BBA LocalLLM Bridge

Expose local models from Ollama or LM Studio in the VS Code chat model picker.

What this extension does

This extension registers a LanguageModelChatProvider named BBA LocalLLM Bridge. When VS Code asks the provider for available models, it:

• discovers models from Ollama via /api/tags
• discovers models from LM Studio via /v1/models
• publishes those models into the chat model picker
• forwards chat requests to the selected local model with streaming responses

Current scope

This is a focused bridge for chat models, not a full assistant stack. It currently:

• supports local text chat streaming
• discovers models automatically
• supports tool-calling passthrough for tool-capable models
• lets you restrict which models appear with settings
• uses naive token counting for VS Code budgeting
• includes a packaging-ready TypeScript extension scaffold

It does not yet:

• support images
• estimate tokens with model-specific tokenizers
• expose embeddings, completions, or rerank endpoints

Extension settings

Open Settings and search for BBA LocalLLM Bridge.

Important settings:

• bbaLocalLlmBridge.chatModelVisibility
• bbaLocalLlmBridge.ollama.baseUrl
• bbaLocalLlmBridge.ollama.models
• bbaLocalLlmBridge.ollama.toolCallingModels
• bbaLocalLlmBridge.lmStudio.baseUrl
• bbaLocalLlmBridge.lmStudio.apiKey
• bbaLocalLlmBridge.lmStudio.models
• bbaLocalLlmBridge.lmStudio.toolCallingModels

Tool Calling

Tool calling support depends on the model you select.

• If the selected model supports tool use, VS Code can use it in agent-style flows that require tools.
• If the selected model does not support tool use well, it may still appear for normal chat but not behave reliably for tool-based tasks.
• LM Studio can expose tool-use capability metadata for some models.
• This extension can also be configured to show likely chat models more aggressively in the Chat picker, while retrying normal chat requests without tools if a backend rejects tool payloads.

If you want strict control, set these allowlists:

• bbaLocalLlmBridge.ollama.toolCallingModels
• bbaLocalLlmBridge.lmStudio.toolCallingModels

For easier setup, use:

• BBA LocalLLM Bridge: Configure Chat Visibility

Visibility modes:

• balanced: recommended for most users, shows likely chat models while keeping a safe retry path
• strict: only advertises tool support from metadata or explicit allowlists
• all-chat-models: shows every non-embedding model in Chat

Project structure

• src/extension.ts contains the provider implementation
• images/icon.png is the marketplace-safe extension icon
• tsconfig.json compiles to out/
• .vscode/launch.json and .vscode/tasks.json support F5 debugging

Typical local setup

Ollama

1. Start Ollama.
2. Ensure http://127.0.0.1:11434/api/tags works locally.
3. Open VS Code chat and manage models.
4. Select a model from BBA LocalLLM Bridge.

LM Studio

1. Start LM Studio local server.
2. Ensure http://127.0.0.1:1234/v1/models works locally.
3. Open VS Code chat and manage models.
4. Select a model from BBA LocalLLM Bridge.

Commands

• BBA LocalLLM Bridge: Manage Provider
• BBA LocalLLM Bridge: Refresh Models

Development

1. Run npm install
2. Open this folder in VS Code
3. Press F5 to launch the Extension Development Host
4. Open Chat and use the model picker

Useful commands:

• npm run compile
• npm run lint
• npm run package

Notes

The provider API used here is documented by VS Code:

• Language Model Chat Provider API
• AI language models in VS Code