Mastering Semantic Kernel by Creating Projects

Discover how kernel, orchestrator of semantic kernel, centralizes config and monitoring, manages services and plugins, routes prompts built from templates to an llm, with events and middleware for responsible ai.

Demonstrates the builder pattern with a pizza example in C#, showing fluent methods and a build function to produce the pizza and how semantic kernel uses it to instantiate kernels.

Create an OpenAI API key and an Azure key, then store it as an environment variable for use in Visual Studio to securely access semantic kernel demonstrations.

Learn to create an Azure OpenAI API key, deploy the service in portal.azure.com, retrieve the endpoint and key, and set environment variables for use in Visual Studio.

Create a Visual Studio console project, install the semantic kernel NuGet package, set environment variables for OpenAI and Azure keys, and build kernels using OpenAI or Azure OpenAI chat completion.

Create and invoke chat completion prompts using semantic kernel, switch between OpenAI and Azure OpenAI kernels, and configure output with max tokens and temperature.

Learn to implement chat completion streaming with semantic kernel, delivering real-time text updates as the model completes each message, demonstrated with OpenAI and Azure kernels.

Learn to generate images from text with semantic kernel by configuring OpenAI and Azure OpenAI kernels, using Dall-E 3 for text-to-image, and adjusting execution settings to customize quality and style.

Learn to generate audio files from text using semantic kernel by configuring text-to-speech with OpenAI and Azure OpenAI, selecting models, voices, and saving mp3 output.

Extract text from an audio file using semantic kernel by configuring OpenAI and Azure OpenAI audio-to-text services, selecting whisper models, and parsing the audio through a text content service.

Create a WordPress blog post generator using semantic kernel, generating title, featured image, and an audio file with structured paragraphs, headings, lists, code snippets, and bold text.

Create a sky post generator project, install semantic kernel and inspector console nuget packages, configure openai and azure openai kernels, and enable environment variables for generating a WordPress post.

This lecture demonstrates generating a WordPress blog post with semantic kernel by defining a blog post region, prompting for a topic, and using Gutenberg blocks.

Create a blog post featured image using Semantic Kernel by configuring a text-to-image service, crafting a topic-driven prompt, and generating a high-quality, vivid landscape image, then display the URL.

Create an audio file for a blog post by cleaning post content, invoking the text-to-audio service, configuring OpenAI options (voice alloy, mp3), and saving to a defined path for WordPress.

Publish generated blog content to a WordPress sandbox, create a temporary WordPress site, craft a post with audio and a featured image, and publish, noting optional SEO optimization.

Learn how to configure and test the chat completion service in semantic kernel by building kernels for OpenAI and Azure OpenAI, then query the capital of France and receive Paris.

Use semantic kernel to store and reuse chat history with a chat history object that accumulates user and assistant messages, guiding responses with system, user, and assistant prompts.

Learn to use semantic kernel for multi-modal chat by passing text and images to extract image tags, leveraging pre-configured chat history and system messages for image analysis.

Build an interactive chat app that writes text and processes images from URLs or local files to describe content and generate tags via prompts.

Create a chat history project by installing the semantic kernel NuGet package and inspector console, then configure program.cs to initialize kernels and manage chat history with a general system prompt.

Display a welcome message using Ansi console formatting and markup lines to show bold green text for the Semantic Kernel Chat, outline text input, the img command, and exit.

Build an interactive chat loop that reads user input, handles exit commands, updates chat history, and asynchronously queries the language model to generate and display assistant responses.

Add image reading to the chat system by prompting for a local path or URL, validating input, inferring mime types, and generating image content for description and prompts.

Discover the foundation of plugins in semantic kernel. Learn how kernel functions add features like image creation, local saving, information extraction, and external queries with automatic schema extraction.

Create kernel functions with semantic kernel by setting up a project, configuring OpenAI and Azure OpenAI, building a kernel, and defining time and poem functions from method and prompt.

Discover how semantic kernel plugins group kernel functions for reuse, expose methods as kernel functions with snake_case names, and use description attributes to guide parameter selection for native plugins.

Create a native plugin for semantic kernel by defining a systeminfo plugin with kernel functions to fetch system memory usage and top memory consuming processes, then invoke them with parameters.

Learn to access built-in semantic kernel plugins through NuGet, install the conversation summary plugin, and invoke it to summarize text. Explore additional plugins for http, file IO, time, and weight.

Define tools metadata and send them with messages to an LLM, letting the model select and call the best function, such as fetching weather, and incorporate results into the response.

Enable function calling in semantic kernel by wiring plugins with kernel functions, selecting simple metadata, using enums for parameters, and sending function schemas with chat history to orchestrate pizza orders.

Observe how to create and attach plugins to a semantic kernel, enable automatic function selection for tool calls, and run a chat workflow with Azure or OpenAI kernels.

Learn how to register plugins in semantic kernel using add from type and add from object, wiring dependencies like file service through dependency injection and the builder.

Explore function choice behavior in semantic kernel by configuring which kernel functions and plugins to expose to the AI model, controlling invocation with auto, required, or known modes.

Configure semantic kernel to auto invoke or manually execute tools with the auto invoke parameter, updating chat history and invoking function calls for tool execution.

Learn how to pass complex objects to the semantic kernel by defining a speaker model with descriptive properties and using add from object to supply a list of speakers.

Learn how to reuse real-world APIs as plugins in Semantic Kernel by importing an OpenAPI specification, configuring a namespace, and running tests to query endpoints.

Build a chat-type application with Semantic Kernel to transcribe videos with timestamps, query content, and generate social content like tweets or blog posts, including identifying viral moments and burning subtitles.

Set up a Visual Studio project with Semantic Kernel, install the Semantic Colonel NuGet package, configure Azure OpenAI chat and audio-to-text, and enable video transcript extraction and analysis.

Install ffmpeg by downloading the Windows build from ffmpeg.org, unzipping, and adding the bin folder to your path. Test by opening a terminal and running ffmpeg to confirm installation.

Learn to extract audio from a video file by building an FFmpeg-based utility, wiring it into a semantic kernel plugin, and handling file existence checks and output paths.

Show how to compress an audio file for transcription with an asynchronous ffmpeg-based reduce size method, converting to a smaller mp3 and validating a successful result.

Learn to implement the Whisper transcription service to convert audio to text, generate SRT transcripts, and integrate it into the Semantic Kernel via a speech-to-text plugin.

Learn to cut video clips with ffmpeg via semantic kernel, using start and end times from srt to generate ai-named micro videos.

Use semantic kernel to burn subtitles into trimmed videos with ffmpeg and SRT files. The workflow extracts audio, transcribes via whisper and a speech-to-text plugin, and outputs a subtitled video.

Craft clear prompts with rich context to guide ai responses. Explore zero-shot and few-shot prompting, examples, personas, and chain-of-thought techniques to shape outputs.

Define ai functions with the semantic kernel prompt template language. Install the semantic kernel package, build a kernel from a prompt, and invoke kernel functions to generate responses.

Convert prompts into chat history instances using semantic kernel by defining system and user messages, guiding tone, and building a chat history object for the AI response.

Learn to use variables in prompt templates by declaring an input variable, inserting ${variable} into prompts, and passing kernel arguments through invoke async to inject the person's name at runtime.

Learn how semantic kernel uses Handlebars prompt templates to combine templates with data, iterate over history, and produce dynamic responses via an Azure kernel function.

Explore Liquid prompt templates in Semantic Kernel, compare Liquid syntax with Handlebars, install the package, configure a Liquid template factory, and run the kernel with client data and history.

Learn to separate prompt templates into YAML files for clarity and versioning in semantic kernel projects, using embedded resources and create function from prompt YAML with topic and length parameters.

Learn to build a podcast generator project by converting an input document to markdown, extracting tone, host name, guest name, and language, and generating an MP3 conversation.

Configure the podcast generator project by installing the semantic kernel and inspector console, set up the kernel with chat completion and text to audio components, and use OpenAI.

Collects podcast project data by prompting the user for tone, host, guest, language, and an input document, using Ansi Console, Figlet titles, and a formatted table to display details.

Master document-to-markdown conversion using NuGet packages like Mark it Down sharp to produce clean markdown from PDFs, Word documents, and more, reducing noise for the AI model.

Create a first draft workflow using semantic kernel and YAML to generate podcast ideas from document content, execute the kernel function, and display results.

Generate a podcast script by creating a generate dialogue.yaml prompt, guiding a host–guest conversation that is informative, engaging, and structured for audience engagement.

Create a generate conclusion yaml file from the preconfigured template, using host name, guest name, language, and script data to generate the podcast closing as a purple conclusion panel.

Convert the podcast script and conclusion to speech using an OpenAI text-to-audio service, configure host and guest voices, synthesize lines, and save the final mp3 file.

Explore how embeddings in semantic kernel use vector stores to enable semantic concept searches and context-aware question answering with llms.

Define a data model for a semantic kernel vector store by creating a glossary class with key, category, term, and definition, and annotate properties with vector store attributes for search.

Learn to generate embeddings with semantic kernel, configure in-memory vector stores, and upsert glossary embeddings using OpenAI or Azure OpenAI services, including collection creation and retrieval.

Execute vectorized search asynchronously by embedding the query, retrieving semantically closest results from a vector store, and refine with filters, top/skip, and vector property options.

Install semantic kernel web, create an OpenAI kernel, and configure Bing text search to augment prompts with web results and asynchronous text search results.

Extend semantic kernel with text search plugins and Bing search; create and add a search plugin to the OpenAI kernel and build prompts to query results.

Enable semantic kernel to auto select the search method via function calling, fill in query parameters, and use Bing to retrieve text search results and quotes, all without prompt templates.

Leverage semantic kernel to perform text search with vector stores by extending the glossary model with text search attributes, and query in-memory vector stores using Azure OpenAI embeddings.