Inline Completions vs. Copilot Chat

Switch between them as the task changes. Each gives you different context.

The Four Chat Surfaces

Surface	Shortcut	Best For	History
Sidebar	Ctrl+Alt+I	Extended conversations, @workspace	Preserved
Inline	Ctrl+I (in editor)	Code-specific edits, explanations	Ephemeral
Quick Chat	Ctrl+Shift+Alt+L	Fast one-off questions	Dismissed
Terminal	Ctrl+I (in terminal)	CLI help, error debugging	Ephemeral

Use the sidebar for a longer thread, inline chat for a focused edit, quick chat for a one-off question, and terminal chat for command output.

Sidebar Chat: Your Primary Surface

• Full-height panel with messaging-style interface
• Preserves conversation history within the session
• Supports multiple threads (click + for a new thread)
• Best for longer conversations, @workspace queries

Try it:

What does this project do?

Then follow up:

What programming languages are used?

Context carries over between messages.

Inline Chat: Focused Code Edits

• Select code → Ctrl+I → type your request
• Result appears as a diff right in the editor
• Accept, reject, or modify without leaving the editor

Try it:

1. Select a function
2. Press Ctrl+I
3. Type: add input validation
4. Review the inline diff

Inline Chat is context-anchored. Copilot sees exactly what you selected.

Chat Participants

Point Copilot at specific knowledge sources with @:

@workspace How is error handling implemented in this project?

Use @workspace for a project question. For a specific file, name that file explicitly.

Slash Commands

• Combine with selections for precise results
• Use /fix + error message for rapid debugging

Context References

Tell Copilot exactly what to look at with # references:

Based on the patterns in #file:services/user_service.py, 
create a new service method for order cancellation.

Name the context that matters. It reduces guesswork.

Model Selection & Cost

• Click the model picker in the Chat panel to switch models
• Auto lets Copilot choose a model based on the task
• Model costs are usage-based: token consumption × per-model rate

Start on Auto. Switch when you need a specific model's strengths.
Available models and pricing change. Verify live docs: supported models and models and pricing.

Conversational Workflows

Build on the previous response when it has the right direction:

1. Ask → Get initial code
2. Refine → "Make it async" or "Add error handling"
3. Constrain → "Use only the standard library"
4. Validate → "Does this handle empty inputs?"

Write a function to parse CSV data.
→ Make it handle quoted fields with commas inside.
→ Add type hints and a docstring.
→ Now write tests for the edge cases we discussed.

Keep the useful parts and ask for the next specific change.

Copilot in the Terminal

GitHub Copilot includes a standalone CLI tool for terminal assistance — shell command explanation, generation, and interactive help.

The current standard is the standalone copilot CLI binary (npm install -g @github/copilot or via your package manager).

The CLI is covered fully in Session 04 — GitHub Copilot in the CLI, including:

• Interactive command explanation and suggestion
• Shell integration and context-aware completions

Skip the CLI demo here. Point trainees to Session 04 for hands-on practice.

Lab Overview

Deliverable: A debugged and tested module, plus documented model comparison observations.

Key Takeaways

1. Four Chat surfaces — Sidebar, Inline, Quick Chat, Terminal — each for a different context
2. @workspace searches your entire project — use it for cross-file questions
3. Slash commands (/explain, /fix, /tests, /doc) accelerate common tasks
4. Context references (#file, #selection) give Copilot precise input
5. Model selection lets you match model strengths to your task
6. Iterative conversations work better than one-shot prompts. Refine instead of restarting.