jihadkhawaja/mem0sharp

Mem0Sharp gives AI applications built with .NET a way to remember things about users over time, instead of forgetting everything as soon as a conversation ends. It is a C# implementation inspired by an existing open source project called Mem0, though it is built independently and is not officially connected to that project. In plain terms, this library lets a program save short pieces of information, like a user's preferences or facts mentioned in a conversation, and later search for the most relevant ones using meaning based search rather than exact keyword matching. For example, an application could save that a user prefers dark mode, and later ask what editor settings that user likes, getting back the relevant saved memory even though the wording is different. The library supports adding, searching, updating, and deleting these memories, and lets developers organize them by user, session, or agent. Each memory can also carry extra metadata. For getting started quickly, it includes a simple in-memory storage option with no external dependencies, useful for testing or local development. For real production use, it also supports connecting to an OpenAI-compatible AI service for generating the underlying search embeddings, and can store memories permanently in a PostgreSQL database using a special extension called pgvector, which is designed for fast similarity search. Installing it is straightforward for a .NET developer: it is distributed as a NuGet package, the standard way to add libraries to .NET projects, and can be added with a single command. The project also includes documentation covering getting started, configuring providers, and a full API reference, along with a test suite that can be run with the standard dotnet test command. Mem0Sharp is licensed under the Apache License 2.0, a permissive license that allows free use, modification, and redistribution, including for commercial purposes, as long as license terms and attribution are respected.