Postg Mem

🚀 PostgMem

PostgMem is a server implementation based on the Model Context Protocol (MCP), using PostgreSQL and pgvector to provide vector memory storage for AI applications.

🚀 Quick Start

PostgMem is a .NET-based service that allows AI agents to store, retrieve, and access memories through similarity searches. It leverages PostgreSQL and the pgvector extension to provide efficient similarity search capabilities.

✨ Features

• Store structured memories using vector embeddings.
• Retrieve a single memory by ID.
• Perform semantic searches via vector similarity.
• Filter search results using tags.
• Integrate with the Model Context Protocol (MCP) for easy use with AI agents.

🔧 Technical Details

• .NET 9.0
• PostgreSQL with the pgvector extension
• Model Context Protocol (MCP)
• ASP.NET Core
• Npgsql for PostgreSQL connection

📦 Installation

Prerequisites

• .NET 9.0 SDK
• A PostgreSQL database with the pgvector extension installed
• A text embedding API (Ollama is used in the default configuration)

Database Configuration

1. Create a PostgreSQL database for the application.
2. Install the pgvector extension in the database:

   CREATE EXTENSION vector;
   

3. Create the memories table:

   CREATE TABLE memories (
       id UUID PRIMARY KEY,
       type TEXT NOT NULL,
       content JSONB NOT NULL,
       source TEXT NOT NULL,
       embedding VECTOR(384) NOT NULL,
       tags TEXT[] NOT NULL,
       confidence DOUBLE PRECISION NOT NULL,
       created_at TIMESTAMP WITH TIME ZONE NOT NULL,
       updated_at TIMESTAMP WITH TIME ZONE NOT NULL
   );
   

Environment Configuration

Configure the application settings in the .env file of the project:

ConnectionStrings__Storage="Host=localhost;Database=mcp_memory;Username=postgres;Password=postgres"
Embeddings__ApiUrl=http://localhost:11434/
Embeddings__Model=all-minilm:33m-l12-v2-fp16

• ConnectionStrings__Storage: PostgreSQL connection string
• Embeddings__ApiUrl: URL of the embedding API (Ollama is used by default)
• Embeddings__Model: Embedding model to use

Running the Application

1. Navigate to the PostgMem directory.
2. Run the application:

   dotnet run
   

3. The MCP server will be available at http://localhost:5000 by default.

💻 Usage Examples

MCP Tools

Storage

Store a new memory in the database.

Parameters:

• type (string): Type of the memory (e.g., 'conversation', 'document', etc.)
• content (string): Memory content as a JSON object
• source (string): Source of the memory (e.g., 'user', 'system', etc.)
• tags (string array): Optional tags for memory categorization
• confidence (double precision float): Confidence score of the memory (0.0 to 1.0)

Search

Search for similar memories by the provided text.

Parameters:

• query (string): Text to search for
• limit (integer): Maximum number of results to return (default: 10)
• minSimilarity (double precision float): Minimum similarity threshold (0.0 to 1.0) (default: 0.7)
• filterTags (string array): Optional tags to filter the search results

Get

Retrieve a specific memory by ID.

Parameters:

• id (Guid): ID of the memory to retrieve

Delete

Delete a memory by ID.

Parameters:

• id (Guid): ID of the memory to delete

📚 Documentation

Implementation Details

• Memory.cs: Defines the memory data model.
• Storage.cs: Handles database operations, including storing and retrieving memories.
• EmbeddingService.cs: Generates text vector embeddings.
• MemoryTools.cs: Implements MCP tools to interact with the memory storage.

📄 License

[Fill in your license information here]

🤝 Contribution

[Add contribution guidelines here, if applicable]

📞 Contact

[Provide contact information or support details here]