Skip to content

Going through the workshop

We’ll build the interview coach in small, runnable steps. Each chapter explains the next capability and the reason for adding it before we edit the code.

Chapter 0 checks the tools and model access by running the finished app. Chapter 1 starts our own copy. We’ll keep that working folder through the remaining lessons.

Chapter 13 practices failure recovery and compares repeated runs. Chapter 14 connects the finished application to the decisions we made.

Start Chapter 0: Get ready and try the interview coach

The starter supplies the Blazor chat interface, Aspire helpers, and database repository. We’ll write the agent instructions, tool connections, and handoff rules. Run the course’s build and Aspire commands from your working folder’s root.

Each code step names the file and shows what to add, replace, or remove. Apply the steps in order. Finish the whole edit group before building because some intermediate edits do not compile alone.

Required explanations appear beside the work. Expandable notes cover optional internals or a problem you may encounter.

After building the app, try the chapter’s input and look for its expected result. Early on, that’s a reply to your question. Later, we’ll look for a tool invocation, a saved record, or a handoff.

Some lessons ask you to write an instruction or choose an expected result before revealing the explanation. Try the question first. Getting it wrong can expose a useful distinction, such as a saved record versus a visible chat message.

Temporary experiments include restoration instructions. Return to the chapter’s baseline before continuing so the next code replacement still matches.

Mark a chapter complete after its run check and learning activities. Read Chapter 14’s synthesis before marking the summary complete. Quiz answers are not stored or used to block progress.

Completion marks appear in the sidebar and chapter list. The landing page continues at your next unfinished chapter. This browser remembers your progress. Earlier curriculum records remain saved separately because this revision adds new practice.

Each build chapter links to a working checkpoint. Extract one into a new folder and keep your existing work alongside it. The starter is cloud-free. Later checkpoints need their own configuration to reuse the approved model.

Samples and checkpoints are available throughout the build. Use the reference documentation for configuration and runtime details. Deployment is optional. The core course runs the local application with Foundry.

Getting started

  1. Get ready and try the interview coach Completed

    Set up your tools, run the finished app, and try a short interview.

  2. Start your workshop app Completed

    Open your own working folder and run the supplied application shell.

  3. Ask your first agent a question Completed

    Create a ChatClientAgent, give it instructions, and get a reply.

  4. Put the coach in the chat UI Completed

    Wire the agent into the supplied Blazor app and stream its replies.

Tools and interview context

  1. Give the coach a C# tool Completed

    Write a function, expose it as a tool, and inspect a real invocation.

  2. Expose interview tools with MCP Completed

    Start InterviewData and discover the tools its server provides.

  3. Connect the coach to those tools Completed

    Let the coach create and fetch a synthetic interview record.

  4. Save interview progress Completed

    Give the agent a session lifecycle and inspect the records it saves.

  5. Read a resume with a tool Completed

    Connect MarkItDown and inspect the text extracted from a sample resume.

  6. Use the resume in the interview Completed

    Connect document extraction and use a resume and job description as context.

Specialist workflows

  1. Make your first handoff Completed

    Watch triage transfer document intake to the receptionist.

  2. Add the interviewers Completed

    Connect behavioural and technical practice using the saved interview context.

  3. Finish with a summary agent Completed

    Complete the five-role workflow and save its final feedback.

  4. Find and fix a failed step Completed

    Recover a failed request and compare repeated runs against expected behavior.

  5. Review what you built Completed

    Connect the workshop concepts to their code and the problems they solve.