graphiti-guide

本条目面向搭 Agent 与 AI 应用场景,提供“graphiti-guide”技能的目录入口。公开标签:learning、知识图谱、AI代理、Neo4j。请通过固定上游版本核对实际功能、依赖与权限;此页为 Skillπ 独立编写的用途介绍和原站链接,不代表运行安全保证。原文与附件依各自许可开放。

仓库
brycewang-stanford/Auto-Empirical-Research-Skills
仓库最近推送
Stars
5,023
Forks
578
Issues / PRs
0

安装方式

先用 Prompt 核对来源,也可切换到安装命令或下载可用的安装包。

安装方式

用 AI 助手安装

复制提示词,让助手先核对固定版本和依赖。

查看 Prompt收起 Prompt
先检查当前 Agent 是否已安装 SkillPI Skill Finder。如果尚未安装,请按官方仓库的说明安装并启用:https://github.com/Findskills/skillpi。然后使用 Skill Finder 打开此 SkillPI 详情页并定位 "graphiti-guide":https://skillpi.com/zh/skills/deepskill-832d4060c6eb033cbe871db7b85e2f73a0bba6b87d99740d8058a1ae5ce5e41f。安装前核对上游来源和版本:https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba/skills/43-wentorai-research-plugins/skills/tools/knowledge-graph/graphiti-guide/SKILL.md。来源不一致时停止并说明差异。核对通过后,用 Skill Finder 安装目标 Skill,并报告安装路径和来源版本。

此命令从 GitHub 来源仓库获取指定技能。运行前请先查看源文件。

运行工具
npx skills add https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/tree/bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba/skills/43-wentorai-research-plugins/skills/tools/knowledge-graph/graphiti-guide --skill graphiti-guide
许可协议: —

来源与配置

查看 GitHub 源码

brycewang-stanford/Auto-Empirical-Research-Skills固定提交 bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba

配置方式待核实

请查看原始来源中的适用工具、依赖和权限要求。

安装实测待补充

本条目面向搭 Agent 与 AI 应用场景,提供“graphiti-guide”技能的目录入口。公开标签:learning、知识图谱、AI代理、Neo4j。请通过固定上游版本核对实际功能、依赖与权限;此页为 Skillπ 独立编写的用途介绍和原站链接,不代表运行安全保证。原文与附件依各自许可开放。

↓ 0 下载 ↗ 0 分享点击

文件浏览

在原站浏览文件 ↗

原始 SKILL.md

打开原文 ↗
---
name: graphiti-guide
description: "Build real-time knowledge graphs for AI agents using Graphiti by Zep"
metadata:
  openclaw:
    emoji: "🕸"
    category: "tools"
    subcategory: "knowledge-graph"
    keywords: ["knowledge-graph", "ai-agents", "temporal-graphs", "neo4j", "memory", "entity-extraction"]
    source: "https://github.com/getzep/graphiti"
---

# Graphiti Guide

## Overview

Graphiti is an open-source framework for building and querying dynamic, temporally-aware knowledge graphs designed specifically for AI agent applications. Developed by Zep, Graphiti enables agents to maintain persistent, structured memory that evolves over time, capturing entities, relationships, and facts extracted from conversational and documentary sources.

Traditional knowledge graphs are static structures that require manual curation and batch updates. Graphiti takes a fundamentally different approach: it incrementally builds and updates the graph in real-time as new information arrives, resolving contradictions, merging duplicate entities, and maintaining temporal metadata that tracks when facts were established and whether they remain current.

For academic researchers, Graphiti offers a powerful framework for constructing domain-specific knowledge graphs from research literature, experimental observations, and collaborative discussions. With over 23,000 GitHub stars, the project has gained significant traction in both the AI engineering and research communities as a practical bridge between unstructured text and structured, queryable knowledge.

## Installation and Setup

Install Graphiti via pip:

```bash
pip install graphiti-core
```

Graphiti requires a Neo4j database for graph storage. Set up Neo4j using Docker:

```bash
docker run -d \
  --name neo4j-graphiti \
  -p 7474:7474 -p 7687:7687 \
  -e NEO4J_AUTH=neo4j/your-password \
  -e NEO4J_PLUGINS='["apoc"]' \
  neo4j:5
```

Configure environment variables for your LLM provider and Neo4j connection:

```bash
export NEO4J_URI=bolt://localhost:7687
export NEO4J_USER=neo4j
export NEO4J_PASSWORD=$NEO4J_PASSWORD
export OPENAI_API_KEY=$OPENAI_API_KEY
```

Initialize Graphiti in your Python project:

```python
from graphiti_core import Graphiti

graphiti = Graphiti(
    neo4j_uri="bolt://localhost:7687",
    neo4j_user="neo4j",
    neo4j_password=$NEO4J_PASSWORD,
)

# Build indices for efficient querying
await graphiti.build_indices()
```

## Core Features

**Incremental Graph Construction**: Add episodes of information that are automatically parsed into entities and relationships:

```python
from graphiti_core.nodes import EpisodeType
from datetime import datetime

# Add a research observation
await graphiti.add_episode(
    name="experiment_log_2026_03_10",
    episode_body="""
    The CRISPR-Cas9 experiment targeting gene BRCA1 in HeLa cells
    showed 87% knockout efficiency. The guide RNA sequence gRNA-42
    was designed using the Benchling platform. Dr. Chen supervised
    the experiment, which used the protocol established in our
    2025 Nature Methods paper.
    """,
    source=EpisodeType.text,
    source_description="Lab notebook entry",
    reference_time=datetime(2026, 3, 10),
)
```

Graphiti automatically extracts entities (BRCA1, HeLa cells, Dr. Chen, gRNA-42, Benchling), establishes relationships (gRNA-42 targets BRCA1, Dr. Chen supervised the experiment), and records temporal metadata.

**Temporal Awareness**: Knowledge graphs built with Graphiti track when facts were established and can reason about changes over time:

```python
# Add an update that modifies a previous fact
await graphiti.add_episode(
    name="experiment_update",
    episode_body="""
    After reanalysis, the CRISPR knockout efficiency for BRCA1
    in HeLa cells was revised to 82% due to off-target effects
    detected in the secondary sequencing run.
    """,
    source=EpisodeType.text,
    source_description="Updated analysis",
    reference_time=datetime(2026, 3, 12),
)
```

The graph updates the efficiency value while maintaining the historical record, enabling queries about both current and historical states of knowledge.

**Semantic Search**: Query the knowledge graph using natural language:

```python
# Search for relevant entities and facts
results = await graphiti.search(
    query="What is the knockout efficiency for BRCA1?",
    num_results=5,
)

for result in results:
    print(f"Fact: {result.fact}")
    print(f"Source: {result.source_description}")
    print(f"Valid from: {result.valid_at}")
    print(f"Confidence: {result.score}")
    print("---")
```

**Entity Resolution**: Graphiti handles duplicate and variant entity references automatically. References to "CRISPR-Cas9", "CRISPR", and "Cas9 system" are resolved to the appropriate entities based on context, reducing manual curation overhead.

## Research Workflow Integration

**Literature Knowledge Base**: Build a continuously growing knowledge graph from your reading notes and paper summaries:

```python
# Process a batch of paper summaries
papers = [
    {
        "title": "Attention Is All You Need",
        "summary": "Vaswani et al. introduced the Transformer architecture...",
        "date": datetime(2017, 6, 12),
    },
    {
        "title": "BERT: Pre-training of Deep Bidirectional Transformers",
        "summary": "Devlin et al. proposed BERT, a masked language model...",
        "date": datetime(2018, 10, 11),
    },
]

for paper in papers:
    await graphiti.add_episode(
        name=paper["title"],
        episode_body=paper["summary"],
        source=EpisodeType.text,
        source_description=f"Paper summary: {paper['title']}",
        reference_time=paper["date"],
    )
```

Then query across your entire reading history to find connections, trace the evolution of ideas, and identify foundational works.

**Experimental Knowledge Management**: Track the relationships between experiments, reagents, instruments, protocols, and personnel. This creates an institutional memory that survives personnel turnover and helps new lab members quickly understand the research context.

**Systematic Review Support**: As you read and annotate papers for a systematic review, feed summaries into Graphiti to build a structured representation of findings, methodologies, and populations studied. Query the resulting graph to identify patterns and gaps.

**Agent Memory for Research Assistants**: Integrate Graphiti as the memory backend for AI research assistants. The agent can accumulate knowledge from each interaction and provide increasingly informed responses:

```python
# Research agent with persistent memory
async def research_agent_step(user_query, conversation_context):
    # Retrieve relevant knowledge
    memories = await graphiti.search(
        query=user_query,
        num_results=10,
    )

    # Include memories in the agent context
    context = format_memories(memories) + conversation_context

    # Generate response using LLM
    response = await generate_response(user_query, context)

    # Store new knowledge from the interaction
    await graphiti.add_episode(
        name=f"conversation_{datetime.now().isoformat()}",
        episode_body=f"User asked: {user_query}\nResponse: {response}",
        source=EpisodeType.message,
        source_description="Research assistant conversation",
        reference_time=datetime.now(),
    )

    return response
```

## Graph Exploration and Maintenance

Access the Neo4j browser at `http://localhost:7474` to visually explore your knowledge graph. Use Cypher queries for advanced exploration:

```cypher
// Find all entities related to a specific gene
MATCH (e:Entity)-[r]-(connected)
WHERE e.name CONTAINS 'BRCA1'
RETURN e, r, connected
LIMIT 50

// Find the most connected entities (research hubs)
MATCH (e:Entity)-[r]-()
RETURN e.name, COUNT(r) AS connections
ORDER BY connections DESC
LIMIT 20
```

Periodically review and clean the graph to maintain quality. Remove incorrectly extracted entities and merge duplicates that automated resolution missed.

## References

- Graphiti repository: https://github.com/getzep/graphiti
- Zep documentation: https://docs.getzep.com/
- Neo4j graph database: https://neo4j.com/
- Temporal knowledge graph concepts in the project documentation

版本历史

  1. v2026.07 GitHub ↗
    原版更新说明

    ## Highlights - **First tagged release.** Every gate is green: `make check` (validate + 183 unit tests + eval-harness + numeric benchmarks), validate-catalog, quality-evals, and OpenSSF Scorecard on `main`. - **Rigor double loop at 13 method families.** Each family closes taxonomy tag → eval scenario (does the agent avoid the method's known traps?) → deterministic numeric benchmark (does it recover the true effect?). New this cycle: heterogeneous treatment effects (CATE) and quantile treatment effects (QTE). - **Paper-WorkFlow flagship at 29/29 executable rigor gates**, now merged to its `main` with a drift-gated README badge — the credibility artifact is checked by CI, so it can never silently go stale ([RIGOR.md](https://github.com/brycewang-stanford/Paper-WorkFlow/blob/main/RIGOR.md)). - **1,150 vendored skills / 69 collections**, all license-audited, security-scanned, tracked in a machine-readable catalog with CI-enforced freshness, and importable as a single whole-repo router skill. - **Hardened weekly upstream syncs**: synced trees must pass the full gate before merge; this cycle both sync PRs were verified, patched (StatsPAI's SkillOpt execution-gate card restored), and merged. - **First-party plugin marketplace** (Python / R / Stata empirical-analysis plugins + guide) shipped in the previous cycle, now under a tagged version. ## Known Follow-Ups - Ecosystem benchmark comparison (AERS vs econometrics agents on the shared benchmark suite) — [PLAN-2026-07 week 3](https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/main/docs/PLAN-2026-07.md). - Decomposition (Oaxaca–Blinder) and shift-share (Bartik) method families (weeks 3–4). - GitHub Pages catalog site, README demo screenshots, five-language README consistency script (week 4). ## Catalog - Top-level collections: **69** - `SKILL.md` files: **1150** - Cataloged agent tools: **335** ([`tools/CATALOG.md`](../tools/CATALOG.md)) ## Methodological rigor - Method families in the coverage map: **13** ([`RIGOR_COVERAGE.md`](RIGOR_COVERAGE.md)) - Eval scenarios: **26** (138 rubric items, 122 auto-checkable) ([`../eval-harness/`](../eval-harness/README.md)) - Benchmark tasks: **13** (70 gold checks, 58 required) ([`../benchmark/`](../benchmark/README.md)) ### Method families (skills tagged) - `iv`: 27 - `rdd`: 25 - `panel-fe`: 19 - `staggered-did`: 17 - `bayesian`: 13 - `did`: 10 - `event-study`: 10 - `matching`: 10 - `synthetic-control`: 10 - `cate`: 8 - `dml`: 8 - `survival`: 3 - `quantile`: 2 ### Eval scenarios by category - `causal-identification`: 14 - `citation-hygiene`: 1 - `reproducibility`: 2 - `research-integrity`: 1 - `runtime-safety`: 1 - `statistical-validity`: 2 - `writing-compliance`: 3 - `writing-style`: 2 ## License buckets - AGPL-3.0: 1 - Apache-2.0: 1 - CC-BY-NC-4.0 (non-commercial): 1 - CC-BY-SA-4.0: 1 - CC-BY-SA-4.0 (repository default): 4 - GPL-3.0: 1 - MIT: 35 - MIT Non-Commercial: 1 - Mixed (per-skill; see folder): 2 - UNKNOWN - check upstream: 22 ## Quality gates - `make check` — catalog/provenance/audit/eval freshness, repo hygiene, Paper-WorkFlow gates, stdlib unit tests, eval-harness lint + graded smoke, and the strict numeric benchmark (reference-gold drift fails the build). - External links and OpenSSF Scorecard run on their scheduled workflows.

当前收录版本

0.0.1

核对来源版本 ↗

质量评测

满分 100。前四项来自 DeepSkill 收录的质量报告。

场景价值92/100
评分依据

提供 Graphiti 框架的完整端到端使用指南,覆盖安装、配置、核心功能与科研工作流集成

触发精准95/100
评分依据

description 清晰描述核心功能,frontmatter keywords 覆盖知识图谱、AI 代理、时序图等关键术语

结构规范93/100
评分依据

文档结构清晰完整,包含 Overview、Installation、Core Features、Research Workflow Integration 等逻辑章节

执行可靠91/100
评分依据

提供完整的安装命令、Neo4j Docker 配置、环境变量设置和 Python 初始化代码

证据可验 (前四项均值)93/100
计算方式

取前四项原始分值的算术平均,四舍五入取整。可通过收录报告核对各项评分与说明,运行记录需单独查证。

权限与使用前检查

根据当前主文档自动整理,命中内容可能是示例或防护说明;不是实际权限声明,也未覆盖附件、外部依赖或运行行为。

  • 安装或获取外部依赖

    使用官方来源并锁定版本,在隔离环境检查安装脚本。当前文档检查不覆盖后续下载的代码。

    主文档提及 · 第 28 行
文档引用的域名或地址

这些地址可能用于文档链接或示例,不是已批准的联网清单。使用前核对实际目的地及发送的数据。

  • docs.getzep.com
  • github.com
  • localhost
  • neo4j.com

本站展示提示不会授予权限。涉及文件改动、密钥、外部发送或费用时,仍由你使用的工具按需确认。

安全评级与访问范围

risk · 静态证据核验,不代表运行安全保证。

使用前请由宿主核对并授权实际文件读写、外部服务和联网目的地;本站的核验不会授予任何运行权限。

来源与使用声明 SkillPi 的部分技能资料来自公开来源或创作者提交。第三方内容的权利归各自权利人所有。安装或运行前,请查看原站的许可、权限要求和使用风险。

来源与使用

请按详情页的作者、原始来源和许可证核对使用条件,运行前检查技能需要的权限。目录收录不等于已验证安装或安全性。