microsoft-foundry-local

Minimal example project showing how to use Foundry Local from Python via the foundry-local-sdk, and then send requests to the locally-hosted model using the OpenAI Python client.

Install Microsoft Foundry Local

Official docs: https://learn.microsoft.com/en-us/azure/foundry-local/get-started

Windows (WinGet)

winget install Microsoft.FoundryLocal

macOS (Homebrew, Apple Silicon)

brew tap microsoft/foundrylocal
brew install foundrylocal

Verify installation

foundry --version

Repository contents

• main.py — Downloads and loads a model with FoundryLocalManager, then sends a sample chat completion request to the local Foundry endpoint using openai.OpenAI.
• requirements.txt — Python dependencies for running the sample (foundry-local-sdk, openai).
• LICENSE — License information.
• .gitignore — Git ignore rules for common Python artifacts.

Prerequisites

• Python 3.9+

Setup

python -m venv .venv
# Windows
.\.venv\Scripts\activate
# macOS/Linux
# source .venv/bin/activate

pip install -r requirements.txt

Run

python main.py

The script will:

1. Download the model alias qwen2.5-0.5b (if needed)
2. Load the model
3. Create an OpenAI client pointing at the local Foundry endpoint
4. Send a sample chat completion request
5. Unload the model

Notes

• To use a different model, change model_alias in main.py.