Getting started with Kernel Memory

Kernel Memory is a multi-modal AI service that enables efficient indexing of datasets through custom data pipelines. It supports RAG, synthetic memory, prompt engineering, and custom semantic memory processing.

This video is a gentle introduction to Kernel Memory, and how we can run it in both Serverless and as a Web Service to index data, and use natural language queries to obtain answers from indexed data!

0:00 Introduction
1:22 Difference between Kernel Memory and Semantic Memory
2:46 Serverless mode
3:57 Kernel Memory Web Service
5:25 Demo - Running Kernel Memory in Serverless Mode
15:26 Demo - Running Kernel Memory in a Web Service
27:05 Wrap up

Connect with me!

[SilverLion]

Would definitely love to see more from this series. I have learned a lot, you are a great teacher.

Could you do a video about combining vector databases, using embeddings, memory, and RAG with semantic kernel?

[JaredWoodruff]

Great video Will, Thanks for sharing 🙂

[LearnWithVarun]

Hi, How to know how many tokens were consumed for the prompt execution?

[rodgerb]

Great video! Do you know why indexing to Azure AI search uses a different index structure than Elasticsearch or Postgres when using Kernel Memory? Elastic has a indexed “context” field but Azure AI search only has a “payload” field in the index you need to extract the text from?

[kai-odie]

Great video! Would love to see an example of using persistent storage, something like Postgres. Either way, keep up the good work!

[DeepikaAnand]

In this example you ask the question the memory so at that point only this document in what the memory aware of. right? And what algo does the memory use to figure out the answers? Is it just Semantic search or is it Hybrid search?

[natious]

Greate video. Extremely helpful. I would love to see one using Azure OpenAI for the Serverless mode scenario

[GWANCHAV]

Heyy

great video here. Will we get more anytime soon ?

THNX

[FabioAngela]

does it allows you to apply taks to uploaded documents? or custom schemes?
e.g. imagine I want to import different kind of documents, like pdf but also emails, can I specify the type of document so that if I ask for email contents it search only into emails and not pdfs?

[devgenesis]

Grt video.. how can we change the kernal memeory web service settings which we have configured through powershell.. is there any file created for this

[omaralejandroibanez]

Does anyone know how I can use km in an api? I ran some tests using this great example video but it still doesn't work for me