samyama-graph

概述

Samyama 是一个基于 Rust 的图向量数据库，能够处理海量数据（单台机器上可容纳高达 7400 万节点和 10 亿边）。它实现了 OpenCypher 查询语言，并提供 Redis 协议支持和 HTTP 端点。该项目包含一个专门的 MCP 服务器，允许 AI 代理通过自然语言查询知识图谱。Samyama 展示了令人印象深刻的性能基准，包括仅用 2.50 美元查询 10 亿条边。其 MCP 集成尤为突出，因为它可以从图模式自动生成 MCP 服务器，为任何图数据集实现即时 AI 代理功能。

试试问 AI

装完之后，这里有 6 个你可以让 AI 做的事：

• query_graph

  Execute OpenCypher queries on the graph database

  试试:Find all drugs tested in cancer clinical trials试试:Who are Alice's friends of friends that work at Google?试试:What is the shortest path between node A and node B?
• run_graph_algorithm

  Execute graph algorithms like PageRank or community detection

  试试:Calculate PageRank for the social network graph试试:Find strongly connected components in the graph试试:Perform triangle counting on the network
• vector_search

  Perform semantic search using vector embeddings

  试试:Find papers similar to this research article试试:Search for documents related to 'machine learning'试试:Find the most similar nodes based on embeddings
• load_graph_data

  Load domain-specific knowledge graphs into the database

  试试:Load the clinical trials dataset into the graph试试:Import the drug interactions knowledge graph试试:Load the cricket simulation data
• analyze_biomedical_data

  Query biomedical knowledge graphs for research insights

  试试:Find drugs most tested in cancer clinical trials试试:What are the known side effects of this medication?试试:Find NCI-funded clinical trials for specific drugs
• natural_language_query

  Convert natural language questions to Cypher queries

  试试:Who are Alice's friends of friends that work at Google?试试:Show me the most tested interventions in diabetes trials试试:Find all articles related to aspirin clinical trials
• create_vector_index

  Create vector indexes for semantic search capabilities

  试试:Create a vector index on Paper nodes with 384 dimensions试试:Set up cosine similarity index for document embeddings试试:Build a vector search index for node properties
• execute_algorithm

  Run specific graph algorithms on the database

  试试:Run PageRank algorithm on the social graph试试:Calculate local clustering coefficients for all nodes试试:Perform breadth-first search from a starting node
• analyze_fraud_patterns

  Detect fraud patterns in banking and financial networks

  试试:Find potential money laundering patterns in transactions试试:Identify clusters of suspicious accounts试试:Trace the flow of high-risk transactions
• analyze_supply_chain

  Analyze supply chain networks for disruptions and optimization

  试试:Identify potential bottlenecks in the supply chain试试:Find alternative routes for disrupted shipping lanes试试:Analyze port congestion impacts on delivery times
• analyze_clinical_trials

  Query clinical trial data for research insights

  试试:Find clinical trials testing specific interventions试试:Match patients to eligible clinical trials试试:Analyze adverse events reported in trials
• analyze_manufacturing

  Analyze manufacturing networks for digital twins and failure prediction

  试试:Simulate failure cascades in production lines试试:Identify critical components in manufacturing processes试试:Optimize maintenance schedules based on component dependencies

说明：Tools were inferred from the README's documentation of capabilities and examples. While there's no explicit 'Tools' section, the README demonstrates various functions through examples and feature descriptions. The tools represent the main f

可对比工具

安装

# 安装 Samyama
git clone https://github.com/samyama-ai/samyama-graph && cd samyama-graph
cargo build --release
./target/release/samyama  # 在 :6379 启动 RESP，在 :8080 启动 HTTP

# 安装 MCP 服务器
pip install samyama[mcp]

# 使用演示数据启动 MCP 服务器
samyama-mcp-serve --demo cricket

{
  "mcpServers": {
    "samyama": {
      "command": "python",
      "args": ["-m", "samyama.mcp"],
      "env": {}
    }
  }
}

FAQ

Samyama 支持哪些查询语言？

Samyama 支持 OpenCypher 查询（约 90% 兼容性），以及使用 Cypher 扩展的向量搜索 API。它还接受会被翻译成 Cypher 的自然语言查询。

Samyama 能否处理多领域知识图谱？

是的，Samyama 已经加载并查询了多个相互连接的知识图谱，包括 PubMed、ClinicalTrials.gov、Reactome 和 DrugBank，同时在单台机器上运行。

MCP 服务器集成是如何工作的？

Samyama 提供了一个 'samyama-mcp-serve' 命令，可以从您的图模式生成 MCP 服务器工具，使 AI 代理能够查询数据库，而无需自定义工具定义。

samyama-graph 对比