Continue

Use Radium with the Continue VS Code extension

Continue

Continue is an open-source VS Code extension for inline AI code assistance. Configure it to use Radium as the backend.

Prerequisites

• Continue extension installed in VS Code
• A Radium API key from the dashboard

Setup

Add Radium to your Continue config.json:

1. Open Continue panel in VS Code
2. Click the gear icon → Open config.json
3. Add the following model config:

{
  "models": [
    {
      "title": "Radium Hal",
      "provider": "openai",
      "model": "hal-1.0",
      "apiBase": "https://api.radium.cloud/v1",
      "apiKey": "YOUR_RADIUM_API_KEY"
    },
    {
      "title": "Radium Clarke",
      "provider": "openai",
      "model": "clarke-1.0",
      "apiBase": "https://api.radium.cloud/v1",
      "apiKey": "YOUR_RADIUM_API_KEY"
    },
    {
      "title": "Radium Tycho",
      "provider": "openai",
      "model": "tycho-1.0",
      "apiBase": "https://api.radium.cloud/v1",
      "apiKey": "YOUR_RADIUM_API_KEY"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Radium Tycho",
    "provider": "openai",
    "model": "tycho-1.0",
    "apiBase": "https://api.radium.cloud/v1",
    "apiKey": "YOUR_RADIUM_API_KEY"
  }
}

Use environment variables for the API key instead of hardcoding:

"apiKey": "${RADIUM_API_KEY}"

Usage

Chat

Open the Continue chat panel (Cmd/Ctrl + L) and select a Radium model from the dropdown. Ask questions about your code, request refactors, or generate new functions.

Tab autocomplete

Continue uses the tabAutocompleteModel for inline suggestions. We recommend tycho-1.0 for this because of its low latency and high throughput.

Inline edits

Highlight code and press Cmd/Ctrl + I to request an inline edit. The selected Radium model will suggest a replacement.

Recommended models

Feature	Recommended model	Why
Chat / Q&A	hal-1.0	Best reasoning and code understanding
Code review	clarke-1.0	Strong on correctness and edge cases
Tab autocomplete	tycho-1.0	Fastest, lowest TTFT
Inline edits	hal-1.0	Best at preserving code style

Context length

Continue sends the current file and recent edits as context. Radium models support up to 999999 tokens, so even large files fit comfortably.

Rate limits

Continue makes frequent small requests. If you experience throttling:

1. Switch to tycho-1.0 for autocomplete
2. Reduce the number of context files sent with each request
3. Check your current limits at Rate limits

Troubleshooting

Symptom	Fix
”Connection error”	Verify apiBase is https://api.radium.cloud/v1 (note the /v1)
“Model not found”	Use exact model IDs: hal-1.0, clarke-1.0, tycho-1.0
Slow autocomplete	Switch tabAutocompleteModel to tycho-1.0
No suggestions	Ensure the file is saved; Continue works best with saved files

Next steps

Quickstart

Your first Radium request

Models

Compare model capabilities

Cline

Cline setup guide

Rate limits

Current limits by tier