---
name: kg-integration---knowledge-graph-integration-for-h
description: Skill for AI agent capabilities
---

# kg-integration - Knowledge Graph Integration for Hermes

## Description

Integrate sqlite-knowledge-graph (kg.db) with Hermes. Provides tools to search papers, query agents/skills, and find the best agent for tasks using the knowledge graph with 2,497 papers, 452 skills, and 25 agents migrated from OpenClaw.

## Activation Keywords

- knowledge graph
- kg search
- find agent
- recommend skill
- search paper
- arxiv paper
- agent selection
- skill recommendation
- knowledge base query

## Core Concepts

### Knowledge Graph Structure

The kg.db contains:
- **2,497 papers** - arXiv papers with utility scores
- **452 skills** - Extracted from papers with activation keywords
- **25 agents** - Specialized agents migrated from OpenClaw
- **743,466 relations** - Connections between entities

### Entity Types

| Type | Count | Description |
|------|-------|-------------|
| paper | 2,497 | arXiv papers with metadata |
| skill | 452 | Reusable skills from papers |
| agent | 25 | Specialized task agents |

### Relation Types

| Type | Count | Description |
|------|-------|-------------|
| derived_from | 115 | Paper → Skill relationships |
| has_skill | 7 | Agent → Skill relationships |
| related_by_keywords | 740,418 | Keyword-based paper similarity |
| similar_to | 2,926 | Skill similarity |

## Tools

### kg_search_papers

Search papers in the knowledge graph by keywords.

**Parameters:**
- `query` (required): Search query keywords
- `limit` (optional): Maximum number of results (default: 10)

**Example:**
```python
kg_search_papers("multi-agent reinforcement learning", limit=5)
```

### kg_get_agent

Get detailed information about an agent.

**Parameters:**
- `agent_name` (required): Agent name (e.g., algorithm-engineer, ml-engineer)

**Available Agents:**
- algorithm-engineer - Algorithm design and optimization
- ml-engineer - Machine learning model development
- fullstack-engineer - Full-stack development
- data-engineer - Data pipeline and infrastructure
- security-engineer - Security analysis and implementation
- neuroscientist - Neuroscience research
- biologist - Biological research
- geneticist - Genetics research
- economist - Economic analysis
- stock-analyst - Stock market analysis
- mathematician - Mathematical modeling
- physicist - Physics research
- philosopher - Philosophical analysis
- linguist - Linguistic analysis
- psychologist - Psychological analysis
- tech-cofounder - Startup technical leadership
- research-agent - Research assistance
- quantitative-analyst - Quantitative analysis
- computational-scientist - Computational research
- population-dynamics-scientist - Population dynamics
- computer-network-scientist - Network research
- applied-scientist - Applied research
- statistician - Statistical analysis
- logician - Logical analysis
- prompt-engineer - Prompt engineering

**Example:**
```python
kg_get_agent("algorithm-engineer")
```

### kg_get_skill

Get detailed information about a skill.

**Parameters:**
- `skill_name` (required): Skill name

**Example:**
```python
kg_get_skill("turboquant-vector-quantization")
```

### kg_find_agent_for_task

Find the best agent for a given task.

**Parameters:**
- `task_description` (required): Description of the task

**Example:**
```python
kg_find_agent_for_task("Implement a graph neural network for brain connectivity analysis")
```

### kg_recommend_skills

Recommend skills related to a topic.

**Parameters:**
- `topic` (required): Topic or keywords
- `min_utility` (optional): Minimum utility score 0-1 (default: 0.8)

**Example:**
```python
kg_recommend_skills("neural network", min_utility=0.9)
```

### kg_high_utility_papers

Get high utility papers (utility >= 0.85).

**Parameters:**
- `threshold` (optional): Utility threshold (default: 0.85)
- `limit` (optional): Maximum number of results (default: 20)

**Example:**
```python
kg_high_utility_papers(threshold=0.9, limit=10)
```

### kg_stats

Get knowledge graph statistics.

**Example:**
```python
kg_stats()
```

## Usage Patterns

### Pattern 1: Find Agent for Task

When user asks for help with a specific task:
1. Use `kg_find_agent_for_task` to find the best agent
2. Present the top 3 agents with their purposes
3. Recommend the best match

### Pattern 2: Skill Discovery

When user wants to learn about a topic:
1. Use `kg_recommend_skills` to find relevant skills
2. Use `kg_search_papers` to find related papers
3. Present skills with high utility scores

### Pattern 3: Agent Delegation

When a task matches a specialized agent:
1. Use `kg_get_agent` to get agent details
2. Check if the agent has relevant skills
3. Suggest delegating to that agent

### Pattern 4: Research Assistance

When user is researching a topic:
1. Use `kg_search_papers` to find relevant papers
2. Use `kg_recommend_skills` to find implementation skills
3. Present papers with high utility scores first

## Examples

### Example 1: Find Agent for ML Task

User: "I need help training a neural network"

```python
agents = kg_find_agent_for_task("training a neural network")
# Returns: ml-engineer (score: 6)
```

Response: "For training neural networks, I recommend the **ml-engineer** agent. Would you like me to delegate this task?"

### Example 2: Find Skills for Brain Analysis

User: "What skills are available for brain network analysis?"

```python
skills = kg_recommend_skills("brain network", min_utility=0.9)
# Returns: hypergraph-unet-brain (utility: 0.95), dcorp-connectivity-refinement (utility: 0.95), ...
```

### Example 3: Search Papers

User: "Find papers about multi-agent systems"

```python
papers = kg_search_papers("multi-agent", limit=5)
# Returns: EvoScientist, The Devil Behind Moltbook, From Spark to Fire, ...
```

## Technical Details

### Database Location

```
~/.openclaw/openclaw/workspace-fullstack-engineer/sqlite-knowledge-graph/kg.db
```

### Data Sources

- **Papers**: arXiv cs.AI papers with utility scores
- **Skills**: Extracted from papers using skill-extractor
- **Agents**: Migrated from OpenClaw workspace

### Migration History

- **Date**: 2025-04-14
- **Papers**: 2,497 (from kg.db)
- **Skills**: 452 (from ~/.openclaw/openclaw/workspace/skills)
- **Agents**: 25 (from ~/.openclaw/openclaw/agents)

## Integration with Hermes

The kg_tool.py script provides Python functions that can be called from Hermes. The tools are registered in the Hermes tool system for easy access.

## Performance

- Search: < 100ms for typical queries
- Agent matching: < 50ms
- Skill recommendation: < 100ms

## Future Enhancements

- Vector-based semantic search
- Graph traversal for related entities
- Automatic agent delegation suggestions
- Skill learning from conversations

## Tools Used

- `Read` - Read existing files and documentation
- `Write` - Create new files and documentation
- `Bash` - Execute commands when needed

## Instructions for Agents

1. Identify user's intent and specific requirements
2. Gather necessary context from files or user input
3. Execute appropriate actions using available tools
4. Provide clear results and suggest next steps