Admin interface for the Graphiti knowledge graph framework. Manage, visualize, and query your knowledge graphs through an interactive web UI.

This project is 100% AI-generated code (vibe-coding with Claude).

Important: This UI requires Brusdeylins/graphiti (fork), not the upstream getzep/graphiti. The fork adds CRUD operations, extended MCP tools, and FalkorDB multi-graph fixes that this UI depends on. The upstream repo does not provide these APIs.

• Dashboard — Service status overview (LLM, Embedder, MCP Server, FalkorDB), MCP configuration snippet for clients
• Graph Visualization — Interactive force-directed graph (2D/3D) with node/edge editing, entity type legend, and Shift+Click edge creation
• Cypher Query Editor — Run queries against your graph database with selectable graph targets
• Entity Types — View and edit entity type definitions used for LLM-based knowledge extraction
• API Keys — Create and manage API keys for MCP server access
• Knowledge Ingestion — Send text to the LLM for automatic entity/relationship extraction

On first start, you'll be prompted to set an admin password. After that, log in with the configured username (default: admin) and your password.

The main workspace. Select a graph from the dropdown at the top, then explore:

• Left-click a node to view its details in the sidebar (attributes, connections, episodes)
• Right-click + drag to pan the view
• Scroll wheel to zoom, middle-click to fit all nodes
• Shift+Click on a second node to create a new edge between two selected nodes
• Node / Edge buttons to create entities directly (without LLM)
• Send Knowledge to submit text for LLM-based extraction into the graph
• Toggle between 2D and 3D rendering modes
• Adjust label visibility, gravity, and node size with the sliders

Shows the health status of all connected services and provides the MCP configuration JSON that clients need to connect.

View the entity type schema that guides LLM extraction. Edit field definitions, add new types, or reset to defaults from the config file.

Generate API keys for authenticating MCP clients. Keys are shown once at creation — copy them immediately.

• Docker and Docker Compose
• A running Graphiti MCP Server (or the upstream getzep/graphiti)
• An LLM endpoint (OpenAI-compatible: OpenAI, Ollama, etc.)

# Clone the repo
git clone https://github.com/Brusdeylins/graphiti-ui.git
cd graphiti-ui

# Prepare environment
cp .env.example .env
# Edit .env with your API keys and settings

# Start the full stack (FalkorDB + MCP Server + UI)
docker compose -f docker-compose.example.yml up -d

# Open browser
open http://localhost:8080

# Build the Docker image
./build.sh

# Or manually
docker build -t graphiti-ui:latest .

The Docker image installs graphiti-core directly from the main branch of the Brusdeylins fork. To use a different branch or the upstream package, edit the RUN uv pip install line in the Dockerfile.

graphiti-ui/
├── Dockerfile                  # Multi-stage Docker build
├── build.sh                    # Build script
├── pyproject.toml              # Python dependencies
├── .env.example                # Environment variable template
├── docker-compose.example.yml  # Local development stack
├── frontend/                   # React frontend (TypeScript)
│   └── src/
│       ├── components/         # Layout, ForceGraphVisualization
│       ├── pages/              # Dashboard, Visualization, Query, ...
│       ├── contexts/           # Auth, Theme
│       └── api/                # Axios HTTP client
└── src/                        # FastAPI backend (Python)
    ├── api/                    # Route handlers
    ├── auth/                   # JWT authentication
    ├── models/                 # Pydantic models
    └── services/               # Business logic

• Graphiti Framework (fork)
• Graphiti (upstream)
• FalkorDB
• Tabler UI
• MCP Specification

MIT