неполадок см. [в Руководстве по установке](docs/guides/installation.md) .\n\n## Конфигурация Клода MCP\n\n### Стандартная конфигурация\n\nДобавьте следующее в файл `claude_desktop_config.json` :\n\n```json\n{\n \"memory\": {\n \"command\": \"uv\",\n \"args\": [\n \"--directory\",\n \"your_mcp_memory_service_directory\", // e.g., \"C:\\\\REPOSITORIES\\\\mcp-memory-service\"\n \"run\",\n \"memory\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"your_chroma_db_path\", // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\"\n \"MCP_MEMORY_BACKUPS_PATH\": \"your_backups_path\" // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n### Конфигурация, специфичная для Windows (рекомендуется)\n\nПользователям Windows мы рекомендуем использовать скрипт-оболочку, чтобы убедиться, что PyTorch установлен правильно:\n\n```json\n{\n \"memory\": {\n \"command\": \"python\",\n \"args\": [\n \"C:\\\\path\\\\to\\\\mcp-memory-service\\\\memory_wrapper.py\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\",\n \"MCP_MEMORY_BACKUPS_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\nСкрипт-обертка будет:\n\n1. Проверьте, установлен ли PyTorch и правильно ли он настроен.\n2. При необходимости установите PyTorch с правильным индексным URL-адресом.\n3. Запустите сервер памяти с соответствующей конфигурацией.\n\n## Руководство по использованию\n\nПодробные инструкции по взаимодействию со службой памяти в Claude Desktop:\n\n* [Руководство по вызову](docs/guides/invocation_guide.md) — изучите конкретные ключевые слова и фразы, которые запускают операции с памятью в Claude\n* [Руководство по установке](docs/guides/installation.md) - Подробные инструкции по настройке\n\nСлужба памяти вызывается посредством команд естественного языка в ваших разговорах с Клодом. Например:\n\n* Сохранить: «Пожалуйста, помните, что крайний срок сдачи моего проекта — 15 мая».\n* Извлечь: «Помнишь, что я говорил тебе о сроках сдачи моего проекта?»\n* Удалить: «Пожалуйста, забудьте то, что я вам говорил о моем адресе».\n\nПолный список команд и подробные примеры использования см. в [Руководстве по вызову.](docs/guides/invocation_guide.md)\n\n## Операции с памятью\n\nСлужба памяти обеспечивает выполнение следующих операций через сервер MCP:\n\n### Операции с основной памятью\n\n1. `store_memory` — сохранение новой информации с дополнительными тегами\n2. `retrieve_memory` — выполнить семантический поиск соответствующих воспоминаний\n3. `recall_memory` - Извлечение воспоминаний с использованием выражений времени на естественном языке\n4. `search_by_tag` — поиск воспоминаний с использованием определенных тегов\n5. `exact_match_retrieve` — поиск воспоминаний с точным совпадением содержания\n6. `debug_retrieve` — извлечение воспоминаний с оценками сходства\n\n### Управление базой данных\n\n7. `create_backup` — Создать резервную копию базы данных\n8. `get_stats` - Получить статистику памяти\n9. `optimize_db` — Оптимизация производительности базы данных\n10. `check_database_health` — получение показателей работоспособности базы данных\n11. `check_embedding_model` — проверка статуса модели\n\n### Управление памятью\n\n12. `delete_memory` - Удалить определенную память по хешу\n13. `delete_by_tag` — Удалить все воспоминания с определенным тегом\n14. `cleanup_duplicates` - Удалить дубликаты записей\n\n## Параметры конфигурации\n\nНастройте через переменные среды:\n\n```\nCHROMA_DB_PATH: Path to ChromaDB storage\nBACKUP_PATH: Path for backups\nAUTO_BACKUP_INTERVAL: Backup interval in hours (default: 24)\nMAX_MEMORIES_BEFORE_OPTIMIZE: Threshold for auto-optimization (default: 10000)\nSIMILARITY_THRESHOLD: Default similarity threshold (default: 0.7)\nMAX_RESULTS_PER_QUERY: Maximum results per query (default: 10)\nBACKUP_RETENTION_DAYS: Number of days to keep backups (default: 7)\nLOG_LEVEL: Logging level (default: INFO)\n\n# Hardware-specific environment variables\nPYTORCH_ENABLE_MPS_FALLBACK: Enable MPS fallback for Apple Silicon (default: 1)\nMCP_MEMORY_USE_ONNX: Use ONNX Runtime for CPU-only deployments (default: 0)\nMCP_MEMORY_USE_DIRECTML: Use DirectML for Windows acceleration (default: 0)\nMCP_MEMORY_MODEL_NAME: Override the default embedding model\nMCP_MEMORY_BATCH_SIZE: Override the default batch size\n```\n\n## Совместимость оборудования\n\n| Платформа | Архитектура | Ускоритель | Статус |\n| --------- | --------------------------- | ---------- | ---------------------------------------- |\n| macOS | Apple Silicon (M1/M2/M3) | МПС | ✅ Полностью поддерживается |\n| macOS | Apple Silicon под Rosetta 2 | Процессор | ✅ Поддерживается с резервными вариантами |\n| macOS | Интел | Процессор | ✅ Полностью поддерживается |\n| Окна | x86\\_64 | CUDA | ✅ Полностью поддерживается |\n| Окна | x86\\_64 | DirectML | ✅ Поддерживается |\n| Окна | x86\\_64 | Процессор | ✅ Поддерживается с резервными вариантами |\n| Линукс | x86\\_64 | CUDA | ✅ Полностью поддерживается |\n| Линукс | x86\\_64 | ROCм | ✅ Поддерживается |\n| Линукс | x86\\_64 | Процессор | ✅ Поддерживается с резервными вариантами |\n| Линукс | ARM64 | Процессор | ✅ Поддерживается с резервными вариантами |\n\n## Тестирование\n\n```bash\n# Install test dependencies\npip install pytest pytest-asyncio\n\n# Run all tests\npytest tests/\n\n# Run specific test categories\npytest tests/test_memory_ops.py\npytest tests/test_semantic_search.py\npytest tests/test_database.py\n\n# Verify environment compatibility\npython scripts/verify_environment_enhanced.py\n\n# Verify PyTorch installation on Windows\npython scripts/verify_pytorch_windows.py\n\n# Perform comprehensive installation verification\npython scripts/test_installation.py\n```\n\n## Поиск неисправностей\n\nПодробные инструкции по устранению неполадок см. [в руководстве по установке](docs/guides/installation.md#troubleshooting-common-installation-issues) .\n\n### Советы по быстрому устранению неполадок\n\n* **Ошибки Windows PyTorch** : используйте `python scripts/install_windows.py`\n* **Конфликты зависимостей Intel в macOS** : используйте `python install.py --force-compatible-deps`\n* **Ошибки рекурсии** : Запустите `python scripts/fix_sitecustomize.py`\n* **Проверка среды** : Запустите `python scripts/verify_environment_enhanced.py`\n* **Проблемы с памятью** : установите `MCP_MEMORY_BATCH_SIZE=4` и попробуйте меньшую модель.\n* **Apple Silicon** : убедитесь, что Python 3.10+ собран для ARM64, установите `PYTORCH_ENABLE_MPS_FALLBACK=1`\n* **Тестирование установки** : Запустите `python scripts/test_installation.py`\n\n## Структура проекта\n\n```\nmcp-memory-service/\n├── src/mcp_memory_service/ # Core package code\n│ ├── __init__.py\n│ ├── config.py # Configuration utilities\n│ ├── models/ # Data models\n│ ├── storage/ # Storage implementations\n│ ├── utils/ # Utility functions\n│ └── server.py # Main MCP server\n├── scripts/ # Helper scripts\n├── memory_wrapper.py # Windows wrapper script\n├── install.py # Enhanced installation script\n└── tests/ # Test suite\n```\n\n## Руководство по разработке\n\n* Python 3.10+ с подсказками типов\n* Используйте классы данных для моделей\n* Строки документации в тройных кавычках для модулей и функций\n* Шаблон async/await для всех операций ввода-вывода\n* Следуйте рекомендациям по стилю PEP 8\n* Включить тесты для новых функций\n\n## Лицензия\n\nЛицензия MIT — подробности см. в файле LICENSE\n\n## Благодарности\n\n* Команда ChromaDB для векторной базы данных\n* Проект Sentence Transformers для внедрения моделей\n* Проект MCP для спецификации протокола\n\n## Контакт\n\n[Телеграмма](t.me/doobeedoo)\n\n## Интеграции\n\nСлужба памяти MCP может быть расширена различными инструментами и утилитами. См. раздел [Интеграции](docs/integrations.md) для списка доступных опций, включая:\n\n* [MCP Memory Dashboard](https://github.com/doobidoo/mcp-memory-dashboard) — веб-интерфейс для просмотра и управления памятью\n* [Контекст памяти Клода](https://github.com/doobidoo/claude-memory-context) — внедрение контекста памяти в инструкции проекта Клода","Russian","ru-RU",{"_29":76,"_33":80,"_38":81},{"_31":77,"_33":78,"_35":79},"为 Claude 提供语义记忆和持久存储，利用 ChromaDB 和句子转换器增强搜索和检索功能。","MCP内存服务","# MCP内存服务\n\n[![许可证：MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT) [![铁匠徽章](https://smithery.ai/badge/@doobidoo/mcp-memory-service)](https://smithery.ai/server/@doobidoo/mcp-memory-service)\n\n一个 MCP 服务器，使用 ChromaDB 和句子转换器为 Claude Desktop 提供语义记忆和持久存储功能。该服务支持长期记忆存储和语义搜索功能，非常适合跨对话和实例维护上下文。\n\n## 帮助\n\n使用[TalkToGitHub](https://talktogithub.com/doobidoo/mcp-memory-service)与 Repo 对话！\n\n## 特征\n\n* 使用句子转换器进行语义搜索\n* **基于自然语言时间的回忆**（例如，“上周”，“昨天早上”）\n* 基于标签的记忆检索系统\n* 使用 ChromaDB 进行持久存储\n* 自动数据库备份\n* 内存优化工具\n* 精确匹配检索\n* 相似性分析的调试模式\n* 数据库健康监控\n* 重复检测和清理\n* 可定制的嵌入模型\n* **跨平台兼容性**（Apple Silicon、Intel、Windows、Linux）\n* 针对不同环境的**硬件感知优化**\n* 有限硬件资源的**优雅回退**\n\n## 安装\n\n### 快速启动（推荐）\n\n增强的安装脚本会自动检测您的系统并安装适当的依赖项：\n\n```bash\n# Clone the repository\ngit clone https://github.com/doobidoo/mcp-memory-service.git\ncd mcp-memory-service\n\n# Create and activate a virtual environment\npython -m venv venv\nsource venv/bin/activate # On Windows: venv\\Scripts\\activate\n\n# Run the installation script\npython install.py\n```\n\n`install.py`脚本将：\n\n1. 检测您的系统架构和可用的硬件加速器\n2. 为您的平台安装适当的依赖项\n3. 为您的环境配置最佳设置\n4. 验证安装并在需要时提供诊断\n\n### Docker 安装\n\n您可以使用 Docker 运行内存服务：\n\n```bash\n# Using Docker Compose (recommended)\ndocker-compose up\n\n# Using Docker directly\ndocker build -t mcp-memory-service .\ndocker run -p 8000:8000 -v /path/to/data:/app/chroma_db -v /path/to/backups:/app/backups mcp-memory-service\n```\n\n我们针对不同的场景提供了多种 Docker Compose 配置：\n\n* `docker-compose.yml` - 使用 pip install 的标准配置\n* `docker-compose.uv.yml` - 使用 UV 包管理器的替代配置\n* `docker-compose.pythonpath.yml` - 具有显式 PYTHONPATH 设置的配置\n\n要使用替代配置：\n\n```bash\ndocker-compose -f docker-compose.uv.yml up\n```\n\n### Windows 安装（特殊情况）\n\n由于特定平台的 Wheel 可用性，Windows 用户可能会遇到 PyTorch 安装问题。请使用我们 Windows 专用的安装脚本：\n\n```bash\n# After activating your virtual environment\npython scripts/install_windows.py\n```\n\n该脚本处理：\n\n1. 检测 CUDA 可用性和版本\n2. 从正确的索引 URL 安装适当的 PyTorch 版本\n3. 安装其他依赖项而不与 PyTorch 冲突\n4. 验证安装\n\n### 通过 Smithery 安装\n\n要通过[Smithery](https://smithery.ai/server/@doobidoo/mcp-memory-service)自动为 Claude Desktop 安装内存服务：\n\n```bash\nnpx -y @smithery/cli install @doobidoo/mcp-memory-service --client claude\n```\n\n### 详细安装指南\n\n有关全面的安装说明和故障排除，请参阅[安装指南](docs/guides/installation.md)。\n\n## Claude MCP 配置\n\n### 标准配置\n\n将以下内容添加到您的`claude_desktop_config.json`文件中：\n\n```json\n{\n \"memory\": {\n \"command\": \"uv\",\n \"args\": [\n \"--directory\",\n \"your_mcp_memory_service_directory\", // e.g., \"C:\\\\REPOSITORIES\\\\mcp-memory-service\"\n \"run\",\n \"memory\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"your_chroma_db_path\", // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\"\n \"MCP_MEMORY_BACKUPS_PATH\": \"your_backups_path\" // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n### Windows 特定配置（推荐）\n\n对于 Windows 用户，我们建议使用包装器脚本来确保 PyTorch 正确安装：\n\n```json\n{\n \"memory\": {\n \"command\": \"python\",\n \"args\": [\n \"C:\\\\path\\\\to\\\\mcp-memory-service\\\\memory_wrapper.py\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\",\n \"MCP_MEMORY_BACKUPS_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n包装脚本将：\n\n1. 检查 PyTorch 是否已安装并正确配置\n2. 如果需要，使用正确的索引 URL 安装 PyTorch\n3. 使用适当的配置运行内存服务器\n\n## 使用指南\n\n有关如何与 Claude Desktop 中的内存服务交互的详细说明：\n\n* [调用指南](docs/guides/invocation_guide.md)——了解在 Claude 中触发内存操作的特定关键字和短语\n* [安装指南](docs/guides/installation.md)- 详细设置说明\n\n在您与 Claude 的对话中，记忆服务会通过自然语言命令调用。例如：\n\n* 对商店说：“请记住，我的项目截止日期是 5 月 15 日。”\n* 检索：“你还记得我告诉你的项目截止日期吗？”\n* 删除：“请忘记我告诉你的我的地址。”\n\n请参阅[调用指南](docs/guides/invocation_guide.md)以获取完整的命令列表和详细的使用示例。\n\n## 内存操作\n\n内存服务通过MCP服务器提供以下操作：\n\n### 核心内存操作\n\n1. `store_memory` - 使用可选标签存储新信息\n2. `retrieve_memory` - 对相关记忆进行语义搜索\n3. `recall_memory` - 使用自然语言时间表达检索记忆\n4. `search_by_tag` - 使用特定标签查找记忆\n5. `exact_match_retrieve` - 查找内容完全匹配的记忆\n6. `debug_retrieve` - 检索具有相似度分数的记忆\n\n### 数据库管理\n\n7. `create_backup` - 创建数据库备份\n8. `get_stats`获取内存统计信息\n9. `optimize_db`优化数据库性能\n10. `check_database_health` - 获取数据库健康指标\n11. `check_embedding_model` - 验证模型状态\n\n### 内存管理\n\n12. `delete_memory` - 通过哈希删除特定内存\n13. `delete_by_tag` - 删除所有带有特定标签的记忆\n14. `cleanup_duplicates` - 删除重复条目\n\n## 配置选项\n\n通过环境变量配置：\n\n```\nCHROMA_DB_PATH: Path to ChromaDB storage\nBACKUP_PATH: Path for backups\nAUTO_BACKUP_INTERVAL: Backup interval in hours (default: 24)\nMAX_MEMORIES_BEFORE_OPTIMIZE: Threshold for auto-optimization (default: 10000)\nSIMILARITY_THRESHOLD: Default similarity threshold (default: 0.7)\nMAX_RESULTS_PER_QUERY: Maximum results per query (default: 10)\nBACKUP_RETENTION_DAYS: Number of days to keep backups (default: 7)\nLOG_LEVEL: Logging level (default: INFO)\n\n# Hardware-specific environment variables\nPYTORCH_ENABLE_MPS_FALLBACK: Enable MPS fallback for Apple Silicon (default: 1)\nMCP_MEMORY_USE_ONNX: Use ONNX Runtime for CPU-only deployments (default: 0)\nMCP_MEMORY_USE_DIRECTML: Use DirectML for Windows acceleration (default: 0)\nMCP_MEMORY_MODEL_NAME: Override the default embedding model\nMCP_MEMORY_BATCH_SIZE: Override the default batch size\n```\n\n## 硬件兼容性\n\n| 平台 | 建筑学 | 加速器 | 地位 |\n| ----- | -------------------------- | -------- | ------ |\n| macOS | 苹果硅片 (M1/M2/M3) | 国会议员 | ✅ 完全支持 |\n| macOS | Rosetta 2 下的 Apple Silicon | 中央处理器 | ✅ 支持后备 |\n| macOS | 英特尔 | 中央处理器 | ✅ 完全支持 |\n| 视窗 | x86\\_64 | CUDA | ✅ 完全支持 |\n| 视窗 | x86\\_64 | DirectML | ✅ 支持 |\n| 视窗 | x86\\_64 | 中央处理器 | ✅ 支持后备 |\n| Linux | x86\\_64 | CUDA | ✅ 完全支持 |\n| Linux | x86\\_64 | ROC | ✅ 支持 |\n| Linux | x86\\_64 | 中央处理器 | ✅ 支持后备 |\n| Linux | ARM64 | 中央处理器 | ✅ 支持后备 |\n\n## 测试\n\n```bash\n# Install test dependencies\npip install pytest pytest-asyncio\n\n# Run all tests\npytest tests/\n\n# Run specific test categories\npytest tests/test_memory_ops.py\npytest tests/test_semantic_search.py\npytest tests/test_database.py\n\n# Verify environment compatibility\npython scripts/verify_environment_enhanced.py\n\n# Verify PyTorch installation on Windows\npython scripts/verify_pytorch_windows.py\n\n# Perform comprehensive installation verification\npython scripts/test_installation.py\n```\n\n## 故障排除\n\n有关详细的故障排除步骤，请参阅[安装指南](docs/guides/installation.md#troubleshooting-common-installation-issues)。\n\n### 快速故障排除提示\n\n* **Windows PyTorch 错误**：使用`python scripts/install_windows.py`\n* **macOS Intel 依赖冲突**：使用`python install.py --force-compatible-deps`\n* **递归错误**：运行`python scripts/fix_sitecustomize.py`\n* **环境验证**：运行`python scripts/verify_environment_enhanced.py`\n* **内存问题**：设置`MCP_MEMORY_BATCH_SIZE=4`并尝试较小的模型\n* **Apple Silicon** ：确保为 ARM64 构建 Python 3.10+，设置`PYTORCH_ENABLE_MPS_FALLBACK=1`\n* **安装测试**：运行`python scripts/test_installation.py`\n\n## 项目结构\n\n```\nmcp-memory-service/\n├── src/mcp_memory_service/ # Core package code\n│ ├── __init__.py\n│ ├── config.py # Configuration utilities\n│ ├── models/ # Data models\n│ ├── storage/ # Storage implementations\n│ ├── utils/ # Utility functions\n│ └── server.py # Main MCP server\n├── scripts/ # Helper scripts\n├── memory_wrapper.py # Windows wrapper script\n├── install.py # Enhanced installation script\n└── tests/ # Test suite\n```\n\n## 开发指南\n\n* 带有类型提示的 Python 3.10+\n* 使用数据类作为模型\n* 模块和函数的三重引号文档字符串\n* 所有 I/O 操作的异步/等待模式\n* 遵循 PEP 8 样式指南\n* 包括新功能的测试\n\n## 执照\n\nMIT 许可证 - 详情请参阅许可证文件\n\n## 致谢\n\n* ChromaDB矢量数据库团队\n* 用于嵌入模型的句子转换器项目\n* 协议规范的 MCP 项目\n\n## 接触\n\n[电报](t.me/doobeedoo)\n\n## 集成\n\nMCP 内存服务可以通过各种工具和实用程序进行扩展。请参阅[集成](docs/integrations.md)以获取可用选项列表，包括：\n\n* [MCP 内存仪表板](https://github.com/doobidoo/mcp-memory-dashboard)- 用于浏览和管理内存的 Web UI\n* [Claude Memory Context](https://github.com/doobidoo/claude-memory-context) - 将内存上下文注入 Claude 项目说明","Chinese","zh-CN","githubUser","locale",{"_85":86,"_87":88,"_13":132,"_140":42,"_141":42,"_142":143,"_147":148,"_171":-5,"_16":172,"_173":174,"_175":176,"_33":43,"_187":161,"_188":-5,"_189":190,"_191":192,"_197":-5,"_183":157,"_198":199,"_202":203,"_162":204,"_205":206},"addedAt","2024-12-27T00:00:12.795795158Z","argumentsJsonSchema",{"_89":90,"_129":130,"_96":131},"properties",{"_91":92,"_98":99,"_102":103,"_106":107,"_110":111,"_114":115,"_118":119,"_122":123,"_126":127},"LOG_LEVEL",{"_93":94,"_31":95,"_96":97},"default","INFO","Logging level (DEBUG, INFO, WARNING, ERROR)","type","string","PYTHONPATH",{"_93":100,"_31":101,"_96":97},"/app","Python module search path","PYTHONUNBUFFERED",{"_93":104,"_31":105,"_96":97},"1","Disable Python output buffering","MCP_MEMORY_FORCE_CPU",{"_93":108,"_31":109,"_96":97},"0","Force CPU-only mode (0 or 1)","SIMILARITY_THRESHOLD",{"_93":112,"_31":113,"_96":97},"0.7","Threshold for similarity search (0.0 to 1.0)","MAX_RESULTS_PER_QUERY",{"_93":116,"_31":117,"_96":97},"10","Maximum number of results returned per query","MCP_MEMORY_CHROMA_PATH",{"_93":120,"_31":121,"_96":97},"/app/chroma_db","Path to ChromaDB storage directory","MCP_MEMORY_BACKUPS_PATH",{"_93":124,"_31":125,"_96":97},"/app/backups","Path to backups storage directory","PYTORCH_ENABLE_MPS_FALLBACK",{"_93":104,"_31":128,"_96":97},"Enable MPS fallback for Apple Silicon (0 or 1)","required",[],"object",[133,134,135,136,137,138,139],"category:knowledge-and-memory","category:vector-databases","category:rag-systems","hosting:hybrid","language:python","license_grade:a","author:claimed","descriptionMarkdown","descriptionPlainText","githubProject",{"_35":144,"_145":146},"# MCP Memory Service\n\n[![License: Apache 2.0](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)\n[![smithery badge](https://smithery.ai/badge/@doobidoo/mcp-memory-service)](https://smithery.ai/server/@doobidoo/mcp-memory-service)\n[![Verified on MseeP](https://mseep.ai/badge.svg)](https://mseep.ai/app/0513fb92-e941-4fe0-9948-2a1dbb870dcf)\n\nAn intelligent MCP server providing semantic memory, persistent storage, and **autonomous memory consolidation** for Claude Desktop. This service combines ChromaDB/SQLite-vec storage with a revolutionary **dream-inspired consolidation system** that automatically organizes, compresses, and manages memories over time, creating a self-evolving knowledge base.\n\n $\"grafik\"$ \n $\"Memory$ \n\n## Help\n- Talk to the Repo with [TalkToGitHub](https://talktogithub.com/doobidoo/mcp-memory-service)!\n- Use Gitprobe to digg deeper: [GitProbe](https://gitprobe.com/doobidoo/mcp-memory-service)!\n\n---\n\n## 🎯 NEW: Claude Code Commands (v2.2.0)\n\n**Get started in 2 minutes with direct memory commands!**\n\n```bash\n# Install with Claude Code commands\npython install.py --install-claude-commands\n\n# Start using immediately\nclaude /memory-store \"Important decision about architecture\"\nclaude /memory-recall \"what did we decide last week?\"\nclaude /memory-search --tags \"architecture,database\"\nclaude /memory-health\n```\n\n✨ **5 conversational commands** following CCPlugins pattern \n🚀 **Zero MCP server configuration** required \n🧠 **Context-aware operations** with automatic project detection \n🎨 **Professional interface** with comprehensive guidance \n\n➡️ [**Quick Start Guide**](docs/guides/claude-code-quickstart.md) | [**Full Integration Guide**](docs/guides/claude-code-integration.md)\n\n## 🚀 NEW: Remote MCP Memory Service (v4.0.0)\n\n**Production-ready remote memory service with native MCP-over-HTTP protocol!**\n\n### Remote Deployment\n\nDeploy the memory service on any server for cross-device access:\n\n```bash\n# On your server\ngit clone https://github.com/doobidoo/mcp-memory-service.git\ncd mcp-memory-service\npython install.py\npython scripts/run_http_server.py\n```\n\n**Server Access Points:**\n- **MCP Protocol**: `http://your-server:8000/mcp` (for MCP clients)\n- **Dashboard**: `http://your-server:8000/` (web interface)\n- **API Docs**: `http://your-server:8000/api/docs` (interactive API)\n\n### Claude Code Integration\n\nConnect Claude Code to your remote memory service:\n\n```bash\n# Test MCP connection\ncurl -X POST http://your-server:8000/mcp \\\n -H \"Content-Type: application/json\" \\\n -d '{\n \"jsonrpc\": \"2.0\",\n \"id\": 1,\n \"method\": \"tools/list\"\n }'\n\n# Store memories remotely\ncurl -X POST http://your-server:8000/mcp \\\n -H \"Content-Type: application/json\" \\\n -d '{\n \"jsonrpc\": \"2.0\", \n \"id\": 1,\n \"method\": \"tools/call\",\n \"params\": {\n \"name\": \"store_memory\",\n \"arguments\": {\n \"content\": \"Your memory content\",\n \"tags\": [\"tag1\", \"tag2\"]\n }\n }\n }'\n```\n\n**Key Benefits:**\n- ✅ **Cross-Device Access**: Connect from any device running Claude Code\n- ✅ **Native MCP Protocol**: Standard JSON-RPC 2.0 implementation \n- ✅ **No Bridge Required**: Direct HTTP/HTTPS connection\n- ✅ **Production Ready**: Proven deployment at scale\n\n### MCP Client Compatibility\n\n| Client | Status | Notes |\n|--------|--------|-------|\n| ✅ Standard MCP Libraries | Working | Python `mcp`, JavaScript SDK |\n| ✅ Claude Desktop | Working | With proper MCP configuration |\n| ⚠️ Claude Code | Limited | Use Claude Commands instead ([compatibility guide](CLAUDE_CODE_COMPATIBILITY.md)) |\n| ✅ Custom MCP Clients | Working | Full protocol compliance |\n| ✅ HTTP API | Working | REST interface on port 8080 |\n\n➡️ [**Dual-Service Guide**](DUAL_SERVICE_DEPLOYMENT.md) | [**Claude Code Compatibility**](CLAUDE_CODE_COMPATIBILITY.md) | [**Migration Log**](MIGRATION_LOG.md)\n\n---\n\n## Features\n\n### 🎯 MCP Protocol Enhancements (NEW in v4.1.0!)\n\n#### 📚 **Enhanced Resources**\nAccess structured memory data through URI-based resources:\n- `memory://stats` - Database statistics and health metrics\n- `memory://tags` - All available memory tags\n- `memory://recent/{n}` - N most recent memories\n- `memory://tag/{tagname}` - Memories with specific tag\n- `memory://search/{query}` - Dynamic search results\n\n#### 📋 **Guided Prompts**\nInteractive workflows for common memory operations:\n- **memory_review** - Review and organize memories from time periods\n- **memory_analysis** - Analyze patterns and themes in memories\n- **knowledge_export** - Export memories in JSON/Markdown/Text formats\n- **memory_cleanup** - Identify and remove duplicates/outdated memories\n- **learning_session** - Store structured learning notes with automatic tagging\n\n#### 📊 **Progress Tracking**\nReal-time progress notifications for long-running operations:\n- Bulk deletion operations with percentage updates\n- Database optimization with step-by-step progress\n- Operation IDs for tracking multiple concurrent tasks\n- MCP-compliant progress notification protocol\n\n#### 🔄 **Database Synchronization** (NEW in v4.5.0!)\nMulti-node database synchronization for distributed memory access:\n- **JSON Export/Import**: Preserve timestamps and metadata across database migrations\n- **Litestream Integration**: Real-time database replication with conflict resolution\n- **3-Node Architecture**: Central server with replica nodes for distributed workflows\n- **Deduplication Logic**: Content hash-based duplicate prevention during imports\n- **Source Tracking**: Automatic tagging to identify memory origin machines\n- **Cross-Platform Sync**: Synchronize memories between Windows, macOS, and Linux systems\n\n#### 🎛️ **Multi-Client Optimization** (NEW in v4.5.1!)\nThe MCP Memory Service automatically detects your MCP client and optimizes its output:\n- **Claude Desktop**: Clean JSON-only communication for maximum compatibility\n - Suppresses diagnostic output to maintain strict JSON-RPC protocol\n - Routes only WARNING/ERROR messages to stderr\n - Ensures seamless integration with Claude's parsing requirements\n- **LM Studio**: Enhanced diagnostic output for easier troubleshooting \n - Shows system diagnostics, dependency checks, and initialization status\n - Provides detailed feedback for development and debugging\n - Maintains full INFO/DEBUG output for comprehensive monitoring\n- **Automatic Detection**: Uses process inspection and environment variables\n- **Manual Override**: Set `CLAUDE_DESKTOP=1` or `LM_STUDIO=1` for explicit control\n- **Fallback Safety**: Defaults to strict JSON mode for unknown clients\n\n### 🧠 Dream-Inspired Memory Consolidation (NEW in v2.0!)\n- **Autonomous memory management** inspired by human sleep cycle processing\n- **Multi-layered time horizons** (daily → weekly → monthly → quarterly → yearly)\n- **Creative association discovery** finding non-obvious connections between memories\n- **Semantic clustering** automatically organizing related memories\n- **Intelligent compression** preserving key information while reducing storage\n- **Controlled forgetting** with safe archival and recovery systems\n- **Performance optimized** for processing 10k+ memories efficiently\n- **Health monitoring** with comprehensive error handling and alerts\n\n### 🔍 Core Memory Operations\n- Semantic search using sentence transformers\n- **Natural language time-based recall** (e.g., \"last week\", \"yesterday morning\")\n- **Enhanced tag deletion system** with flexible multi-tag support\n- Tag-based memory retrieval system\n- Exact match retrieval\n- Debug mode for similarity analysis\n- Duplicate detection and cleanup\n\n### 🗄️ Storage & Performance\n- **Dual storage backends**: ChromaDB (full-featured) and SQLite-vec (lightweight, fast)\n- Automatic database backups\n- Memory optimization tools\n- Database health monitoring\n- Customizable embedding model\n- **Cross-platform compatibility** (Apple Silicon, Intel, Windows, Linux)\n- **Hardware-aware optimizations** for different environments\n- **Graceful fallbacks** for limited hardware resources\n\n### 🔗 Integration & Coordination\n- **🆕 Modern Dashboard UI (v3.3.0)** - Professional web interface with live stats and interactive endpoint documentation\n- **🆕 Claude Code Commands (v2.2.0)** - Conversational memory commands following CCPlugins pattern\n- **🆕 Multi-client coordination** for Claude Desktop + Claude Code concurrent access\n- **🆕 Intelligent coordination modes** with automatic WAL/HTTP detection\n- **🆕 mDNS Service Discovery (v2.1.0)** - Zero-configuration networking with automatic service discovery\n- **🆕 HTTPS Support** with auto-generated certificates for secure connections\n- **7 new MCP tools** for consolidation operations\n- Environment variable-based configuration\n\n### Recent Enhancements\n\n#### v3.3.0 - Modern Dashboard UI\n- ✅ **Professional Web Interface**: Modern gradient design with card-based layout\n- ✅ **Live Statistics**: Real-time memory count, model info, server status, and response times\n- ✅ **Interactive API Documentation**: Organized endpoint cards with direct links to API docs\n- ✅ **Tech Stack Display**: Visual representation of underlying technologies\n- ✅ **Mobile Responsive**: Optimized for desktop and mobile devices\n- ✅ **Auto-Refresh**: Live stats update every 30 seconds automatically\n\n\n*Dashboard screenshot coming soon - shows modern gradient design with live stats, interactive endpoint cards, and tech stack badges*\n\n**Access the Dashboard:**\n- Local: `http://localhost:8000`\n- mDNS: `http://mcp-memory-service.local:8000`\n- API Docs: `http://localhost:8000/api/docs`\n\n#### v3.2.0 - SQLite-vec Embedding Fixes & Diagnostics\n- ✅ **Zero Vector Repair**: Comprehensive diagnostic and repair tools for corrupted embeddings\n- ✅ **Enhanced Error Handling**: Robust validation and initialization for SQLite-vec backend\n- ✅ **Migration Tools**: Safe migration utilities that preserve existing memories\n- ✅ **Dependency Management**: Moved core ML dependencies to main requirements for reliability\n- ✅ **Database Diagnostics**: Advanced tools for analyzing and fixing embedding issues\n- ✅ **HTTP API Improvements**: Fixed search endpoint compatibility and error handling\n\n#### v3.1.0 - Cross-Platform Service Installation\n- ✅ **Native Service Support**: Install as system service on Windows, macOS, and Linux\n- ✅ **Auto-Startup Configuration**: Automatic boot/login startup with service management\n- ✅ **mDNS Port Flexibility**: Clean access via port 443 without Pi-hole conflicts\n- ✅ **Service Management Commands**: Start, stop, status, and uninstall operations\n\n#### v3.0.0 - Autonomous Multi-Client Memory Service (MAJOR RELEASE)\n- 🧠 **Dream-Inspired Consolidation**: Autonomous memory processing with exponential decay and creative association discovery\n- 🌐 **Multi-Client Architecture**: Production FastAPI HTTPS server with automatic SSL certificates\n- 🔍 **mDNS Service Discovery**: Zero-configuration networking with `_mcp-memory._tcp.local.` advertisement\n- 📡 **Server-Sent Events**: Real-time updates with 30-second heartbeat for live synchronization \n- 🚀 **Production Deployment**: Complete systemd service integration with professional management scripts\n- 🔒 **Security**: API key authentication and user-space execution for enhanced security\n- 📖 **Documentation Overhaul**: Comprehensive production setup and service lifecycle guides\n- ⚡ **Performance**: Mathematical consolidation using existing embeddings (no external AI dependencies)\n\n#### v2.2.0 - Claude Code Commands Integration\n- **5 conversational commands** for direct memory operations: `/memory-store`, `/memory-recall`, `/memory-search`, `/memory-context`, `/memory-health`\n- **Optional installation** integrated into main installer with intelligent prompting\n- **CCPlugins-compatible** markdown-based conversational command format\n- **Context-aware operations** with automatic project and session detection\n- **Cross-platform support** with comprehensive error handling and fallback systems\n\n#### v2.1.0 - Zero-Configuration Networking\n- ✅ **mDNS Service Discovery**: Automatic service advertisement and discovery using `_mcp-memory._tcp.local.`\n- ✅ **HTTPS Integration**: SSL/TLS support with automatic self-signed certificate generation\n- ✅ **Enhanced HTTP-MCP Bridge**: Auto-discovery mode with health validation and fallback\n- ✅ **Zero-Config Deployment**: No manual endpoint configuration needed for local networks\n\n#### Previous Enhancements\n- ✅ **PyTorch Optional**: Now works without PyTorch for basic functionality when using SQLite-vec backend\n- ✅ **Improved SQLite-vec**: Robust error handling and validation for the lightweight backend\n- ✅ **Intelligent Health Checks**: Backend-specific health monitoring with detailed diagnostics\n- ✅ **Comprehensive Testing**: Added test scripts for all critical functions\n- ✅ **API Consistency**: Enhanced `delete_by_tag` to support both single and multiple tags\n- ✅ **New Delete Methods**: Added `delete_by_tags` (OR logic) and `delete_by_all_tags` (AND logic)\n- ✅ **Backward Compatibility**: All existing code continues to work unchanged\n- ✅ **Dashboard Integration**: Enhanced UI with multiple tag selection capabilities\n\n## Installation Methods\n\n[![Docker](https://img.shields.io/badge/Docker-Fastest_Setup-008fe2?style=for-the-badge&logo=docker&logoColor=white)](#docker-installation)\n[![Smithery](https://img.shields.io/badge/Smithery-Auto_Install-9f7aea?style=for-the-badge&logo=npm&logoColor=white)](#installing-via-smithery)\n[![Python](https://img.shields.io/badge/Python-Intelligent_Installer-ffd343?style=for-the-badge&logo=python&logoColor=black)](#-intelligent-installer-recommended)\n[![uvx](https://img.shields.io/badge/uvx-Isolated_Install-00d2d3?style=for-the-badge&logo=python&logoColor=white)](#uvx-installation)\n\n### 🚀 Quick Start Options\n\n| Method | Best For | Setup Time | Features |\n|--------|----------|------------|----------|\n| **Docker** | Production, Multi-platform | 2 minutes | ✅ Isolated, ✅ Multi-client ready |\n| **Smithery** | Claude Desktop users | 1 minute | ✅ Auto-config, ✅ One command |\n| **Python Installer** | Developers, Customization | 5 minutes | ✅ Hardware detection, ✅ Full control |\n| **uvx** | Temporary use, Testing | 3 minutes | ✅ No virtual env, ✅ Clean install |\n\n### 🚀 Intelligent Installer (Recommended)\n\nThe new unified installer automatically detects your hardware and selects the optimal configuration:\n\n```bash\n# Clone the repository\ngit clone https://github.com/doobidoo/mcp-memory-service.git\ncd mcp-memory-service\n\n# Create and activate a virtual environment\npython -m venv venv\nsource venv/bin/activate # On Windows: venv\\Scripts\\activate\n\n# Run the intelligent installer\npython install.py\n\n# ✨ NEW: Multi-client setup is now integrated!\n# You'll be prompted to configure universal MCP client access\n# for Claude Desktop, VS Code, Continue, and other MCP applications\n```\n\n### 🎯 Hardware-Specific Installation\n\n**For Intel Macs:**\nFor detailed setup instructions specific to Intel Macs, see our [Intel Mac Setup Guide](docs/platforms/macos-intel.md).\n\n**For Legacy Hardware (2013-2017 Intel Macs):**\n```bash\npython install.py --legacy-hardware\n```\n\n**For Server/Headless Deployment:**\n```bash\npython install.py --server-mode\n```\n\n**For HTTP/SSE API Development:**\n```bash\npython install.py --enable-http-api\n```\n\n**For Migration from ChromaDB:**\n```bash\npython install.py --migrate-from-chromadb\n```\n\n**For Multi-Client Setup:**\n```bash\n# Automatic multi-client setup during installation\npython install.py --setup-multi-client\n\n# Skip the interactive multi-client prompt\npython install.py --skip-multi-client-prompt\n```\n\n**For Claude Code Commands:**\n```bash\n# Install with Claude Code commands (prompts if CLI detected)\npython install.py --install-claude-commands\n\n# Skip the interactive Claude Code commands prompt\npython install.py --skip-claude-commands-prompt\n```\n\n### 🧠 What the Installer Does\n\n1. **Hardware Detection**: CPU, GPU, memory, and platform analysis\n2. **Intelligent Backend Selection**: ChromaDB vs SQLite-vec based on your hardware\n3. **Platform Optimization**: macOS Intel fixes, Windows CUDA setup, Linux variations\n4. **Dependency Management**: Compatible PyTorch and ML library versions\n5. **Auto-Configuration**: Claude Desktop config and environment variables\n6. **Migration Support**: Seamless ChromaDB to SQLite-vec migration\n\n### 📊 Storage Backend Selection\n\nMCP Memory Service supports two optimized storage backends:\n\n#### SQLite-vec 🪶 (Lightweight & Fast)\n**Best for**: 2015 MacBook Pro, older Intel Macs, low-memory systems, Docker deployments\n\n- ✅ **10x faster startup** (2-3 seconds vs 15-30 seconds)\n- ✅ **Single file database** (easy backup/sharing)\n- ✅ **Minimal memory usage** (~150MB vs ~600MB)\n- ✅ **No external dependencies**\n- ✅ **HTTP/SSE API support**\n\n#### ChromaDB 📦 (Full-Featured)\n**Best for**: Modern Macs (M1/M2/M3), GPU-enabled systems, production deployments\n\n- ✅ **Advanced vector search** with multiple metrics\n- ✅ **Rich metadata support** and complex queries\n- ✅ **Battle-tested scalability**\n- ✅ **Extensive ecosystem** integration\n\n**The installer automatically recommends the best backend for your hardware**, but you can override with:\n```bash\npython install.py --storage-backend sqlite_vec # Lightweight\npython install.py --storage-backend chromadb # Full-featured\n```\n\n### Docker Installation\n\n#### Docker Hub (Recommended)\n\nThe easiest way to run the Memory Service is using our pre-built Docker images:\n\n```bash\n# Pull the latest image\ndocker pull doobidoo/mcp-memory-service:latest\n\n# Run with default settings (for MCP clients like Claude Desktop)\ndocker run -d -p 8000:8000 \\\n -v $(pwd)/data/chroma_db:/app/chroma_db \\\n -v $(pwd)/data/backups:/app/backups \\\n doobidoo/mcp-memory-service:latest\n\n# Run in standalone mode (for testing/development)\ndocker run -d -p 8000:8000 \\\n -e MCP_STANDALONE_MODE=1 \\\n -v $(pwd)/data/chroma_db:/app/chroma_db \\\n -v $(pwd)/data/backups:/app/backups \\\n doobidoo/mcp-memory-service:latest\n```\n\n#### Docker Compose\n\nWe provide multiple Docker Compose configurations for different scenarios:\n- `docker-compose.yml` - Standard configuration for MCP clients (Claude Desktop)\n- `docker-compose.standalone.yml` - **Standalone mode** for testing/development (prevents boot loops)\n- `docker-compose.uv.yml` - Alternative configuration using UV package manager\n- `docker-compose.pythonpath.yml` - Configuration with explicit PYTHONPATH settings\n\n```bash\n# Using Docker Compose (recommended)\ndocker-compose up\n\n# Standalone mode (prevents boot loops)\ndocker-compose -f docker-compose.standalone.yml up\n```\n\n#### Building from Source\n\nIf you need to build the Docker image yourself:\n\n```bash\n# Build the image\ndocker build -t mcp-memory-service .\n\n# Run the container\ndocker run -p 8000:8000 \\\n -v $(pwd)/data/chroma_db:/app/chroma_db \\\n -v $(pwd)/data/backups:/app/backups \\\n mcp-memory-service\n```\n\n### uvx Installation\n\nYou can install and run the Memory Service using uvx for isolated execution:\n\n```bash\n# Install uv (which includes uvx) if not already installed\npip install uv\n# Or use the installer script:\n# curl -LsSf https://astral.sh/uv/install.sh | sh\n\n# Install and run the memory service\nuvx mcp-memory-service\n\n# Or install from GitHub\nuvx --from git+https://github.com/doobidoo/mcp-memory-service.git mcp-memory-service\n```\n\n### Windows Installation (Special Case)\n\nWindows users may encounter PyTorch installation issues due to platform-specific wheel availability. Use our Windows-specific installation script:\n\n```bash\n# After activating your virtual environment\npython scripts/install_windows.py\n```\n\nThis script handles:\n1. Detecting CUDA availability and version\n2. Installing the appropriate PyTorch version from the correct index URL\n3. Installing other dependencies without conflicting with PyTorch\n4. Verifying the installation\n\n### Installing via Smithery\n\nTo install Memory Service for Claude Desktop automatically via [Smithery](https://smithery.ai/server/@doobidoo/mcp-memory-service):\n\n```bash\nnpx -y @smithery/cli install @doobidoo/mcp-memory-service --client claude\n```\n\n### Detailed Installation Guide\n\nFor comprehensive installation instructions and troubleshooting, see the [Installation Guide](docs/installation/master-guide.md).\n\n## Claude MCP Configuration\n\n### Standard Configuration\n\nAdd the following to your `claude_desktop_config.json` file:\n\n```json\n{\n \"memory\": {\n \"command\": \"uv\",\n \"args\": [\n \"--directory\",\n \"your_mcp_memory_service_directory\", // e.g., \"C:\\\\REPOSITORIES\\\\mcp-memory-service\"\n \"run\",\n \"memory\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"your_chroma_db_path\", // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\"\n \"MCP_MEMORY_BACKUPS_PATH\": \"your_backups_path\" // e.g., \"C:\\\\Users\\\\John.Doe\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n### Windows-Specific Configuration (Recommended)\n\nFor Windows users, we recommend using the wrapper script to ensure PyTorch is properly installed:\n\n```json\n{\n \"memory\": {\n \"command\": \"python\",\n \"args\": [\n \"C:\\\\path\\\\to\\\\mcp-memory-service\\\\memory_wrapper.py\"\n ],\n \"env\": {\n \"MCP_MEMORY_CHROMA_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\chroma_db\",\n \"MCP_MEMORY_BACKUPS_PATH\": \"C:\\\\Users\\\\YourUsername\\\\AppData\\\\Local\\\\mcp-memory\\\\backups\"\n }\n }\n}\n```\n\n### SQLite-vec Configuration (Lightweight)\n\nFor a lighter-weight configuration that doesn't require PyTorch:\n\n```json\n{\n \"memory\": {\n \"command\": \"python\",\n \"args\": [\"-m\", \"mcp_memory_service.server\"],\n \"cwd\": \"/path/to/mcp-memory-service\",\n \"env\": {\n \"MCP_MEMORY_STORAGE_BACKEND\": \"sqlite_vec\",\n \"MCP_MEMORY_SQLITE_PATH\": \"/path/to/mcp-memory/sqlite_vec.db\",\n \"MCP_MEMORY_BACKUPS_PATH\": \"/path/to/mcp-memory/backups\",\n \"MCP_MEMORY_USE_ONNX\": \"1\",\n \"PYTHONPATH\": \"/path/to/mcp-memory-service\"\n }\n }\n}\n```\n\nThe wrapper script will:\n1. Check if PyTorch is installed and properly configured\n2. Install PyTorch with the correct index URL if needed\n3. Run the memory server with the appropriate configuration\n\n## 🌐 Multi-Client Deployment\n\n**NEW**: Deploy MCP Memory Service for multiple clients sharing the same memory database!\n\n### 🚀 Centralized Server Deployment (Recommended)\n\nPerfect for distributed teams, multiple devices, or cloud deployment:\n\n```bash\n# Install and start HTTP/SSE server\npython install.py --server-mode --enable-http-api\nexport MCP_HTTP_HOST=0.0.0.0 # Allow external connections\nexport MCP_API_KEY=\"your-secure-key\" # Optional authentication\npython scripts/run_http_server.py\n```\n\n**✅ Benefits:**\n- 🔄 **Real-time sync** across all clients via Server-Sent Events (SSE)\n- 🌍 **Cross-platform** - works from any device with HTTP access\n- 🔒 **Secure** with optional API key authentication\n- 📈 **Scalable** - handles many concurrent clients\n- ☁️ **Cloud-ready** - deploy on AWS, DigitalOcean, Docker, etc.\n\n**Access via:**\n- **API Docs**: `http://your-server:8000/api/docs`\n- **Web Dashboard**: `http://your-server:8000/`\n- **REST API**: All MCP operations available via HTTP\n\n### ⚠️ Why NOT Cloud Storage (Dropbox/OneDrive/Google Drive)\n\n**Direct SQLite on cloud storage DOES NOT WORK** for multi-client access:\n\n❌ **File locking conflicts** - Cloud sync breaks SQLite's locking mechanism \n❌ **Data corruption** - Incomplete syncs can corrupt the database \n❌ **Sync conflicts** - Multiple clients create \"conflicted copy\" files \n❌ **Performance issues** - Full database re-upload on every change \n\n**✅ Solution**: Use centralized HTTP server deployment instead!\n\n### 📖 Complete Documentation\n\nFor detailed deployment guides, configuration options, and troubleshooting:\n\n📚 **[Multi-Client Deployment Guide](docs/integration/multi-client.md)**\n\nCovers:\n- **Centralized HTTP/SSE Server** setup and configuration\n- **Shared File Access** for local networks (limited scenarios)\n- **Cloud Platform Deployment** (AWS, DigitalOcean, Docker)\n- **Security & Authentication** setup\n- **Performance Tuning** for high-load environments\n- **Troubleshooting** common multi-client issues\n\n## Usage Guide\n\nFor detailed instructions on how to interact with the memory service in Claude Desktop:\n\n- [Invocation Guide](docs/guides/invocation_guide.md) - Learn the specific keywords and phrases that trigger memory operations in Claude\n- [Installation Guide](docs/installation/master-guide.md) - Detailed setup instructions\n- **[Demo Session Walkthrough](docs/tutorials/demo-session-walkthrough.md)** - Real-world development session showcasing advanced features\n\nThe memory service is invoked through natural language commands in your conversations with Claude. For example:\n- To store: \"Please remember that my project deadline is May 15th.\"\n- To retrieve: \"Do you remember what I told you about my project deadline?\"\n\n### Claude Code Commands Usage\nWith the optional Claude Code commands installed, you can also use direct command syntax:\n```bash\n# Store information with context\nclaude /memory-store \"Important architectural decision about database backend\"\n\n# Recall memories by time\nclaude /memory-recall \"what did we decide about the database last week?\"\n\n# Search by tags or content\nclaude /memory-search --tags \"architecture,database\"\n\n# Capture current session context\nclaude /memory-context --summary \"Development planning session\"\n\n# Check service health\nclaude /memory-health\n```\n- To delete: \"Please forget what I told you about my address.\"\n\nSee the [Invocation Guide](docs/guides/invocation_guide.md) for a complete list of commands and detailed usage examples.\n\n## Storage Backends\n\nThe MCP Memory Service supports multiple storage backends to suit different use cases:\n\n### ChromaDB (Default)\n- **Best for**: Large memory collections (>100K entries), high-performance requirements\n- **Features**: Advanced vector indexing, excellent query performance, rich ecosystem\n- **Memory usage**: Higher (~200MB for 1K memories)\n- **Setup**: Automatically configured, no additional dependencies\n\n### SQLite-vec (Lightweight Alternative)\n- **Best for**: Smaller collections (<100K entries), resource-constrained environments\n- **Features**: Single-file database, 75% lower memory usage, better portability\n- **Memory usage**: Lower (~50MB for 1K memories)\n- **Setup**: Requires `sqlite-vec` package\n\n#### Quick Setup for SQLite-vec\n\n```bash\n# Install sqlite-vec (if using installation script, this is handled automatically)\npip install sqlite-vec\n\n# Configure the backend\nexport MCP_MEMORY_STORAGE_BACKEND=sqlite_vec\nexport MCP_MEMORY_SQLITE_PATH=/path/to/sqlite_vec.db\n\n# Optional: For CPU-only mode without PyTorch (much lighter resource usage)\nexport MCP_MEMORY_USE_ONNX=1\n\n# Restart Claude Desktop\n```\n\n#### SQLite-vec with Optional PyTorch\n\nThe SQLite-vec backend now works with or without PyTorch installed:\n\n- **With PyTorch**: Full functionality including embedding generation\n- **Without PyTorch**: Basic functionality using pre-computed embeddings and ONNX runtime\n- **With Homebrew PyTorch**: Integration with macOS Homebrew PyTorch installation\n \nTo install optional machine learning dependencies:\n\n```bash\n# Add ML dependencies for embedding generation\npip install 'mcp-memory-service[ml]'\n```\n\n#### Homebrew PyTorch Integration\n\nFor macOS users who prefer to use Homebrew's PyTorch installation:\n\n```bash\n# Install PyTorch via Homebrew\nbrew install pytorch\n\n# Run MCP Memory Service with Homebrew PyTorch integration\n./run_with_homebrew.sh\n```\n\nThis integration offers several benefits:\n- Uses Homebrew's isolated Python environment for PyTorch\n- Avoids dependency conflicts with Claude Desktop\n- Reduces memory usage in the main process\n- Provides better stability in resource-constrained environments\n\nFor detailed documentation on the Homebrew PyTorch integration:\n- [Homebrew Integration Guide](docs/integration/homebrew.md) - Technical journey and solution architecture\n\n#### Migration Between Backends\n\n```bash\n# Migrate from ChromaDB to SQLite-vec\npython migrate_to_sqlite_vec.py\n\n# Full migration with backup\npython scripts/migrate_storage.py \\\n --from chroma --to sqlite_vec \\\n --backup --backup-path backup.json\n```\n\nFor detailed SQLite-vec setup, migration, and troubleshooting, see the [SQLite-vec Backend Guide](docs/sqlite-vec-backend.md).\n\n## Memory Operations\n\nThe memory service provides the following operations through the MCP server:\n\n### Core Memory Operations\n\n1. `store_memory` - Store new information with optional tags\n2. `retrieve_memory` - Perform semantic search for relevant memories\n3. `recall_memory` - Retrieve memories using natural language time expressions \n4. `search_by_tag` - Find memories using specific tags\n5. `exact_match_retrieve` - Find memories with exact content match\n6. `debug_retrieve` - Retrieve memories with similarity scores\n\n### Database Management\n\n7. `create_backup` - Create database backup\n8. `get_stats` - Get memory statistics\n9. `optimize_db` - Optimize database performance\n10. `check_database_health` - Get database health metrics\n11. `check_embedding_model` - Verify model status\n\n### Memory Management\n\n12. `delete_memory` - Delete specific memory by hash\n13. `delete_by_tag` - **Enhanced**: Delete memories with specific tag(s) - supports both single tags and multiple tags\n14. `delete_by_tags` - **New**: Explicitly delete memories containing any of the specified tags (OR logic)\n15. `delete_by_all_tags` - **New**: Delete memories containing all specified tags (AND logic)\n16. `cleanup_duplicates` - Remove duplicate entries\n\n### API Consistency Improvements\n\n**Issue 5 Resolution**: Enhanced tag deletion functionality for consistent API design.\n\n- **Before**: `search_by_tag` accepted arrays, `delete_by_tag` only accepted single strings\n- **After**: Both operations now support flexible tag handling\n\n```javascript\n// Single tag deletion (backward compatible)\ndelete_by_tag(\"temporary\")\n\n// Multiple tag deletion (new!)\ndelete_by_tag([\"temporary\", \"outdated\", \"test\"]) // OR logic\n\n// Explicit methods for clarity\ndelete_by_tags([\"tag1\", \"tag2\"]) // OR logic \ndelete_by_all_tags([\"urgent\", \"important\"]) // AND logic\n```\n\n### Example Usage\n\n```javascript\n// Store memories with tags\nstore_memory(\"Project deadline is May 15th\", {tags: [\"work\", \"deadlines\", \"important\"]})\nstore_memory(\"Grocery list: milk, eggs, bread\", {tags: [\"personal\", \"shopping\"]})\nstore_memory(\"Meeting notes from sprint planning\", {tags: [\"work\", \"meetings\", \"important\"]})\n\n// Search by multiple tags (existing functionality)\nsearch_by_tag([\"work\", \"important\"]) // Returns memories with either tag\n\n// Enhanced deletion options (new!)\ndelete_by_tag(\"temporary\") // Delete single tag (backward compatible)\ndelete_by_tag([\"temporary\", \"outdated\"]) // Delete memories with any of these tags\ndelete_by_tags([\"personal\", \"shopping\"]) // Explicit multi-tag deletion\ndelete_by_all_tags([\"work\", \"important\"]) // Delete only memories with BOTH tags\n```\n\n## 🧠 Dream-Inspired Memory Consolidation\n\nThe memory consolidation system operates autonomously in the background, inspired by how human memory works during sleep cycles. It automatically organizes, compresses, and manages your memories across multiple time horizons.\n\n### Quick Start\nEnable consolidation with a single environment variable:\n```bash\nexport MCP_CONSOLIDATION_ENABLED=true\n```\n\n### How It Works\n- **Daily consolidation** (light processing): Updates memory relevance and basic organization\n- **Weekly consolidation**: Discovers creative associations between memories\n- **Monthly consolidation**: Performs semantic clustering and intelligent compression\n- **Quarterly/Yearly consolidation**: Deep archival and long-term memory management\n\n### New MCP Tools Available\nOnce enabled, you get access to powerful new consolidation tools:\n- `consolidate_memories` - Manually trigger consolidation for any time horizon\n- `get_consolidation_health` - Monitor system health and performance\n- `get_consolidation_stats` - View processing statistics and insights\n- `schedule_consolidation` - Configure autonomous scheduling\n- `get_memory_associations` - Explore discovered memory connections\n- `get_memory_clusters` - Browse semantic memory clusters\n- `get_consolidation_recommendations` - Get AI-powered memory management advice\n\n### Advanced Configuration\nFine-tune the consolidation system through environment variables:\n```bash\n# Archive location (default: ~/.mcp_memory_archive)\nexport MCP_CONSOLIDATION_ARCHIVE_PATH=/path/to/archive\n\n# Retention periods (days)\nexport MCP_RETENTION_CRITICAL=365 # Critical memories\nexport MCP_RETENTION_REFERENCE=180 # Reference materials \nexport MCP_RETENTION_STANDARD=30 # Standard memories\nexport MCP_RETENTION_TEMPORARY=7 # Temporary memories\n\n# Association discovery settings\nexport MCP_ASSOCIATION_MIN_SIMILARITY=0.3 # Sweet spot range\nexport MCP_ASSOCIATION_MAX_SIMILARITY=0.7 # for creative connections\n\n# Autonomous scheduling (cron-style)\nexport MCP_SCHEDULE_DAILY=\"02:00\" # 2 AM daily\nexport MCP_SCHEDULE_WEEKLY=\"SUN 03:00\" # 3 AM on Sundays\nexport MCP_SCHEDULE_MONTHLY=\"01 04:00\" # 4 AM on 1st of month\n```\n\n### Performance\n- Designed to process 10k+ memories efficiently\n- Automatic hardware optimization (CPU/GPU/MPS)\n- Safe archival system - no data is ever permanently deleted\n- Full recovery capabilities for all archived memories\n\n## Configuration Options\n\nConfigure through environment variables:\n\n```\nCHROMA_DB_PATH: Path to ChromaDB storage\nBACKUP_PATH: Path for backups\nAUTO_BACKUP_INTERVAL: Backup interval in hours (default: 24)\nMAX_MEMORIES_BEFORE_OPTIMIZE: Threshold for auto-optimization (default: 10000)\nSIMILARITY_THRESHOLD: Default similarity threshold (default: 0.7)\nMAX_RESULTS_PER_QUERY: Maximum results per query (default: 10)\nBACKUP_RETENTION_DAYS: Number of days to keep backups (default: 7)\nLOG_LEVEL: Logging level (default: INFO)\n\n# HTTP API and Authentication\nMCP_API_KEY: API key for HTTP authentication (optional, no default)\nMCP_HTTP_ENABLED: Enable HTTP server mode (default: false)\nMCP_HTTP_HOST: HTTP server bind address (default: 127.0.0.1)\nMCP_HTTP_PORT: HTTP server port (default: 8000)\n\n# Hardware and backend configuration\nMCP_MEMORY_STORAGE_BACKEND: Storage backend to use (chromadb or sqlite_vec)\nMCP_MEMORY_SQLITE_PATH: Path to SQLite-vec database file\nPYTORCH_ENABLE_MPS_FALLBACK: Enable MPS fallback for Apple Silicon (default: 1)\nMCP_MEMORY_USE_ONNX: Use ONNX Runtime for CPU-only deployments (default: 0)\nMCP_MEMORY_USE_DIRECTML: Use DirectML for Windows acceleration (default: 0)\nMCP_MEMORY_MODEL_NAME: Override the default embedding model\nMCP_MEMORY_BATCH_SIZE: Override the default batch size\n```\n\n### API Key Security\n\nFor production deployments with HTTP API enabled, always set a secure API key:\n\n```bash\n# Generate a secure API key\nexport MCP_API_KEY=\"$(openssl rand -base64 32)\"\n\n# Or use your preferred method\nexport MCP_API_KEY=\"your-secure-api-key-here\"\n```\n\nThe API key is required in the `Authorization: Bearer ` header for all HTTP requests when set.\n\n## 🚀 Service Installation (NEW!)\n\nInstall MCP Memory Service as a native system service for automatic startup:\n\n### Cross-Platform Service Installer\n\n```bash\n# Install as a service (auto-detects OS)\npython install_service.py\n\n# Start the service\npython install_service.py --start\n\n# Check service status\npython install_service.py --status\n\n# Stop the service\npython install_service.py --stop\n\n# Uninstall the service\npython install_service.py --uninstall\n```\n\nThe installer provides:\n- ✅ **Automatic OS detection** (Windows, macOS, Linux)\n- ✅ **Native service integration** (systemd, LaunchAgent, Windows Service)\n- ✅ **Automatic startup** on boot/login\n- ✅ **Service management commands**\n- ✅ **Secure API key generation**\n- ✅ **Platform-specific optimizations**\n\nFor detailed instructions, see the [Service Installation Guide](docs/guides/service-installation.md).\n\n## Hardware Compatibility\n\n| Platform | Architecture | Accelerator | Status | Notes |\n|----------|--------------|-------------|--------|-------|\n| macOS | Apple Silicon (M1/M2/M3) | MPS | ✅ Fully supported | Best performance |\n| macOS | Apple Silicon under Rosetta 2 | CPU | ✅ Supported with fallbacks | Good performance |\n| macOS | Intel | CPU | ✅ Fully supported | Good with optimized settings |\n| Windows | x86_64 | CUDA | ✅ Fully supported | Best performance |\n| Windows | x86_64 | DirectML | ✅ Supported | Good performance |\n| Windows | x86_64 | CPU | ✅ Supported with fallbacks | Slower but works |\n| Linux | x86_64 | CUDA | ✅ Fully supported | Best performance |\n| Linux | x86_64 | ROCm | ✅ Supported | Good performance |\n| Linux | x86_64 | CPU | ✅ Supported with fallbacks | Slower but works |\n| Linux | ARM64 | CPU | ✅ Supported with fallbacks | Slower but works |\n| Any | Any | No PyTorch | ✅ Supported with SQLite-vec | Limited functionality, very lightweight |\n\n## Testing\n\n```bash\n# Install test dependencies\npip install pytest pytest-asyncio\n\n# Run all tests\npytest tests/\n\n# Run specific test categories\npytest tests/test_memory_ops.py\npytest tests/test_semantic_search.py\npytest tests/test_database.py\n\n# Verify environment compatibility\npython scripts/verify_environment_enhanced.py\n\n# Verify PyTorch installation on Windows\npython scripts/verify_pytorch_windows.py\n\n# Perform comprehensive installation verification\npython scripts/test_installation.py\n```\n\n## FAQ\n\n### Can I run the MCP Memory Service across multiple applications simultaneously?\n\n**Yes!** The MCP Memory Service now features **universal multi-client coordination** that enables seamless concurrent access from any MCP-compatible applications. Share memories between Claude Desktop, Claude Code, VS Code, Continue, Cursor, and other MCP clients with automatic coordination.\n\n**🚀 Integrated Setup (Recommended):**\n```bash\n# During installation, you'll be prompted:\npython install.py\n\n# Would you like to configure multi-client access? (y/N): y\n# ✅ Automatically detects and configures all your MCP clients!\n```\n\n**🔧 Manual Setup:**\n```bash\npython examples/setup/setup_multi_client_complete.py\n```\n\n**🌐 Universal Compatibility:**\n- ✅ **Claude Desktop + Claude Code**: Original use case with automatic config\n- ✅ **VS Code with MCP Extension**: Seamless integration instructions\n- ✅ **Continue IDE**: Automatic configuration file updates\n- ✅ **Cursor IDE**: MCP extension support with guidance\n- ✅ **Any MCP Client**: Generic configuration for future applications\n\n**Key Benefits:**\n- ✅ **Automatic Coordination**: Intelligent detection of optimal access mode\n- ✅ **Universal Setup**: Works with any MCP-compatible application\n- ✅ **Shared Memory**: All clients access the same memory database\n- ✅ **No Lock Conflicts**: WAL mode prevents database locking issues\n- ✅ **IDE-Agnostic**: Switch between development tools while maintaining context\n\n**Multi-Client Features:**\n- **Phase 1: WAL Mode** - Direct SQLite access with Write-Ahead Logging (default)\n- **Phase 2: HTTP Coordination** - Advanced server-based coordination (optional)\n- **Automatic Retry Logic** - Handles transient lock conflicts gracefully\n- **Performance Optimized** - Tuned for concurrent access patterns\n\n**Technical Implementation:**\n- **SQLite WAL Mode**: Multiple readers + single writer coordination\n- **HTTP Auto-Detection**: Intelligent server coordination when beneficial\n- **Connection Retry**: Exponential backoff for robust access\n- **Shared Database**: Single source of truth across all clients\n\n**Setup Guides:**\n- **🚀 Quick Start:** Integrated into `python install.py` with automatic detection\n- **📖 Universal Setup:** [Universal Multi-Client Setup Guide](docs/integration/multi-client.md)\n- **🔧 Manual Setup:** [Multi-Client Setup Guide](docs/integration/multi-client.md) \n- **⚙️ Legacy Setup:** `python examples/setup/setup_multi_client_complete.py`\n\n## Troubleshooting\n\nSee the [Installation Guide](docs/installation/master-guide.md) and [Troubleshooting Guide](docs/troubleshooting/general.md) for detailed troubleshooting steps.\n\n### Quick Troubleshooting Tips\n\n- **Windows PyTorch errors**: Use `python scripts/install_windows.py`\n- **macOS Intel dependency conflicts**: Use `python install.py --force-compatible-deps`\n- **Recursion errors**: Run `python scripts/fix_sitecustomize.py` \n- **Environment verification**: Run `python scripts/verify_environment_enhanced.py`\n- **Memory issues**: Set `MCP_MEMORY_BATCH_SIZE=4` and try a smaller model\n- **Apple Silicon**: Ensure Python 3.10+ built for ARM64, set `PYTORCH_ENABLE_MPS_FALLBACK=1`\n- **Installation testing**: Run `python scripts/test_installation.py`\n\n## 📚 Comprehensive Documentation\n\n### Installation & Setup\n- **[Master Installation Guide](docs/installation/master-guide.md)** - Complete installation guide with hardware-specific paths\n- **[Storage Backend Guide](docs/guides/STORAGE_BACKENDS.md)** ⭐ **NEW** - Comprehensive CLI options including multi-client setup\n- **[Multi-Client Setup](docs/integration/multi-client.md)** ⭐ **NEW** - Integrated setup for any MCP application\n- **[Storage Backend Comparison](docs/guides/STORAGE_BACKENDS.md)** - Detailed comparison and selection guide\n- **[Migration Guide](docs/guides/migration.md)** - ChromaDB to SQLite-vec migration instructions\n\n### Platform-Specific Guides\n- **[Intel Mac Setup Guide](docs/platforms/macos-intel.md)** - Comprehensive guide for Intel Mac users\n- **[Legacy Mac Guide](docs/platforms/macos-intel.md)** - Optimized for 2015 MacBook Pro and older Intel Macs\n- **[Windows Setup](docs/guides/windows-setup.md)** - Windows-specific installation and troubleshooting\n- **[Ubuntu Setup](docs/guides/UBUNTU_SETUP.md)** - Linux server installation guide\n\n### API & Integration\n- **[HTTP/SSE API](docs/IMPLEMENTATION_PLAN_HTTP_SSE.md)** - New web interface documentation\n- **[Claude Desktop Integration](docs/guides/claude_integration.md)** - Configuration examples\n- **[Integrations](docs/integrations.md)** - Third-party tools and extensions\n\n### Advanced Topics\n- **[Multi-Client Architecture](docs/development/multi-client-architecture.md)** ⭐ **NEW** - Technical implementation details\n- **[Homebrew PyTorch Integration](docs/integration/homebrew.md)** - Using system PyTorch\n- **[Docker Deployment](docs/deployment/docker.md)** - Container-based deployment\n- **[Performance Optimization](docs/implementation/performance.md)** - Tuning for different hardware\n\n### Troubleshooting & Support\n- **[General Troubleshooting](docs/troubleshooting/general.md)** - Common issues and solutions\n- **[Hardware Compatibility](docs/DOCUMENTATION_AUDIT.md)** - Compatibility matrix and known issues\n\n### Quick Commands\n```bash\n# Get personalized setup recommendations\npython install.py --help-detailed\n\n# Generate hardware-specific setup guide\npython install.py --generate-docs\n\n# Test your installation\npython scripts/test_memory_simple.py\n```\n\n## Project Structure\n\n```\nmcp-memory-service/\n├── src/mcp_memory_service/ # Core package code\n│ ├── __init__.py\n│ ├── config.py # Configuration utilities\n│ ├── models/ # Data models\n│ ├── storage/ # Storage implementations\n│ ├── utils/ # Utility functions\n│ └── server.py # Main MCP server\n├── scripts/ # Helper scripts\n├── memory_wrapper.py # Windows wrapper script\n├── install.py # Enhanced installation script\n└── tests/ # Test suite\n```\n\n## Development Guidelines\n\n- Python 3.10+ with type hints\n- Use dataclasses for models\n- Triple-quoted docstrings for modules and functions\n- Async/await pattern for all I/O operations\n- Follow PEP 8 style guidelines\n- Include tests for new features\n\n### Git Setup for Contributors\n\nAfter cloning the repository, run the setup script to configure automated `uv.lock` conflict resolution:\n\n```bash\n./scripts/setup-git-merge-drivers.sh\n```\n\nThis enables automatic resolution of `uv.lock` merge conflicts by:\n1. Using the incoming version to resolve conflicts\n2. Automatically running `uv sync` to regenerate the lock file\n3. Ensuring consistent dependency resolution across all environments\n\nThe setup is required only once per clone and benefits all contributors by eliminating manual conflict resolution.\n\n## License\n\nMIT License - See LICENSE file for details\n\n## Acknowledgments\n\n- ChromaDB team for the vector database\n- Sentence Transformers project for embedding models\n- MCP project for the protocol specification\n\n## 🎯 Why Sponsor MCP Memory Service?\n\n## 🏆 In Production\n- Deployed on Glama.ai\n- Managing 300+ enterprise memories\n- Processing queries in <1 second\n\n### Production Impact\n- **319+ memories** actively managed\n- **828ms** average query response time\n- **100%** cache hit ratio performance\n- **20MB** efficient vector storage\n\n### Developer Community\n- Complete MCP protocol implementation\n- Cross-platform compatibility\n- React dashboard with real-time statistics\n- Comprehensive documentation\n\n### Enterprise Features\n- Semantic search with sentence-transformers\n- Tag-based categorization system\n- Automatic backup and optimization\n- Health monitoring dashboard\n\n## Contact\n\n[Telegram](https://t.me/doobeedoo)\n\n## Integrations\n\nThe MCP Memory Service can be extended with various tools and utilities. See [Integrations](docs/integrations.md) for a list of available options, including:\n\n- [MCP Memory Dashboard](https://github.com/doobidoo/mcp-memory-dashboard) - Web UI for browsing and managing memories\n- [Claude Memory Context](https://github.com/doobidoo/claude-memory-context) - Inject memory context into Claude project instructions\n","url","https://github.com/doobidoo/mcp-memory-service","githubRepository",{"_149":150,"_151":152,"_153":154,"_155":156,"_33":157,"_158":159,"_164":165,"_166":167,"_169":170},"defaultBranch","main","fullName","doobidoo/mcp-memory-service","githubTreeId",170789,"language","Python","mcp-memory-service","owner",{"_160":161,"_162":163},"login","doobidoo","uid","ezae8d08rn","recentStargazers",32,"spdxLicense",{"_33":168},"Apache 2.0","stargazers",566,"glamaJson",469,"inspectable",false,"integrations",[177],{"_178":179,"_31":185,"_162":186},"brand",{"_180":181,"_33":182,"_183":184},"logoSvgPath","M13.762 4.29a6.51 6.51 0 0 0-5.669 3.332 3.571 3.571 0 0 0-1.558-.36 3.571 3.571 0 0 0-3.516 3A4.918 4.918 0 0 0 0 14.796a4.918 4.918 0 0 0 4.92 4.914 4.93 4.93 0 0 0 .617-.045h14.42c2.305-.272 4.041-2.258 4.043-4.589v-.009a4.594 4.594 0 0 0-3.727-4.508 6.51 6.51 0 0 0-6.511-6.27z","iCloud","slug","icloud","Mentioned as a potential cloud storage option where users should ensure sync is complete before accessing from another device.","210wn3xzvx","namespace","npmPackage","redditPostForeignId","1hn367f","scores",{"_193":194,"_195":-5,"_196":-5},"license",100,"quality","security","simpleIcon","supportedPlatforms",[200,201],"WINDOWS","LINUX","toolCount",3,"bzvl3lz34o","updatedAt","2025-08-13T17:20:35.067067250Z","meta",[209,212,214,221,224,226,229,232,235,238,241,243,246,249,251,253,255,257,259],{"_210":211},"title","MCP Memory Service | Glama",{"_213":42,"_33":31},"content",{"_215":216,"_217":218,"_219":220},"href","https://glama.ai/mcp/servers/@doobidoo/mcp-memory-service","rel","canonical","tagName","link",{"_213":43,"_222":223},"property","og:title",{"_213":42,"_222":225},"og:description",{"_213":227,"_222":228},"https://glama.ai/logo.png","og:logo",{"_213":230,"_222":231},"https://glama.ai/mcp/servers/@doobidoo/mcp-memory-service/og-image","og:image",{"_213":233,"_222":234},"630","og:image:height",{"_213":236,"_222":237},"1200","og:image:width",{"_213":239,"_222":240},"website","og:type",{"_213":216,"_222":242},"og:url",{"_213":244,"_222":245},"Glama – MCP Hosting Platform","og:site_name",{"_213":247,"_33":248},"summary_large_image","twitter:card",{"_213":244,"_33":250},"twitter:title",{"_213":42,"_33":252},"twitter:description",{"_213":230,"_33":254},"twitter:image",{"_213":233,"_33":256},"twitter:image:height",{"_213":236,"_33":258},"twitter:image:width",{"_260":261},"script:ld+json",{"_262":263,"_264":265},"@context","http://schema.org","@graph",[266],{"_262":267,"_268":269,"_270":271},"https://schema.org","@type","BreadcrumbList","itemListElement",[272,279,283,286],{"_268":273,"_274":275,"_33":276,"_277":278},"ListItem","item","https://glama.ai/","Glama","position",1,{"_268":273,"_274":280,"_33":281,"_277":282},"https://glama.ai/mcp","MCP",2,{"_268":273,"_274":284,"_33":285,"_277":203},"https://glama.ai/mcp/servers","Servers",{"_268":273,"_274":216,"_33":43,"_277":287},4,"visitorIsMaintainer","routes/_public/mcp/servers/~namespace/~slug/_pages/_explorer/blob/~branch/~/_route",{"_291":292,"_296":150,"_29":297,"_207":320,"_344":345,"_346":347},"blob",{"_213":293,"_294":295},"# Auto-resolve lock file conflicts and regenerate lock file\nuv.lock merge=uv-lock-merge","fileName",".gitattributes","branch",{"_85":86,"_87":298,"_13":310,"_140":42,"_141":42,"_142":311,"_147":312,"_171":-5,"_16":172,"_173":174,"_175":315,"_33":43,"_187":161,"_188":-5,"_189":190,"_191":318,"_197":-5,"_183":157,"_198":319,"_202":203,"_162":204,"_205":206},{"_89":299,"_129":309,"_96":131},{"_91":300,"_98":301,"_102":302,"_106":303,"_110":304,"_114":305,"_118":306,"_122":307,"_126":308},{"_93":94,"_31":95,"_96":97},{"_93":100,"_31":101,"_96":97},{"_93":104,"_31":105,"_96":97},{"_93":108,"_31":109,"_96":97},{"_93":112,"_31":113,"_96":97},{"_93":116,"_31":117,"_96":97},{"_93":120,"_31":121,"_96":97},{"_93":124,"_31":125,"_96":97},{"_93":104,"_31":128,"_96":97},[],[133,134,135,136,137,138,139],{"_35":144,"_145":146},{"_149":150,"_151":152,"_153":154,"_155":156,"_33":157,"_158":313,"_164":165,"_166":314,"_169":170},{"_160":161,"_162":163},{"_33":168},[316],{"_178":317,"_31":185,"_162":186},{"_180":181,"_33":182,"_183":184},{"_193":194,"_195":-5,"_196":-5},[200,201],[321,323,325,327,329,330,331,333,334,335,336,337,338,339,340,341,342,343],{"_210":322},"MCP Server | Glama",{"_213":324,"_33":31},"Enterprise-grade security, privacy, and features like agents, MCP, prompt templates, and more.",{"_215":326,"_217":218,"_219":220},"https://glama.ai/mcp/servers/@doobidoo/mcp-memory-service/blob/main/",{"_213":328,"_222":223},"MCP Server",{"_213":324,"_222":225},{"_213":227,"_222":228},{"_213":332,"_222":231},"https://glama.ai/generated-images/og?title=MCP+Server",{"_213":233,"_222":234},{"_213":236,"_222":237},{"_213":239,"_222":240},{"_213":326,"_222":242},{"_213":244,"_222":245},{"_213":247,"_33":248},{"_213":244,"_33":250},{"_213":324,"_33":252},{"_213":332,"_33":254},{"_213":233,"_33":256},{"_213":236,"_33":258},"path","","treeMembers",[348,351,353,355,357,359,361,363,365,367,369,371,373,375,376,378,380,382,384,386,388,390,392,394,396,398,400,402,404,406,408],{"_344":349,"_96":350},"archive","tree",{"_344":352,"_96":350},".claude",{"_344":354,"_96":350},"claude_commands",{"_344":356,"_96":350},"docs",{"_344":358,"_96":350},"examples",{"_344":360,"_96":350},".github",{"_344":362,"_96":350},"scripts",{"_344":364,"_96":350},"src",{"_344":366,"_96":350},"tests",{"_344":368,"_96":350},"tools",{"_344":370,"_96":350},".uv",{"_344":372,"_96":291},"CHANGELOG.md",{"_344":374,"_96":291},"CLAUDE.md",{"_344":295,"_96":291},{"_344":377,"_96":291},".gitignore",{"_344":379,"_96":291},"install.py",{"_344":381,"_96":291},"install_service.py",{"_344":383,"_96":291},"install_service.sh",{"_344":385,"_96":291},"LICENSE",{"_344":387,"_96":291},".mcp.json.backup",{"_344":389,"_96":291},".mcp.json.template",{"_344":391,"_96":291},"mcp-memory.service",{"_344":393,"_96":291},"NOTICE",{"_344":395,"_96":291},"pyproject.toml",{"_344":397,"_96":291},"pytest.ini",{"_344":399,"_96":291},"PYTORCH_DOWNLOAD_FIX.md",{"_344":401,"_96":291},"README.md",{"_344":403,"_96":291},"ROADMAP.md",{"_344":405,"_96":291},"smithery.yaml",{"_344":407,"_96":291},"SPONSORS.md",{"_344":409,"_96":291},"uv.lock","actionData","errors"]

MCP directory API