Semantic Kernel + Mem0 Memory C# .NET Example

This repository demonstrates how to integrate memory into Microsoft Semantic Kernel using a lightweight in-process memory provider called Mem0. The solution is structured as a multi-project .NET application that showcases storing, retrieving, and augmenting prompts with contextual memory in a Retrieval-Augmented Generation (RAG) flow.

---

Solution structure

• api/ ASP.NET Core API exposing endpoints for working with memory and AI agents.
• service/ Core services and abstractions, with dependency injection setup.
• infrastructure/ Agents and infrastructure wiring, including Semantic Kernel + Mem0 integration.
• model/ Shared models/DTOs (e.g., AgentResponse).

---

Key concepts

• Semantic Kernel orchestrates calls to LLMs, and memory.
• Mem0 minimal in-memory memory store that saves embeddings + text for semantic retrieval.

---

Features

• ASP.NET API to interact with Semantic Kernel and memory.
• Save and query conversation snippets in Mem0.
• Retrieve top-K relevant items with cosine similarity search.
• Demonstration of augmenting prompts with contextual memory.

---

Prerequisites

• .NET 8 SDK
• An LLM provider (Azure OpenAI or OpenAI API key)
• (Optional) Docker Desktop for running via docker-compose

---