Offline Assist

Offline Assist is a standalone VS Code extension that exposes Foundry Local chat-capable models through VS Code's language model picker.

What it does:

• Lists chat-capable models from the local Foundry catalog.
• Registers them as a languageModelChatProvider vendor named offline-assist.
• Downloads and loads a selected model lazily the first time you send a chat request.
• Adds an OfflineAssist: Download Chat Model command so you can pre-download a model before chatting.

Requirements

• VS Code Insiders.
• Proposed API access for chatProvider@4.
• Foundry Local installed on your machine.
• Node.js 18+.

Install Foundry Local on macOS:

brew install microsoft/foundrylocal/foundrylocal

Run

cd Offline-Assist
npm install
npm run compile

Then open this folder in VS Code Insiders and launch an Extension Development Host.

Use the Extension

1. Open the chat view in the Extension Development Host.
2. Open the model picker.
3. Choose a model under Offline Assist.
4. Send a prompt. The extension will download execution providers if needed, then download and load the model on first use.

Optional command:

• OfflineAssist: Download Chat Model pre-downloads a model from a quick pick.

Notes

• This project intentionally handles text chat only.
• It does not yet bridge VS Code tool-calling to Foundry Local tool-calling.
• Non-chat models such as transcription and embeddings are filtered out of the picker.