Show HN: graphiti – Temporal Knowledge Graphs for Agentic Applications
2 comments
100ms is crazy fast. Approx what's the split b/t the embeddings calls and the graph traversal work (the library stuff)?
Running neo4j on reasonable hardware (& depending on graph size), approximately 70% of the latency is calling OpenAI's embedding API. We've seen latency up to 750ms, for just the 3rd-party API call. :-/
As a result, we host embedding models on rented GPUs and get a P90 latency of 10ms.
As a result, we host embedding models on rented GPUs and get a P90 latency of 10ms.
Paul, Preston, and Daniel from Zep here. We’re excited to show you graphiti, a library for building and searching dynamic, temporally aware knowledge graphs.
https://git.new/graphiti
With graphiti, you can model complex, evolving relationships between entities over time. graphiti ingests both unstructured and structured data and the resulting graph may be queried using a fusion of time, full-text, semantic, and graph algorithm approaches.
With graphiti, you can build LLM applications such as:
- Assistants that learn from user interactions, fusing personal knowledge with dynamic data from business systems like CRMs and billing platforms.
- Agents that autonomously execute complex tasks, reasoning with state changes from multiple dynamic sources as varied as traffic conditions or streaming voice transcriptions.
graphiti differs from GraphRAG and other graph libraries. It’s purpose-built for dynamic data and agentic use:
- Smart Graph Updates: Automatically evaluates new entities against the current graph, revising both to reflect the latest context.
- Rich Edge Semantics: Generates human-readable, semantic, and full-text searchable representations for edges during graph construction, enabling search and enhancing interpretability.
- Temporal Awareness: Extracts and updates time-based edge metadata from input data, enabling reasoning over changing relationships.
- Hybrid Search: Offers semantic, BM25, and graph-based search with the ability to fuse results.
- Fast: Search results in < 100ms, with latency primarily determined by the 3rd-party embedding API call.
- Schema Consistency: Maintains a coherent graph structure by reusing existing schema, preventing unnecessary proliferation of node and edge types.
We built graphiti to power Zep Memory, a long-term memory layer for building personalized and accurate LLM apps. We believe graphiti’s potential extends beyond memory applications and have open sourced it to support and grow these use cases.
github: https://git.new/graphiti
documentation: https://help.getzep.com/graphiti
We’d appreciate your feedback, contributions, or just to hear about the awesome projects you’ve built with graphiti!
- Paul, Preston, & Daniel