Architecture
Understanding RA-H's multi-agent system
Overview
RA-H uses a multi-agent architecture with specialized AI agents that collaborate to manage your knowledge base. The system is built around three core concepts:
- Nodes - Your knowledge items (papers, ideas, people, projects)
- Edges - How things connect ("relates to", "inspired by")
- Dimensions - Flexible tags that emerge from your content
Core Elements
Nodes
Knowledge items stored in the database. Each node can be a paper, idea, person, project, video, tweet, or any other piece of information. Nodes have:
- Title and content
- Dimensions (multi-tag categorization)
- Metadata (structured information)
- Embeddings (for semantic search)
- Links (to external sources)
Edges
Directed relationships between nodes. Edges capture how your knowledge connects - not just simple links, but contextual relationships like "this paper inspired this idea" or "these concepts relate to this project."
Dimensions
Multi-select categorization tags. Unlike rigid folders, nodes can have multiple dimensions. Some dimensions can be marked as "priority" for focused context, helping you filter and organize based on what's important right now.
Agent System
RA-H uses multiple AI agents working together, each optimized for different tasks.
Easy/Hard Mode Toggle
Switch between two orchestrator agents depending on your needs:
Easy Mode (⚡) - GPT-5 Mini
- Fast responses for everyday questions
- Low cost, high speed
- Default mode for most tasks
Hard Mode (🔥) - Claude Sonnet 4.5
- Deep thinking for complex analysis
- Advanced reasoning capabilities
- Use when you need thorough, thoughtful responses
Seamless switching: You can change modes mid-conversation and maintain context. Your choice persists across sessions.
Background Workers
Oracle (wise-rah) - Complex workflows
- Executes predefined workflows like "integrate" (finding connections across your entire knowledge base)
- Deep analysis tasks
- Direct database write access
Delegates (mini-rah) - Spawned workers
- Handle specific tasks in the background
- Content extraction from YouTube, websites, PDFs
- Batch operations
- Return summaries to keep your conversation clean
How Agents Work Together
All agents share access to your knowledge base:
- They can see pinned nodes (up to 10 that stay in context)
- They work with your currently focused node
- Background workers execute tasks independently and report back
- Your main conversation stays clean and focused
Context Management
RA-H maintains clean context boundaries:
- Main conversation - Full history with the orchestrator (ra-h or ra-h-easy)
- Isolated workers - Background tasks don't pollute your conversation
- Pinned context - Important nodes stay accessible across all agents
- Focused node - Current working context shared by all agents
This architecture ensures fast responses while maintaining the flexibility to handle complex, long-running tasks in the background.