Apart Intelligence

A knowledge graph for institutional knowledge. AI writes. AI reads. Humans curate. The graph grows as agents work.

Apart Intelligence captures decisions, processes, policies, and architecture as interconnected nodes — so your team's knowledge compounds instead of getting lost in Slack threads and stale wikis.

Install

npm install -g @apart-tech/apart-intelligence

Or run directly:

npx @apart-tech/apart-intelligence --help

Quick Start

Option A: Connect to a hosted server

ai login https://your-server.run.app
ai init

Option B: Use a local PostgreSQL database

Prerequisites: PostgreSQL 16+ with the pgvector extension.

ai init
# Follow the prompts to configure your database and OpenAI API key

Commands

Core

| Command | Description | |---------|-------------| | ai add <content> | Add a knowledge node (decision, process, policy, etc.) | | ai search <query> | Hybrid semantic + keyword search | | ai context <query> | Assemble a context package for a topic | | ai get <id> | Get a node by ID with its edges | | ai update <id> | Update a node's content or metadata | | ai delete <id> | Delete a node or edge |

Organization

| Command | Description | |---------|-------------| | ai link <source> <target> | Link two nodes with a relationship | | ai status <id> <status> | Set node status (draft/reviewed/approved/archived) | | ai drafts | List nodes pending curation | | ai types | List all node and relationship types in use | | ai domains | Manage knowledge domains |

Graph Quality

| Command | Description | |---------|-------------| | ai health | Comprehensive graph health report | | ai orphans | Find nodes with no connections | | ai duplicates | Find near-duplicate nodes by embedding similarity | | ai suggest-links | Suggest missing links based on similarity | | ai islands | Detect disconnected clusters | | ai validate | Validate edge integrity and detect issues | | ai normalize | Analyze and normalize relationship types |

Codebase Mapping

| Command | Description | |---------|-------------| | ai map [directory] | Map a codebase into the knowledge graph | | ai graph | Generate an interactive HTML visualization |

Auth & Config

| Command | Description | |---------|-------------| | ai init | Set up Apart Intelligence in the current directory | | ai login [url] | Log in to an Apart Intelligence server | | ai logout | Log out from the server | | ai whoami | Show current login info |

Examples

# Capture a decision
ai add "We chose Hono over Express for the API server because it's lightweight \
and has native TypeScript support" --type decision --title "Use Hono for API"

# Search for knowledge
ai search "API framework"

# Get context for an AI agent
ai context "onboarding process" --max-nodes 20

# Map your codebase
ai map ./src --domain engineering/backend

# Review and approve drafts
ai drafts
ai status <node-id> approved

# Visualize the knowledge graph
ai graph --open

Claude Code Integration

Apart Intelligence ships with Claude Code skills for capturing knowledge as a natural byproduct of development:

| Skill | Description | |-------|-------------| | /capture <text> | Quick-capture a decision, process, or knowledge node | | /document [path] | Analyze recent changes, generate documentation nodes | | /sync [range] | Incremental knowledge graph update from git diff |

Copy the .claude/ directory from the repository to your project to enable these skills.

Architecture

Apart Intelligence uses a PostgreSQL database with pgvector for semantic search and tsvector for keyword search. Nodes are embedded using OpenAI's embedding model and connected via typed, weighted edges.

Two modes of operation:

• Direct mode — connects to PostgreSQL directly (for local development or self-hosted)
• API mode — connects to a hosted Apart Intelligence server (via ai login)

License

MIT