Skip to main content

VS Code

VS Code integration

Author: Aleksander Lovric

Integrate your editor with locally deployed language models using Loc.ai and the Continue extension for VS Code.

What is Continue?​

Continue is an open-source AI assistant for VS Code and JetBrains. It supports chat, explanation, refactoring, and Tab Autocomplete on your hardware.

Serving models via Loc.ai lets you run inference on a dedicated edge device while you code on your main machine.

Prerequisites​

1. Connect your device​

Register your device with Loc.ai. See the device registration guide.

  1. Open the Devices tab in the Loc.ai dashboard
  2. Click Register Device (or Add Device) to open the registration flow and generate your registration command
Q2: Register Device dialog with device name, TTL hours, and Generate Registration Key

Q2. Register Device popup – enter device name, TTL, and generate key (same as Quickstart)

Q3: Generated one-line curl install command with copy affordance

Q3. Generated one-line installation command ready to copy (same as Quickstart)

Paste the provided curl command on the device terminal. When registered, the device shows Online.

2. Deploy and serve​

Choose a GGUF language model — see Supported models.

  1. Upload a model or pick one from ML MODELS (e.g. Qwen2.5-Coder-1.5B Q4_K_M.gguf) with type Language Model
  2. Deploy to your registered device (same deployment dialog as in the Quickstart)
  3. After deploy, click Start Serving and note the IP address and port (see Serving a model)
Q4: Deploy Model dialog with target device selection and deployment summary

Q4. Deploy Model dialog – select target device and review deployment summary (same as Quickstart)

info

Note the IP address and port from the serving panel for Continue configuration. See also Serving a model.

3. Configuration​

In VS Code, open Continue → gear → config.yaml and add:

config.yaml
version: 0.0.1

models:
  - name: chat_model
    provider: llama.cpp
    model: <YOUR_MODEL_NAME>
    apiBase: http://<YOUR_DEVICE_IP>:<PORT>/

  - name: autocomplete_model
    provider: llama.cpp
    model: <YOUR_MODEL_NAME>
    apiBase: http://<YOUR_DEVICE_IP>:<PORT>/
    roles:
      - autocomplete
tip

Replace <YOUR_DEVICE_IP> and <PORT> with values from the serving panel.

4. Usage​

  • Chat: Select your Loc.ai model in the Continue sidebar
  • Autocomplete: Grey inline suggestions; Tab to accept; Ctrl + . for inline actions
VS Code editor with Continue inline prompt: Generate code and Add Context above a Python function
Inline AI prompt in the editor (Generate code, Add Context)
tip

Multiple models on different ports can be configured as separate providers. See also OpenClaw.