Skip to main content
The Ryumem dashboard provides a modern web interface for visualizing and managing your knowledge graph memory system.

Accessing the Dashboard

The dashboard is included with your self-hosted Ryumem installation. If you followed the Setup Guide:
  1. Navigate to http://localhost:3000 (or your configured dashboard URL)
  2. Enter your API key (starts with ryu_)
  3. Click “Sign in”
Your API key is stored securely in your browser and used for all API requests.
Need an API key? See Setup Guide for instructions on generating one via the /register endpoint.

Dashboard Overview

The dashboard provides 7 main tabs for different operations (5 tabs when entity extraction is disabled):

Chat & Query

Natural language search interface

Graph

Visual knowledge graph explorer

Entities

Browse and filter entities

Episodes

Manage memories and episodes

Queries

View query history and patterns

Tool Analytics

Track tool execution metrics

Agent Settings

Configure agent behavior
Conditional Features: The Graph and Entities tabs only appear when entity extraction is enabled. This can be toggled in Settings to save 30-50% on LLM token costs.

Tab Reference

Chat & Query

What it does: Natural language interface for searching your knowledge graph. Key Features:
  • Enter queries in natural language (e.g., “Where does Alice work?”)
  • Select search strategy: Hybrid (recommended), Semantic, BM25, or Traversal
  • View results as entity cards with scores
  • See relationships and facts with metadata
  • Example queries for quick testing
Common Workflows:
  • Search for specific information: Use natural language queries
  • Find related entities: Use Hybrid strategy for best results
  • Exact keyword matching: Use BM25 strategy
  • Explore relationships: Use Traversal strategy

Graph Visualization

Requires entity extraction to be enabled
What it shows: Interactive force-directed visualization of your knowledge graph. Key Features:
  • Entity nodes colored by type (Person, Organization, Place, etc.)
  • Relationship edges with labeled connections
  • Zoom and pan controls
  • Filter by entity type
  • Click entities for detailed information
When to use:
  • Understand entity relationships visually
  • Explore community clusters
  • Identify connection patterns
  • Present knowledge graph structure

Entities

Requires entity extraction to be enabled
What it displays: Searchable list of all entities in your knowledge graph. Key Features:
  • Paginated entity list (50 items per page)
  • Filter by entity type
  • Sort by mentions or creation date
  • Entity detail panel with relationships
  • Connected entities navigation
  • Source episodes tracking
Exploration Workflows:
  • Browse all entities: Use pagination
  • Find specific entity types: Apply filters
  • Explore relationships: Click entity for details
  • Trace information sources: View source episodes

Episodes

What you can do: View and add memories to your knowledge graph. Key Features:
  • View all episodes with pagination
  • Filter by user_id, agent_id, or session_id
  • Add new episodes via form
  • See episode metadata (source, timestamps)
  • Link to tool analytics
Adding Episodes:
  1. Click “Add Episode” button
  2. Enter content (text, message, or JSON format)
  3. Select source type
  4. Add optional metadata (user_id, agent_id, session_id)
  5. Submit - entities are extracted automatically
Filtering Options:
  • By user: View memories for specific users
  • By agent: See what specific agents learned
  • By session: Track conversation threads

Query History

What it tracks: Historical queries made to the system with augmentation details. Key Features:
  • View all queries with timestamps
  • See query augmentation with historical context
  • Track similar query patterns
  • Analyze search effectiveness
Use Cases:
  • Understand query patterns
  • See how augmentation improves responses
  • Debug search behavior
  • Identify common information needs

Tool Analytics

What metrics are shown: Execution statistics and performance for tracked tools. Key Features:
  • Tool execution counts
  • Performance metrics (timing, success rates)
  • Parameter and result tracking
  • Filter by tool name
  • Link from episodes
How to Analyze:
  • Identify frequently used tools
  • Track tool performance over time
  • Debug tool execution issues
  • Optimize tool usage patterns

Agent Settings

What you can configure: Agent instructions and behavioral parameters. Key Features:
  • Edit agent instructions
  • Customize agent personality
  • Set behavioral parameters
  • Save and apply changes instantly
Configuration Options:
  • System instructions
  • Response style and tone
  • Tool usage preferences
  • Memory search behavior

Settings Page

Access system settings via the header navigation or System Settings button. 7 Configuration Categories:
Configure OpenAI and Google Gemini API keys. Keys are masked for security.
Provider (OpenAI, Gemini, Ollama, LiteLLM), model name, temperatures, max tokens, timeouts.
Provider selection, model configuration, dimensions, batch size.
Enable/disable toggle (affects Graph/Entities tabs), similarity thresholds, max context episodes.
Track tools, track queries, augment queries, similarity thresholds.
Enable/disable community detection, detection threshold.
Settings Features:
  • ✅ Real-time validation before saving
  • ✅ Visual change indicators
  • ✅ Unsaved changes warning
  • ✅ Hot-reload without server restart
  • ✅ Reset to defaults option

Common Workflows

Quick Search Workflow

  1. Go to Chat & Query tab
  2. Enter your question
  3. Select Hybrid strategy (recommended)
  4. Review entity and relationship results

Adding Memories

  1. Go to Episodes tab
  2. Click “Add Episode”
  3. Enter content and metadata
  4. Submit and wait for processing
  5. View in Graph/Entities tabs

Exploring Entity Relationships

  1. Go to Entities tab (or Graph tab)
  2. Search or filter for entity
  3. Click entity to view details
  4. Explore connected entities
  5. View source episodes

Analyzing Tool Usage

  1. Go to Tool Analytics tab
  2. View tool execution statistics
  3. Filter by specific tools
  4. Analyze patterns and performance

Configuring the System

  1. Click Settings in header
  2. Select configuration category
  3. Update values
  4. Save changes
  5. Changes apply immediately (hot-reload)

Tips & Best Practices

Search Strategy Selection: Use Hybrid for most queries. Switch to BM25 for exact keyword matching or Traversal for relationship-focused searches.
Entity Extraction Cost Optimization: Disable entity extraction in Settings to save 30-50% on LLM token costs if you don’t need graph visualization.
Multi-Tenancy: Always filter by user_id, agent_id, or session_id in production to isolate data properly.
Query Augmentation: Enable in Settings to let agents learn from past similar queries and improve over time.

Next Steps