- To provide users with simple learning plans that allow you to quickly learn Semantic Kernel starting with the basic to advanced concepts
- Use a lesson based approach for understanding of concepts that allow you to progressively learn advanced concepts
- Allow much faster adoption of AI using by leveraging proven practices
- Provide users with reusable building blocks resulting in ability to deliver solutions faster
I am not using DevContainers because I have an issue with Docker that will requirement to reformat my laptop, and I simply have not had time to do that at the moment.
In this lesson we cover the fundimental concepts of Semantic Kernel and what is needed to build your first console app.
In this lesson I reiterate the importance of understanding the ChatCompletion Message structure then we dive right into SK Plugins and how they are the body of your AI solution. We will drill deeper into SK Prompts and the various ways you can use prompts and how everything is a Plugin and why this is important.
We are building a foundation of Semantic Kernel Concepts that will set you up for success with your AI engagements.
In this lesson we will look at two options for how to implement a Retrieval Augmentation Generation solution that implement allows us to chat with our documents. We leverage Azure AI Search for the Vector and Semantic Search capabilities. You can of course use other systems like CosmosDB for MongoDB vCore. We will look at using the DataSources extension with the Azure Open AI Completion Endpoint as well as leveraging a custom Semantic Kernel Plugin.
TBD - Will add a description when I get time to come back to this one.
In this lession we take a look at word and phrase normaliztion.
Normalization is the process of converting a word or phrase into a standard or canonical form. It often involves removing variations, redundancies, or unnecessary details to make the term consistent for comparison, storage, or further processing. This can include:
- Singularization: Converting plural forms into their singular counterparts (e.g., ""Ducks"" → ""Duck"").
- Removing Articles or Prepositions: Stripping words like ""the,"" ""of,"" or ""in"" if they are not essential to the meaning (e.g., ""The Walking Dead"" → ""Walking Dead"").
- Standardizing Case: Ensuring consistent letter casing (e.g., lowercase or title case).
- Simplifying Structure: Reordering words or removing non-critical ones to clarify the core meaning (e.g., ""Walking Dead, the"" → ""Walking Dead"").
- Lemmatization: Converting inflected forms of words to their base or dictionary form (e.g., ""running"" → ""run"").
Normalization of words or phrases are often used in combination with a semantic similarity search. The idea is to take a word or a phrase convert it into a normalized version then use this to retrieve records using some sort of LIKE operation and using those records leverage the LLM to perform a similarity match returning the best match of the records you shared with the LLM. Very similar to have documents indexed in Azure AI Search and perform a semantic similarity search against those documents.
Each lesson will have a separate project in the Semantic-Kernel-101.sln. If you need to see a working example for the lesson you can find them here. The idea is that you do the homework, put hands on the keyboard and you create the project that that covers the topics outlined in each lesson. If you need a little help you can use this solution to get you moving.