VS Code Extension Project
BBA LocalLLM Bridge
Expose local models from Ollama or LM Studio in the VS Code chat model picker, so you can work with local AI models inside the editor without changing your flow.
This project is built around a focused bridge: discover local models automatically, publish them into chat, and forward requests to the selected model with streaming responses.
Local-first chat models · Ollama support · LM Studio support · Streaming responses · Tool-calling passthrough
A practical bridge between local models and VS Code chat
BBA LocalLLM Bridge is designed for developers who want to use local AI models without leaving the VS Code chat experience. Instead of creating a separate assistant interface, the extension publishes local models directly into the model picker.
That keeps the workflow simple: start Ollama or LM Studio, open VS Code chat, choose a local model, and keep working inside the editor.
The project focuses on the essentials that matter most in a real developer setup: model discovery, chat streaming, local server integration, and a configuration layer that lets you decide which models should appear in the picker.
Main features
Automatic model discovery
The extension discovers models from Ollama through /api/tags and from LM Studio through /v1/models, then publishes them into the chat picker.
Streaming chat responses
Messages are forwarded to the selected local model with streaming responses, which keeps the VS Code chat experience responsive and familiar.
Tool-calling passthrough
The bridge supports tool-calling passthrough for models that can handle tools well, so the model can participate in agent-style flows.
Model visibility controls
You can restrict which models appear with settings, including visibility modes such as balanced, strict, and all-chat-models.
Local server integration
The extension is built around local endpoints, with configurable base URLs and model lists for both Ollama and LM Studio.
Packaging-ready scaffold
The project ships as a TypeScript extension scaffold built for compilation, packaging, and editor debugging.
How it works
1. Start a local model server
Run Ollama or LM Studio locally so the extension can discover available models from your machine.
2. Open VS Code chat
The extension registers its provider so the local models appear in the chat model picker inside VS Code.
3. Pick a local model
Select a model from BBA LocalLLM Bridge and send chat requests directly to that local runtime.
4. Tune visibility and tool support
Use settings to choose which models appear and how aggressively the extension should expose chat-capable models.
5. Keep the workflow local
The goal is to keep your AI workflow inside your editor while relying on models you run and control locally.
6. Debug and package like a normal extension
The scaffold includes the usual extension development pieces, so it can be built, debugged, and packaged in a familiar way.
Settings and visibility modes
The extension includes a settings surface that gives you control over the local providers and which models show up in chat.
Connection settings
bbaLocalLlmBridge.ollama.baseUrlbbaLocalLlmBridge.ollama.modelsbbaLocalLlmBridge.ollama.toolCallingModelsbbaLocalLlmBridge.lmStudio.baseUrlbbaLocalLlmBridge.lmStudio.apiKeybbaLocalLlmBridge.lmStudio.modelsbbaLocalLlmBridge.lmStudio.toolCallingModels
Visibility modes
- Balanced: recommended for most users
- Strict: only shows models with explicit or reliable tool support
- All-chat-models: shows every non-embedding model in chat
Project limits
- No image support yet
- No embeddings, completions, or rerank endpoints yet
- Uses naive token counting for budgeting
- Focused on local chat models, not a full assistant stack
Why this project matters
For many developers, the hard part is not running a model locally. The hard part is making the local model feel native inside the editor where real work happens.
This project solves that handoff. It gives local AI models a place in the existing VS Code chat workflow, which makes it easier to test local models, compare providers, and keep sensitive development work closer to your machine.
It also makes a strong portfolio piece because it shows product thinking, integration work, and a clear technical boundary: bridge local runtimes into an editor-native experience without overbuilding the assistant layer.
FAQ
Does this extension send my prompts to a third-party API?
No. The extension is designed to work with local providers such as Ollama and LM Studio that run on your machine or your local network.
What makes this different from a full AI assistant?
This project is intentionally narrower. It acts as a bridge into VS Code chat instead of trying to replace the editor’s AI workflow with a separate assistant stack.
Can I use both Ollama and LM Studio?
Yes. The project is built to discover models from both providers and present them in one place.
Is the extension meant for chat only?
Yes. The current scope is focused on local text chat streaming and tool-calling passthrough, not embeddings, image handling, or other model endpoint types.
What should I mention when presenting this project to clients?
Position it as a local-first developer productivity tool that integrates local AI models into a familiar editor workflow and gives teams a controlled way to test model-based chat locally.
Bring local AI into the editor workflow
BBA LocalLLM Bridge helps VS Code users work with Ollama and LM Studio models in one familiar place.
Project URL: https://builtbyamit.com/projects/local-llm-bridge/