Windows Copilot Runtime Server

A REST API server that provides OpenAI-compatible endpoints for Windows Copilot+ PC, allowing you to use local AI models through standard OpenAI client libraries.

Quick Installation

1. Download the latest release:

   • WindowsCopilotRuntimeServer.zip

2. Extract the ZIP file to your preferred location

3. Run PowerShell as Administrator and execute:

cd <extraction-path>
.\Install.ps1

4. Start the server from Start Menu and search for "Windows Copilot Runtime Server"

Prerequisites

• Windows 11 Insider Preview Build 26120.3073 (Dev and Beta Channels) or later must be installed on your device.

For Developers

If you're planning to build the project from source:

• MSIX Packaging Tool is required for packaging
• Build and use the WindowsCopilotRuntimeServer (Package) project instead of the non-packaged version due to experimental Windows App SDK dependencies

API Support Status

Feature	Status	Notes
Chat Completions	✅	Full support with OpenAI compatibility
Tool Calling	✅	Supports OpenAI function calling format
Vision	❌	Planned
Streaming	✅	Supports SSE with token-by-token streaming
Custom Model Loading	❌	Not supported by runtime
System Messages	✅	Supported in chat context
Temperature	✅	Range 0-1
Top P	✅	Range 0-1
Top K	✅	Windows Copilot specific parameter
Max Tokens	❌	Not supported by runtime
Stop Sequences	❌	Not supported by runtime

Usage with OpenAI Clients

Python with OpenAI Client

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:5001/v1",
    api_key="not-needed"  # required by client but not used
)

chat_completion = client.chat.completions.create(
    model="PhiSlica",  # or use "windows-local"
    messages=[{
        "role": "user",
        "content": "What is the golden ratio?"
    }],
    temperature=0.7,
    top_p=0.9,
    top_k=40
)

print(chat_completion.choices[0].message.content)

PowerShell

# Simple request using Invoke-RestMethod
$headers = @{
    "Content-Type" = "application/json"
}

$body = @{
    model = "PhiSlica"
    messages = @(
        @{
            role = "user"
            content = "What is the golden ratio?"
        }
    )
    temperature = 0.7
    top_p = 0.9
    top_k = 40
} | ConvertTo-Json

$response = Invoke-RestMethod -Method Post `
    -Uri "http://localhost:5001/v1/chat/completions" `
    -Headers $headers `
    -Body $body

$response.choices[0].message.content

# Stream response using curl
curl -X POST "http://localhost:5001/v1/chat/completions" `
     -H "Content-Type: application/json" `
     -d '{
           "model": "PhiSlica",
           "messages": [{"role": "user", "content": "What is the golden ratio?"}],
           "temperature": 0.7,
           "top_p": 0.9,
           "top_k": 40
         }'

Streaming Support

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:5001/v1",
    api_key="not-needed"
)

stream = client.chat.completions.create(
    model="PhiSlica",
    messages=[{"role": "user", "content": "Write a story about a cat"}],
    stream=True  # Enable streaming
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="")

Using curl:

curl http://localhost:5001/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "PhiSlica",
    "messages": [{"role": "user", "content": "Write a story about a cat"}],
    "stream": true
  }'

API Reference

Chat Completions

POST /v1/chat/completions

Request body:

{
    "model": "PhiSlica",
    "messages": [
        {
            "role": "user",
            "content": "Your prompt here"
        }
    ],
    "temperature": 0.9,
    "top_p": 0.9,
    "top_k": 40
}

Features

• OpenAI-compatible REST API
• Local model inference using Windows Copilot Runtime
• CORS support for web applications
• Swagger UI at /swagger
• JSON source generation for optimal performance

Troubleshooting

If you encounter any issues:

1. Make sure Windows App SDK 1.7-experimental3 is installed
2. Check if Windows Copilot Runtime is properly installed
3. Run PowerShell as Administrator when executing Install.ps1
4. Check the logs in the application directory

License

This project is licensed under the MIT License.