DEV.co
RAG Frameworks · getzep

graphiti

Graphiti is an open-source Python framework that builds temporal knowledge graphs for AI agents, tracking how facts change over time with full provenance to source data. It enables incremental updates and hybrid retrieval (semantic, keyword, and graph-based) without batch recomputation, making it suitable for dynamic, context-aware agent applications.

Source: GitHub — github.com/getzep/graphiti
28.5k
GitHub stars
2.9k
Forks
Python
Primary language
Apache-2.0
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
Repositorygetzep/graphiti
Ownergetzep
Primary languagePython
LicenseApache-2.0 — OSI-approved
Stars28.5k
Forks2.9k
Open issues415
Latest releasev0.29.2 (2026-06-08)
Last updated2026-07-08
Sourcehttps://github.com/getzep/graphiti

What graphiti is

Graphiti constructs bi-temporal context graphs with entities, relationships (facts with validity windows), episodes (raw data provenance), and pluggable graph backends. It supports prescribed ontology via Pydantic models and learned structure, offering sub-second hybrid retrieval combining embeddings, BM25, and graph traversal for evolving, real-world data streams.

Quickstart

Get the graphiti source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/getzep/graphiti.gitcd graphiti# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Dynamic Agent Memory & Context

Deploy AI agents that maintain evolving context about users, entities, and relationships with temporal awareness. Track what is true now vs. historical facts without losing provenance to source data.

Real-Time Knowledge Integration

Ingest structured and unstructured data continuously into a coherent graph without batch recomputation. Support incremental updates to facts, entities, and relationships as new information arrives.

Hybrid Semantic Search with History

Build retrieval systems combining semantic embeddings, keyword search, and graph traversal for low-latency, context-aware queries across time. Avoid reliance on LLM summarization for contradiction resolution.

Implementation considerations

  • Requires selection and integration of a graph backend (database, embedding store, vector index); no default single deployment artifact provided.
  • Custom ontology definition via Pydantic models is optional but recommended for structured data; learned ontology may require iteration and validation.
  • Episode ingestion pipeline must be designed to feed raw data reliably; schema evolution and data quality considerations fall on the implementer.
  • Hybrid retrieval tuning (semantic weight vs. keyword vs. graph traversal) is custom per use-case; no pre-optimized defaults documented.
  • Temporal query semantics (validity windows, fact invalidation logic) must be understood and correctly applied in business logic.

When to avoid it — and what to weigh

  • Static Document Summarization — If your primary need is batch-oriented knowledge extraction from fixed document sets, consider GraphRAG or traditional RAG instead. Graphiti is optimized for continuous evolution, not one-shot processing.
  • Turnkey Production Deployment — If you need managed infrastructure, governance, sub-200ms retrieval guarantees, or enterprise support out-of-the-box, use Zep (the commercial platform). Graphiti requires self-hosting and custom operational setup.
  • Simple Chatbot Memory — For basic conversation history or session-level context, the complexity and overhead of a temporal knowledge graph may be unjustified. Consider simpler vector stores or traditional conversation buffers.
  • Graph-Backend Agnostic Requirements — Graphiti requires a pluggable graph backend. If your infrastructure lacks a suitable graph database or embedding store, integration work is required before deployment can begin.

License & commercial use

Apache License 2.0 (Apache-2.0). Permissive OSI-approved license allowing commercial use, modification, and distribution under Apache terms.

Apache-2.0 permits commercial use, but you must retain license notices and copyright attribution in source and distributed code. No warranty or liability provided by the licensor. For production AI agent services, review liability, indemnification, and support implications with legal counsel. This is the open-source engine only; enterprise features and support are available through Zep's managed platform.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationAdequate
License clarityClear
Deployment complexityHigh
DEV.co fitGood
Assessment confidenceHigh
Security considerations

Graphiti itself is an open-source Python library; security posture depends on your graph backend, embedding store, and agent integration. No encryption, authentication, or multi-tenancy baked in. Provenance tracking to episodes helps audit data lineage. Self-hosting requires you to manage secrets, access control, and compliance; no managed security guarantees provided.

Alternatives to consider

Zep (Managed Platform)

Commercial platform offering managed temporal context graphs with enterprise governance, sub-200ms retrieval, dashboards, and support. Choose if you want turnkey production infrastructure and can accept vendor lock-in.

GraphRAG

Microsoft's batch-oriented knowledge graph extraction for static document sets. Better for one-shot summarization; not optimized for continuous, evolving agent context.

Traditional Vector RAG + Conversation State

Simpler alternative using embeddings + vector search + chat history. Sufficient for many use-cases; avoids graph complexity if temporal fact management and provenance are not required.

Software development agency

Build on graphiti with DEV.co software developers

Graphiti is a powerful fit for AI agents requiring evolving context and temporal fact management. Devco engineers can help you assess backend integration, ontology design, and production deployment. Contact us to explore fit and implementation effort.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Related on DEV.co

Explore the category and the services that help you build with it.

graphiti FAQ

Can I use Graphiti in production?
Yes, but it is self-hosted and requires you to operate the graph backend, embedding store, and ingestion pipeline. For managed, SLA-backed production use, Zep is the recommended choice.
Do I have to define an ontology upfront?
No. Graphiti supports both prescribed ontology (via Pydantic models) and learned structure. Start simple and evolve as patterns emerge from your data.
What graph databases are supported?
Documentation excerpt does not specify. Graphiti's architecture includes pluggable backends; review the repo's backend adapters or contact the maintainers to confirm compatibility with your chosen database.
How does Graphiti handle contradictory facts?
Automatically via temporal invalidation. When new data contradicts an old fact, the old fact's validity window is closed and marked superseded. Full history is preserved for historical queries.

Software development & web development with DEV.co

DEV.co helps companies turn open-source tools like graphiti into production software. Our software development services cover the full lifecycle — architecture, web development, integration, and maintenance — delivered by software developers and web developers who ship. Engage our software development agency to implement or customize it for your rag frameworks stack.

Evaluate Graphiti for Your Agent Architecture

Graphiti is a powerful fit for AI agents requiring evolving context and temporal fact management. Devco engineers can help you assess backend integration, ontology design, and production deployment. Contact us to explore fit and implementation effort.