Knowledge Connector

Knowledge Connector

Build a lightweight knowledge graph from documents and conversations.

Core capabilities

• Extract concepts and entities from text
• Create relationships between concepts
• Store and query a knowledge graph
• Recommend related concepts
• Export graph views for inspection

Commands

Extract knowledge

kc extract -f document.txt
kc extract -t "Artificial intelligence is a branch of computer science"
kc extract -f document.txt --save

Create or import connections

kc connect --auto
kc connect --from "Artificial Intelligence" --to "Machine Learning" --relation "includes"
kc connect --file relations.json

Query knowledge

kc query "Artificial Intelligence"
kc query --concept "Deep Learning" --detail
kc query --concept "Neural Network" --related
kc query --ask "What is Machine Learning?"

Visualize the graph

kc visualize
kc visualize --format html --output graph.html
kc visualize --format json --output graph.json
kc visualize --concept "Artificial Intelligence" --depth 2

Manage the knowledge base

kc stats
kc export --output backup.json
kc import --file backup.json
kc clear --confirm

Core data structures

Concept

{
  "id": "uuid",
  "name": "Artificial Intelligence",
  "type": "domain",
  "aliases": ["AI"],
  "description": "A branch of computer science",
  "source": "document.txt",
  "createdAt": "2024-01-01T00:00:00Z",
  "updatedAt": "2024-01-01T00:00:00Z",
  "metadata": {}
}

Relation

{
  "id": "uuid",
  "from": "concept-id-1",
  "to": "concept-id-2",
  "type": "includes",
  "weight": 0.85,
  "source": "auto-extract",
  "createdAt": "2024-01-01T00:00:00Z"
}

Common relation types

• includes
• is-a
• related-to
• causes
• follows
• opposite
• similar

Configuration

Default config file:

{
  "dataDir": "~/.local/share/knowledge-connector",
  "autoExtract": true,
  "autoConnect": true,
  "defaultDepth": 2,
  "maxResults": 20,
  "language": "en"
}

Example workflow

# Extract and save
echo "Python is an interpreted programming language. Java is an object-oriented programming language." | kc extract --save

# Add relations
kc connect --from "Python" --to "Programming Language" --relation "is-a"
kc connect --from "Java" --to "Programming Language" --relation "is-a"

# Inspect graph
kc visualize --format html --output languages.html
kc query --concept "Python" --related

Programmatic usage

const KnowledgeConnector = require('knowledge-connector');
const kc = new KnowledgeConnector();

const concepts = await kc.extract('JavaScript is a dynamically typed language');
await kc.connect({ from: 'JavaScript', to: 'Dynamically Typed Language', type: 'is-a' });
const results = await kc.query('JavaScript');
const recommendations = await kc.recommend('JavaScript');

Notes

• Create regular backups when the graph becomes important.
• Query performance may slow down as the graph grows.
• Concept names may be case-sensitive depending on implementation.

License

MIT