Installation

Requirements

Before installing Graphiti, ensure you have:

Python 3.10 or higher
A graph database (one of the following):
- Neo4j 5.26+
- FalkorDB 1.1.2+
- Kuzu 0.11.2+
- Amazon Neptune Database Cluster or Neptune Analytics Graph
An LLM provider API key (OpenAI recommended)

Graphiti works best with LLM services that support Structured Output (such as OpenAI and Gemini). Using other services may result in incorrect output schemas and ingestion failures, particularly with smaller models.

Basic Installation

Install Graphiti using your preferred package manager:

pip install graphiti-core

The basic installation includes:

Neo4j driver support (default)
OpenAI LLM and embedding support
Core dependencies (Pydantic, tenacity, numpy)

Graph Database Backends

Graphiti supports multiple graph database backends through optional dependencies.

Neo4j
FalkorDB
Kuzu
Amazon Neptune

Neo4j is the default backend and is included in the basic installation.

Setup

Install Neo4j Desktop

Download and install Neo4j Desktop for a user-friendly interface to manage Neo4j instances.

Create a Database

In Neo4j Desktop:

Create a new project
Add a local DBMS
Set a password
Start the DBMS

Configure Connection

from graphiti_core import Graphiti

graphiti = Graphiti(
    uri="bolt://localhost:7687",
    user="neo4j",
    password="your_password"
)

Neo4j Desktop provides a browser interface at http://localhost:7474 for visualizing your knowledge graph.

FalkorDB is a lightweight, Redis-compatible graph database that’s easy to set up with Docker.

pip install graphiti-core[falkordb]
# or
uv add graphiti-core[falkordb]

Setup with Docker

docker run -p 6379:6379 -p 3000:3000 -it --rm falkordb/falkordb:latest

Configure Connection

from graphiti_core import Graphiti
from graphiti_core.driver.falkordb_driver import FalkorDriver

driver = FalkorDriver(
    host="localhost",
    port=6379,
    username=None,  # Optional
    password=None,  # Optional
    database="my_graph"  # Optional, defaults to "default_db"
)

graphiti = Graphiti(graph_driver=driver)

FalkorDB provides a browser interface at http://localhost:3000 when using the Docker image.

Kuzu is an embedded graph database that doesn’t require a separate server.

Installation

pip install graphiti-core[kuzu]
# or
uv add graphiti-core[kuzu]

Configure Connection

from graphiti_core import Graphiti
from graphiti_core.driver.kuzu_driver import KuzuDriver

driver = KuzuDriver(db="/tmp/graphiti.kuzu")
graphiti = Graphiti(graph_driver=driver)

Kuzu is ideal for development and testing since it requires no server setup.

Amazon Neptune is a fully managed graph database service in AWS.

Installation

pip install graphiti-core[neptune]
# or
uv add graphiti-core[neptune]

Configure Connection

from graphiti_core import Graphiti
from graphiti_core.driver.neptune_driver import NeptuneDriver

driver = NeptuneDriver(
    host="neptune-db://<cluster-endpoint>",  # For Neptune Database
    # OR
    # host="neptune-graph://<graph-identifier>",  # For Neptune Analytics
    aoss_host="<opensearch-serverless-host>",
    port=8182,  # Optional, defaults to 8182
    aoss_port=443  # Optional, defaults to 443
)

graphiti = Graphiti(graph_driver=driver)

Amazon Neptune requires an OpenSearch Serverless collection as the full-text search backend.

LLM Providers

Graphiti supports multiple LLM providers. OpenAI is the default and recommended option.

OpenAI is included in the basic installation and is the default LLM provider.

Setup

export OPENAI_API_KEY=your_openai_api_key

from graphiti_core import Graphiti

# Uses OpenAI by default
graphiti = Graphiti(uri, user, password)

Graphiti defaults to using gpt-4o-mini for inference and text-embedding-3-small for embeddings.

Use Azure’s OpenAI service for enterprise deployments.

Setup

from openai import AsyncOpenAI
from graphiti_core import Graphiti
from graphiti_core.llm_client.azure_openai_client import AzureOpenAILLMClient
from graphiti_core.llm_client.config import LLMConfig
from graphiti_core.embedder.azure_openai import AzureOpenAIEmbedderClient

# Initialize Azure OpenAI client
azure_client = AsyncOpenAI(
    base_url="https://your-resource.openai.azure.com/openai/v1/",
    api_key="your-api-key",
)

# Create clients
llm_client = AzureOpenAILLMClient(
    azure_client=azure_client,
    config=LLMConfig(
        model="gpt-4o-mini",  # Your deployment name
        small_model="gpt-4o-mini"
    )
)

embedder_client = AzureOpenAIEmbedderClient(
    azure_client=azure_client,
    model="text-embedding-3-small"  # Your deployment name
)

# Initialize Graphiti
graphiti = Graphiti(
    uri, user, password,
    llm_client=llm_client,
    embedder=embedder_client,
)

Use Claude models for LLM inference.

Installation

pip install graphiti-core[anthropic]
# or
uv add graphiti-core[anthropic]

Setup

from graphiti_core import Graphiti
from graphiti_core.llm_client.anthropic_client import AnthropicClient
from graphiti_core.llm_client.config import LLMConfig

llm_client = AnthropicClient(
    config=LLMConfig(
        api_key="your-anthropic-api-key",
        model="claude-3-5-sonnet-20241022"
    )
)

graphiti = Graphiti(
    uri, user, password,
    llm_client=llm_client
)

Use Google’s Gemini models for LLM, embeddings, and reranking.

Installation

pip install graphiti-core[google-genai]
# or
uv add graphiti-core[google-genai]

Setup

from graphiti_core import Graphiti
from graphiti_core.llm_client.gemini_client import GeminiClient, LLMConfig
from graphiti_core.embedder.gemini import GeminiEmbedder, GeminiEmbedderConfig
from graphiti_core.cross_encoder.gemini_reranker_client import GeminiRerankerClient

api_key = "your-google-api-key"

graphiti = Graphiti(
    uri, user, password,
    llm_client=GeminiClient(
        config=LLMConfig(
            api_key=api_key,
            model="gemini-2.0-flash"
        )
    ),
    embedder=GeminiEmbedder(
        config=GeminiEmbedderConfig(
            api_key=api_key,
            embedding_model="embedding-001"
        )
    ),
    cross_encoder=GeminiRerankerClient(
        config=LLMConfig(
            api_key=api_key,
            model="gemini-2.5-flash-lite"
        )
    )
)

Run local LLMs for privacy and cost savings.

Installation

# Install Ollama from https://ollama.ai

# Pull models
ollama pull deepseek-r1:7b
ollama pull nomic-embed-text

Setup

from graphiti_core import Graphiti
from graphiti_core.llm_client.config import LLMConfig
from graphiti_core.llm_client.openai_generic_client import OpenAIGenericClient
from graphiti_core.embedder.openai import OpenAIEmbedder, OpenAIEmbedderConfig
from graphiti_core.cross_encoder.openai_reranker_client import OpenAIRerankerClient

llm_config = LLMConfig(
    api_key="ollama",  # Placeholder
    model="deepseek-r1:7b",
    small_model="deepseek-r1:7b",
    base_url="http://localhost:11434/v1",
)

llm_client = OpenAIGenericClient(config=llm_config)

graphiti = Graphiti(
    uri, user, password,
    llm_client=llm_client,
    embedder=OpenAIEmbedder(
        config=OpenAIEmbedderConfig(
            api_key="ollama",
            embedding_model="nomic-embed-text",
            embedding_dim=768,
            base_url="http://localhost:11434/v1",
        )
    ),
    cross_encoder=OpenAIRerankerClient(
        client=llm_client,
        config=llm_config
    ),
)

Use OpenAIGenericClient (not OpenAIClient) for Ollama and other OpenAI-compatible providers.

Use Groq for fast inference with open-source models.

Installation

pip install graphiti-core[groq]
# or
uv add graphiti-core[groq]

Setup

from graphiti_core import Graphiti
from graphiti_core.llm_client.groq_client import GroqClient
from graphiti_core.llm_client.config import LLMConfig

llm_client = GroqClient(
    config=LLMConfig(
        api_key="your-groq-api-key",
        model="llama-3.3-70b-versatile"
    )
)

graphiti = Graphiti(
    uri, user, password,
    llm_client=llm_client
)

Optional Dependencies

Graphiti offers several optional dependencies for extended functionality:

Voyage AI Embeddings

Use Voyage AI for state-of-the-art embeddings:

pip install graphiti-core[voyageai]

Sentence Transformers

Use local embedding models:

pip install graphiti-core[sentence-transformers]

OpenTelemetry Tracing

Enable distributed tracing for observability:

pip install graphiti-core[tracing]

Multiple Extras

Install multiple optional dependencies at once:

pip install graphiti-core[falkordb,anthropic,google-genai,voyageai]

Performance Tuning

Concurrency Control

Graphiti’s ingestion pipelines support high concurrency. Control this with the SEMAPHORE_LIMIT environment variable:

# Default (low concurrency to avoid rate limits)
export SEMAPHORE_LIMIT=10

# Higher concurrency for better performance (if your LLM provider allows)
export SEMAPHORE_LIMIT=50

# Lower concurrency to avoid 429 rate limit errors
export SEMAPHORE_LIMIT=5

By default, SEMAPHORE_LIMIT is set to 10 to help prevent 429 rate limit errors from your LLM provider. Increase this value if your provider supports higher throughput.

Environment Variables

Common environment variables for Graphiti:

.env

# LLM Provider
OPENAI_API_KEY=your_openai_api_key

# Neo4j Connection
NEO4J_URI=bolt://localhost:7687
NEO4J_USER=neo4j
NEO4J_PASSWORD=password

# FalkorDB Connection
FALKORDB_HOST=localhost
FALKORDB_PORT=6379
FALKORDB_USERNAME=  # Optional
FALKORDB_PASSWORD=  # Optional

# Amazon Neptune
NEPTUNE_HOST=neptune-db://your-cluster-endpoint
NEPTUNE_PORT=8182
AOSS_HOST=your-opensearch-host
AOSS_PORT=443

# Performance
SEMAPHORE_LIMIT=10

# Telemetry (opt-out)
GRAPHITI_TELEMETRY_ENABLED=false

Telemetry

Graphiti collects anonymous usage statistics to improve the framework. This includes:

System information (OS, Python version)
Graphiti version
Provider choices (LLM, database, embedder)

Data NOT collected:

Personal information or identifiers
API keys or credentials
Your data, queries, or graph content
IP addresses or hostnames

Disable Telemetry

To opt out of telemetry:

export GRAPHITI_TELEMETRY_ENABLED=false

Or in your Python code:

import os
os.environ['GRAPHITI_TELEMETRY_ENABLED'] = 'false'

Telemetry is automatically disabled during test runs when pytest is detected.

Verify Installation

Verify your installation with this simple script:

verify.py

import asyncio
from graphiti_core import Graphiti
from graphiti_core.nodes import EpisodeType
from datetime import datetime, timezone

async def verify():
    # Initialize Graphiti (adjust connection parameters)
    graphiti = Graphiti(
        "bolt://localhost:7687",
        "neo4j",
        "password"
    )
    
    try:
        # Add a test episode
        await graphiti.add_episode(
            name="Test Episode",
            episode_body="Test content",
            source=EpisodeType.text,
            reference_time=datetime.now(timezone.utc)
        )
        print("✓ Installation verified successfully!")
    finally:
        await graphiti.close()

if __name__ == "__main__":
    asyncio.run(verify())

Next Steps

Quickstart

Build your first knowledge graph

Core Concepts

Learn about episodes, nodes, and edges

Search Guide

Master hybrid search and reranking

Examples

Explore example projects

Get Started

Core Concepts

Guides

Integrations

Advanced

​Requirements

​Basic Installation

​Graph Database Backends

​Setup

​Installation

​Setup with Docker

​Configure Connection

​Installation

​Configure Connection

​Installation

​Configure Connection

​LLM Providers

​Setup

​Setup

​Installation

​Setup

​Installation

​Setup

​Installation

​Setup

​Installation

​Setup

​Optional Dependencies

​Voyage AI Embeddings

​Sentence Transformers

​OpenTelemetry Tracing

​Multiple Extras

​Performance Tuning

​Concurrency Control

​Environment Variables

​Telemetry

​Disable Telemetry

​Verify Installation

​Next Steps

Quickstart

Core Concepts

Search Guide

Examples

Build docs developers (and LLMs) love

Requirements

Basic Installation

Graph Database Backends

Setup

Installation

Setup with Docker

Configure Connection

Installation

Configure Connection

Installation

Configure Connection

LLM Providers

Setup

Setup

Installation

Setup

Installation

Setup

Installation

Setup

Installation

Setup

Optional Dependencies

Voyage AI Embeddings

Sentence Transformers

OpenTelemetry Tracing

Multiple Extras

Performance Tuning

Concurrency Control

Environment Variables

Telemetry

Disable Telemetry

Verify Installation

Next Steps