Copilot Sdk Setup

Copilot SDK Setup

An interactive skill that walks users through integrating the GitHub Copilot SDK into their projects. It handles both new project creation and modification of existing projects.

When to Use

• User wants to add the Copilot SDK to a project
• User says "set up copilot sdk" or "add copilot to my project"
• User wants to create an AI-powered application
• User asks about integrating Copilot programmatically
• User wants to build agents or tools using the Copilot SDK

Instructions

Step 1: Determine Project Type (REQUIRED FIRST STEP)

Ask the user:

Are you creating a new project or modifying an existing one?
1. New project - I want to create a fresh project with the Copilot SDK
2. Existing project - I want to add the Copilot SDK to my current codebase

Do not proceed until the user answers.

Step 2A: New Project Flow

If the user selected "New project":

2A.1: Ask for Language

Which language would you like to use?
1. TypeScript (Node.js) - npm install @github/copilot-sdk
2. Python - pip install github-copilot-sdk
3. Go - go get github.com/github/copilot-sdk/go
4. .NET (C#) - dotnet add package GitHub.Copilot.SDK

2A.2: Ask What They Want to Build

What would you like to build with the Copilot SDK?
Examples:
- A CLI assistant that answers questions
- A chatbot with custom tools
- An automated code reviewer
- A document analyzer
- A custom agent for specific tasks

Please describe your use case:

2A.3: Ask for Project Name

What would you like to name this project? (use kebab-case, e.g., my-copilot-app)

Validate the name:

• Only lowercase letters, numbers, and hyphens
• No spaces or special characters
• Not starting/ending with hyphen

2A.4: Initialize Project Based on Language

TypeScript:

mkdir <project-name> && cd <project-name>
npm init -y
npm pkg set type="module"
npm install @github/copilot-sdk tsx typescript @types/node
npx tsc --init --module NodeNext --moduleResolution NodeNext --target ES2022 --outDir dist --rootDir src
mkdir -p src

Python:

mkdir <project-name> && cd <project-name>
uv init --name <project-name>
uv add github-copilot-sdk
mkdir -p src/<project_name>
touch src/<project_name>/__init__.py

Or without uv:

mkdir <project-name> && cd <project-name>
python -m venv .venv
source .venv/bin/activate  # or .venv\Scripts\activate on Windows
pip install github-copilot-sdk
mkdir -p src

Go:

mkdir <project-name> && cd <project-name>
go mod init <project-name>
go get github.com/github/copilot-sdk/go

.NET:

dotnet new console -n <ProjectName> && cd <ProjectName>
dotnet add package GitHub.Copilot.SDK

Step 2B: Existing Project Flow

If the user selected "Existing project":

2B.1: Detect Project Language(s)

Examine the current directory for project files:

# Check for project markers
ls -la

Detection rules:

• package.json → TypeScript/JavaScript
• pyproject.toml, setup.py, requirements.txt → Python
• go.mod → Go
• *.csproj, *.sln → .NET
• Cargo.toml → Rust (not yet supported, inform user)

2B.2: Handle Multiple Languages

If multiple project types detected:

I detected multiple project types in this directory:
- package.json (TypeScript/JavaScript)
- pyproject.toml (Python)

Which language would you like to add the Copilot SDK to?
1. TypeScript/JavaScript
2. Python

2B.3: Ask About Use Case

What would you like to use the Copilot SDK for in this project?
Examples:
- Add an AI assistant feature
- Implement automated code analysis
- Create custom tools that Copilot can call
- Build an agent for a specific workflow

Please describe your use case:

2B.4: Install SDK for Detected Language

TypeScript/JavaScript:

npm install @github/copilot-sdk
# If TypeScript project, also:
npm install -D tsx typescript @types/node

Python:

# With uv:
uv add github-copilot-sdk

# With pip:
pip install github-copilot-sdk
# Add to requirements.txt if it exists:
echo "github-copilot-sdk" >> requirements.txt

Go:

go get github.com/github/copilot-sdk/go

.NET:

dotnet add package GitHub.Copilot.SDK

Step 3: Create Starter Code

Based on the user's language and use case, create appropriate starter code.

TypeScript Starter (src/index.ts):

import { CopilotClient, defineTool } from "@github/copilot-sdk";

// Example custom tool - modify based on user's use case
const exampleTool = defineTool("example_tool", {
    description: "Description of what this tool does",
    parameters: {
        type: "object",
        properties: {
            input: { type: "string", description: "Input parameter" },
        },
        required: ["input"],
    },
    handler: async (args: { input: string }) => {
        // Implement tool logic here
        return { result: `Processed: ${args.input}` };
    },
});

async function main() {
    const client = new CopilotClient();
    const session = await client.createSession({
        model: "gpt-4.1",
        streaming: true,
        tools: [exampleTool],
    });

    // Listen for response chunks
    session.on("assistant.message_delta", (event) => {
        process.stdout.write(event.data.deltaContent);
    });
    session.on("session.idle", () => {
        console.log();
    });

    // Send a message
    await session.sendAndWait({ prompt: "Hello! What can you help me with?" });

    await client.stop();
    process.exit(0);
}

main().catch(console.error);

Python Starter (src/main.py):

import asyncio
import sys
from copilot import CopilotClient
from copilot.tools import define_tool
from copilot.generated.session_events import SessionEventType
from pydantic import BaseModel, Field


class ExampleParams(BaseModel):
    """Parameters for the example tool."""
    input: str = Field(description="Input parameter")


@define_tool(description="Description of what this tool does")
async def example_tool(params: ExampleParams) -> dict:
    """Implement tool logic here."""
    return {"result": f"Processed: {params.input}"}


async def main():
    client = CopilotClient()
    await client.start()

    session = await client.create_session({
        "model": "gpt-4.1",
        "streaming": True,
        "tools": [example_tool],
    })

    def handle_event(event):
        if event.type == SessionEventType.ASSISTANT_MESSAGE_DELTA:
            sys.stdout.write(event.data.delta_content)
            sys.stdout.flush()
        if event.type == SessionEventType.SESSION_IDLE:
            print()

    session.on(handle_event)

    await session.send_and_wait({"prompt": "Hello! What can you help me with?"})

    await client.stop()


if __name__ == "__main__":
    asyncio.run(main())

Go Starter (main.go):

package main

import (
	"fmt"
	"log"
	"os"

	copilot "github.com/github/copilot-sdk/go"
)

// ExampleParams defines parameters for the example tool
type ExampleParams struct {
	Input string `json:"input" jsonschema:"Input parameter"`
}

// ExampleResult defines the result structure
type ExampleResult struct {
	Result string `json:"result"`
}

func main() {
	// Define a custom tool
	exampleTool := copilot.DefineTool(
		"example_tool",
		"Description of what this tool does",
		func(params ExampleParams, inv copilot.ToolInvocation) (ExampleResult, error) {
			// Implement tool logic here
			return ExampleResult{Result: fmt.Sprintf("Processed: %s", params.Input)}, nil
		},
	)

	client := copilot.NewClient(nil)
	if err := client.Start(); err != nil {
		log.Fatal(err)
	}
	defer client.Stop()

	session, err := client.CreateSession(&copilot.SessionConfig{
		Model:     "gpt-4.1",
		Streaming: true,
		Tools:     []copilot.Tool{exampleTool},
	})
	if err != nil {
		log.Fatal(err)
	}

	// Listen for response chunks
	session.On(func(event copilot.SessionEvent) {
		if event.Type == "assistant.message_delta" {
			fmt.Print(*event.Data.DeltaContent)
		}
		if event.Type == "session.idle" {
			fmt.Println()
		}
	})

	_, err = session.SendAndWait(copilot.MessageOptions{
		Prompt: "Hello! What can you help me with?",
	}, 0)
	if err != nil {
		log.Fatal(err)
	}
	os.Exit(0)
}

.NET Starter (Program.cs):

using GitHub.Copilot.SDK;
using Microsoft.Extensions.AI;
using System.ComponentModel;

// Define a custom tool
var exampleTool = AIFunctionFactory.Create(
    ([Description("Input parameter")] string input) =>
    {
        // Implement tool logic here
        return new { result = $"Processed: {input}" };
    },
    "example_tool",
    "Description of what this tool does"
);

await using var client = new CopilotClient();
await using var session = await client.CreateSessionAsync(new SessionConfig
{
    Model = "gpt-4.1",
    Streaming = true,
    Tools = [exampleTool],
});

// Listen for response chunks
session.On(ev =>
{
    if (ev is AssistantMessageDeltaEvent deltaEvent)
    {
        Console.Write(deltaEvent.Data.DeltaContent);
    }
    if (ev is SessionIdleEvent)
    {
        Console.WriteLine();
    }
});

await session.SendAndWaitAsync(new MessageOptions
{
    Prompt = "Hello! What can you help me with?",
});

Step 4: Customize Based on Use Case

Modify the starter code based on the user's described use case:

For CLI assistants:

• Add readline/input handling for interactive mode
• Implement conversation history

For chatbots with custom tools:

• Define tools specific to the user's domain
• Add appropriate parameter schemas

For code reviewers:

• Add file reading tools
• Implement diff analysis
• Connect to Git operations

For document analyzers:

• Add file/document reading capabilities
• Implement parsing tools for specific formats

Step 5: Add Run Scripts

TypeScript (package.json):

{
  "scripts": {
    "start": "tsx src/index.ts",
    "build": "tsc",
    "dev": "tsx watch src/index.ts"
  }
}

Python (pyproject.toml):

[project.scripts]
app = "src.main:main"

Go: Already runnable with go run main.go

.NET: Already runnable with dotnet run

Step 6: Verify Setup

Run a test to ensure everything works:

TypeScript:

npm start
# or
npx tsx src/index.ts

Python:

python src/main.py
# or with uv:
uv run python src/main.py

Go:

go run main.go

.NET:

dotnet run

Step 7: Summary

Report to the user:

✅ Copilot SDK setup complete!

📦 Installed:
- GitHub Copilot SDK for <language>
- Starter code with example tool

🚀 Quick start:
  <run command>

📚 Next steps:
1. Modify the example tool to match your use case
2. Add more tools as needed
3. Customize the system message and model
4. Check the SDK documentation: https://github.com/github/copilot-sdk

💡 Key concepts:
- Tools let Copilot call your code
- Streaming shows responses in real-time
- Sessions maintain conversation context
- Custom agents can have specialized behavior

Output Format

Always provide:

1. Confirmation of choices made by the user
2. Commands executed with their output
3. Files created with brief descriptions
4. Run instructions for testing
5. Next steps for customization

SDK Quick Reference

Installation Commands

Key Features

• Custom Tools: Define functions that Copilot can call
• Streaming: Real-time response output
• Sessions: Maintain conversation context
• MCP Servers: Connect to external tool providers
• Custom Agents: Create specialized AI personas

Resources

• GitHub Copilot SDK Repository
• Getting Started Guide
• Cookbook Examples
• MCP Documentation

Reference Documentation

When you need more details about SDK usage, fetch documentation from the github/copilot-sdk repository:

Core Documentation Files

Language-Specific READMEs

Cookbook Recipes (by language)

Each language has practical recipes in cookbook/<language>/:

• error-handling.md - Handle connection failures, timeouts, cleanup
• multiple-sessions.md - Manage multiple conversations simultaneously
• managing-local-files.md - Organize files using AI-powered grouping
• pr-visualization.md - Generate PR charts using GitHub MCP Server
• persisting-sessions.md - Save and resume sessions across restarts

How to Fetch Documentation

Use the GitHub MCP tools or gh CLI to fetch the latest docs:

# Fetch getting started guide
gh api repos/github/copilot-sdk/contents/docs/getting-started.md -H "Accept: application/vnd.github.raw"

# Fetch language-specific README
gh api repos/github/copilot-sdk/contents/nodejs/README.md -H "Accept: application/vnd.github.raw"

# Fetch a cookbook recipe
gh api repos/github/copilot-sdk/contents/cookbook/nodejs/error-handling.md -H "Accept: application/vnd.github.raw"

Or use the github-mcp-server-get_file_contents tool:

• owner: github
• repo: copilot-sdk
• path: docs/getting-started.md (or other paths above)

Error Handling

Common Issues

"copilot command not found":

The Copilot CLI is required. Please install it first:
https://docs.github.com/en/copilot/how-tos/set-up/install-copilot-cli

Authentication errors:

Please authenticate with the Copilot CLI:
copilot auth login

Package installation failures:

For TypeScript: Ensure Node.js 18+ is installed
For Python: Ensure Python 3.8+ is installed
For Go: Ensure Go 1.21+ is installed
For .NET: Ensure .NET 8.0+ SDK is installed

Constraints

• Always verify the Copilot CLI is installed before proceeding
• Use the latest stable SDK versions
• Create minimal, focused starter code
• Let users customize based on their needs
• Do not include secrets or API keys in generated code

Source: danielmeppiel/meeting-2-code — distributed by TomeVault.