Context and Memory
How RA-H maintains context and learns from conversations
Context Management
RA-H assembles the right information for each agent during conversations, ensuring they have what they need without overwhelming them with irrelevant details.
What Agents See
Every agent receives:
Base Context - Understanding of how nodes, edges, and dimensions work
Focused Nodes - The node(s) you're currently working with
Pinned Nodes - Up to 10 nodes you've pinned for persistent context across all conversations
Tools - Role-specific capabilities (search, create, update, etc.)
Workflows - Available workflows like "integrate" for finding connections
Pinned Nodes
You can pin up to 10 nodes for persistent context across all conversations. These nodes stay accessible to the AI in every conversation, helping it maintain understanding of your ongoing projects, identity, and important concepts.
How to pin:
- Open a node in the Focus panel
- Click the pin button
- The node will appear in all future conversations
Best practices:
- Pin identity nodes (who you are, your projects)
- Pin frequently referenced frameworks or concepts
- Pin active projects you're currently working on
- Unpin stale nodes to keep context fresh
Memory System
RA-H automatically extracts insights from your conversations and creates memory nodes that help it better understand you over time.
How Memory Works
Automatic extraction - After every conversation, RA-H analyzes what it learned about you
Memory categories:
- Identity - Who you are, your roles, background
- Interests - What you're curious about, learning, exploring
- Models - How you think, mental frameworks, approaches
- Preferences - What you like/dislike, priorities, values
- Relationships - Connections to people, projects, concepts
Subject types: Memory nodes are organized by what they're about:
- People (yourself or collaborators)
- Projects (initiatives, products)
- Concepts (ideas, theories, mental models)
- Resources (tools, articles, books)
- Organizations (companies, institutions)
- Workflows (processes, systems)
Memory in Action
As you use RA-H, it builds a growing understanding of:
- Your current projects and priorities
- How you prefer to organize information
- Your areas of expertise and interest
- Your thinking style and mental frameworks
- Important people and resources in your work
This memory feeds back into future conversations, allowing RA-H to provide more contextually relevant suggestions, find better connections, and adapt to your specific needs.
Privacy
All memory extraction happens locally on your machine. Memory nodes are stored in your local SQLite database, just like any other node. You can view, edit, or delete any memory node at any time.