Skip to main content

User Guide

Welcome to the Radium User Guide. This comprehensive guide covers everything you need to know to effectively use Radium for building and managing AI agent systems.

Getting Started​

New to Radium? Start here:

Core Features​

Agent Configuration​

Configure and manage AI agents for your specific needs.

  • Agent Configuration Guide - Complete guide to configuring agents
    • TOML configuration format
    • Model selection and parameters
    • Prompt management
    • Persona system integration
    • Self-hosted model configuration

Orchestration​

Intelligent task routing that automatically selects and coordinates agents.

Persona System​

Intelligent model selection with cost optimization and automatic fallback.

  • Persona System - Model selection and cost optimization
    • Performance profiles
    • Fallback chains
    • Cost optimization strategies

Memory & Context​

Maintain continuity across sessions and provide context to agents.

Learning System​

Track mistakes, preferences, and successes to build pattern recognition.

  • Learning System - Learn from experience
    • Mistake tracking
    • Preference learning
    • ACE Skillbook

Vibe Check​

Metacognitive oversight to prevent reasoning lock-in and improve agent alignment.

  • Vibe Check - Chain-Pattern Interrupt system
    • Risk assessment
    • Pattern detection
    • Phase-aware feedback

Constitution Rules​

Session-scoped rules for workflow constraints.

Custom Commands​

Reusable command definitions with template substitution.

Guides​

Step-by-step guides for common tasks:

Feature Documentation​

Planning & Execution​

Security & Policies​

Advanced Features​

Integration Guides​

Extensions​

MCP Integration​

Self-Hosted Models​

CLI Reference​

Roadmap & Vision​

Best Practices​

Agent Design​

  • Use clear, specific prompts
  • Define agent roles and responsibilities
  • Configure appropriate models for tasks
  • Use persona system for cost optimization

Orchestration​

  • Write natural, descriptive requests
  • Let orchestrator select agents automatically
  • Use multi-agent workflows for complex tasks

Memory & Context​

  • Leverage plan-scoped memory
  • Use context sources for external information
  • Build on previous agent outputs

Security​

  • Configure policies for tool execution
  • Use approval modes appropriately
  • Set up session constitutions for sensitive tasks

Troubleshooting​

Common issues and solutions:

  • Agent not found: Check agent discovery and configuration
  • Orchestration not working: Verify API keys and configuration
  • Memory issues: Check memory storage and retrieval
  • Performance problems: Review session analytics and optimize

For detailed troubleshooting, see:

Next Steps​

Ready to build? Start with Agent Configuration or explore Orchestration for intelligent task routing.