Aider

Use Radium with Aider for AI pair programming

Aider

Aider is a command-line tool for AI pair programming. Use Radium as the backend for aider’s coding assistance.

Prerequisites

• Aider installed (pip install aider-install)
• A Radium API key from the dashboard

Setup

Via environment variables

export RADIUM_API_KEY="YOUR_RADIUM_API_KEY"
export OPENAI_API_BASE="https://api.radium.cloud/v1"

aider --model openai/hal-1.0

Via .aider.model.settings.yml

Create .aider.model.settings.yml in your project root:

- name: radium-hal
  edit_format: diff
  weak_model_name: radium-tycho
  use_repo_map: true
  reminder: sys
  examples_as_sys_msg: true
  extra_params:
    model: hal-1.0
    api_base: https://api.radium.cloud/v1
    api_key: YOUR_RADIUM_API_KEY
  caches_by_default: true

- name: radium-clarke
  edit_format: diff
  weak_model_name: radium-tycho
  use_repo_map: true
  reminder: sys
  examples_as_sys_msg: true
  extra_params:
    model: clarke-1.0
    api_base: https://api.radium.cloud/v1
    api_key: YOUR_RADIUM_API_KEY
  caches_by_default: true

- name: radium-tycho
  edit_format: whole
  weak_model_name: radium-tycho
  use_repo_map: false
  reminder: user
  examples_as_sys_msg: false
  extra_params:
    model: tycho-1.0
    api_base: https://api.radium.cloud/v1
    api_key: YOUR_RADIUM_API_KEY

Then run:

aider --model radium-hal

Using a .env file

# .env
RADIUM_API_KEY=YOUR_RADIUM_API_KEY
OPENAI_API_BASE=https://api.radium.cloud/v1

aider --model openai/hal-1.0 --env-file .env

Recommended models

Mode	Model	Why
Main editor	hal-1.0	Best reasoning, diff editing
Commits / linting	tycho-1.0	Fast, cheap for small tasks
Code review	clarke-1.0	Strong correctness analysis

Model aliases

Create convenient aliases in your shell profile:

alias aider-hal='aider --model openai/hal-1.0 --api-key $RADIUM_API_KEY'
alias aider-clarke='aider --model openai/clarke-1.0 --api-key $RADIUM_API_KEY'
alias aider-tycho='aider --model openai/tycho-1.0 --api-key $RADIUM_API_KEY'

Context windows

Aider sends the entire repo map and relevant files as context. Radium models support up to 1M tokens:

Model	Context	Good for
hal-1.0	250K	Medium-large codebases
clarke-1.0	1M	Large codebases, multi-file changes
tycho-1.0	125K	Small files, quick edits

Troubleshooting

Symptom	Fix
”Model not found”	Ensure you’re using openai/hal-1.0 format with --model
”401 Unauthorized”	Check API key is valid and starts with your Radium key prefix
Slow diffs	Switch to tycho-1.0 or reduce repo map size with --map-tokens
Context exceeded	Use clarke-1.0 or split the task into smaller edits

Next steps

Quickstart

Your first Radium request

Models

Compare model capabilities

Tool calling

How Radium handles function calling

Rate limits

Current limits by tier