Kernel Memory is a multi-modal memory service and .NET library for indexing and RAG, with Docker and Azure deployment, focusing on memory pipelines and semantic search.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Kernel Memory is a memory solution for AI/LLM scenarios, providing multi-modal indexing, retrieval augmented generation (RAG), and memory processing. It is available as a Web Service, a Docker container, a Plugin for Copilot/Semantic Kernel, and a .NET library for embedded use. The repository describes a reference implementation and best practices for memory in AI applications.
How it works
The project describes a pipeline-based ingestion model where documents are extracted, chunked, vectorized, and stored. It supports custom pipelines by allowing users to replace or add handlers in the orchestrator, enabling flexible data processing. Retrieval uses embeddings and a memory store, returning answers with citations and source references. Extensions cover various AI services, vector stores, file storage options, and ingestion queues. The service can be deployed as a remote web service or embedded serverless component.
Getting started
Key usage examples shown include:
var builder = DistributedApplication.CreateBuilder();
builder.AddContainer("kernel-memory", "kernelmemory/service")
.WithEnvironment("KernelMemory__TextGeneratorType", "OpenAI")
.WithEnvironment("KernelMemory__DataIngestion__EmbeddingGeneratorTypes__0", "OpenAI")
.WithEnvironment("KernelMemory__Retrieval__EmbeddingGeneratorType", "OpenAI")
.WithEnvironment("KernelMemory__Services__OpenAI__APIKey", "...your OpenAI key...");
builder.Build().Run();
For Docker testing:
docker run -e OPENAI_API_KEY="..." -it --rm -p 9001:9001 kernelmemory/service
To customize settings via development.json and run the Docker image with the file mounted:
# Windows
docker run --volume .\appsettings.Development.json:/app/appsettings.Production.json -it --rm -p 9001:9001 kernelmemory/service
# Linux/macOS
docker run --volume ./appsettings.Development.json:/app/appsettings.Production.json -it --rm -p 9001:9001 kernelmemory/service
Recent releases
Latest changes show versioned Docker-related updates and compatibility improvements:
- 0.98.250508.3 (2025-05-09): Core.csproj update and split character for Japanese
- 0.98.250324.1 (2025-03-24): Add ForcePathStyle support to AWSS3Config for MinIO compatibility
- 0.98.250323.1 (2025-03-24): Install ICU in Docker image to support SQL Server memory
New contributors include @eersinyildizz.
Traction
The repository has 2174 stars and 397 forks. It shows 0 open issues as listed.
Behind the repo
Kernel Memory is positioned as an archived research project with multiple deployment options (Web Service, Docker, Plugin, and embedded .NET library). It references Azure deployment guidance and external services for embeddings, vector stores, and file storage.
Caveats
License: MIT. The repository description notes the project is archived, intended as learning material rather than production software, with no formal support. The README cautions that this is an archived research project and should be used with caution.