Cline

Use Radium with the Cline VS Code extension

Cline

Cline is a VS Code extension for autonomous coding. Point it at Radium for AI-powered coding assistance.

Prerequisites

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

Setup

OpenAI-compatible provider

Cline supports custom OpenAI-compatible providers. Add Radium in your Cline settings:

1. Open VS Code → Settings → Extensions → Cline
2. Under API Provider, select OpenAI Compatible
3. Set:
   • Base URL: https://api.radium.cloud/v1
   • API Key: your Radium API key
   • Model ID: hal-1.0
4. Save and restart the Cline panel

Alternative: .clinerules file

Create a .clinerules file in your project root:

{
  "provider": "openai-compatible",
  "baseUrl": "https://api.radium.cloud/v1",
  "apiKey": "YOUR_RADIUM_API_KEY",
  "model": "hal-1.0",
  "maxTokens": <span style="color:red">99999</span>,
  "temperature": 0.7
}

Do not commit .clinerules with your API key. Use an env var reference or the VS Code settings UI instead.

Recommended models

Task	Model	Context
Code generation	hal-1.0	Up to 999999 tokens
Code review	clarke-1.0	Up to 999999 tokens
Quick fixes / autocomplete	tycho-1.0	Up to 999999 tokens

System prompt

Cline sends its own system prompt for autonomous operation. You can prepend Radium-specific instructions:

You are an expert software engineer. Use Radium's hal-1.0 model for reasoning.
- Prefer TypeScript for new code
- Add JSDoc comments to all public functions
- Run tests before declaring a task complete

Rate limits

Cline can make many requests in rapid succession during an agent loop. If you hit rate limits:

1. Reduce the max requests per task in Cline settings
2. Switch to tycho-1.0 for faster throughput
3. Consider upgrading your Radium tier for higher RPM

Current limits: see Rate limits.

Streaming

Cline supports streaming responses. Ensure streaming is enabled in Cline settings for the lowest perceived latency.

Troubleshooting

Symptom	Cause	Fix
”Invalid API key”	Key not copied correctly	Regenerate in dashboard
”Model not found”	Wrong model ID	Use hal-1.0, clarke-1.0, or tycho-1.0
Slow responses	Model overloaded	Switch to tycho-1.0 or retry
Tool calls fail	Function spec mismatch	Verify JSON schema matches Radium format

Next steps

Tool calling

How function calling works in Radium

Rate limits

Current RPM and TPM limits

Models

Compare hal-1.0, clarke-1.0 and tycho-1.0

Quickstart

Your first Radium request