graph-sieve 1.1.0

Full Spectrum Graph Sieve - Automated Technical Term Extraction and Relationship Mapping

Graph-Sieve 🕸️📊

Full Spectrum Graph Sieve - Automated Technical Term Extraction and Relationship Mapping

graph-sieve is a standalone utility and service designed to extract relationship-aware domain knowledge from internal documents (.docx, .pptx, .msg, .pdf). It uses a multi-gate verifiable pipeline with local or remote models (OpenAI, Ollama, vLLM) to build a structured knowledge graph of technical terms and their relationships.

Features

• Multi-Gate Extraction: A robust pipeline (Detection -> Extraction -> Validation) ensuring high-fidelity term capture.
• Relationship Mapping: Beyond simple term lookup—builds a Property Graph of how terms relate.
• Multi-Format Support: Handles PDF, PPTX, DOCX, MSG, and images (via OCR) using Microsoft MarkItDown.
• Hebrew & Mixed-Language Handling: Specialized BIDI (Bi-Directional) support for Hebrew-English technical documents, ensuring technical terms are correctly extracted from mixed-language contexts.
• Flexible LLM Backend: Run locally with Ollama/vLLM for privacy, or use OpenAI for scale.
• Interactive Visualization: Generate dynamic, relationship-aware graph visualizations.
• MCP Server: Integrated Model Context Protocol (MCP) server for seamless AI agent integration.

Installation

pip install graph-sieve

Quick Start

1. Configure Your LLM: Create a .env file in your working directory:

   LLM_PROVIDER=openai
   OPENAI_API_KEY=your_key_here
   

   Or use Ollama (default):

   LLM_PROVIDER=ollama
   OLLAMA_BASE_URL=http://localhost:11434
   MODEL_NAME=llama3
   

2. Scan a Directory:

   graph-sieve-scan ./path/to/documents --dict my_dictionary.json
   

3. Visualize the Results:

   graph-sieve-visualize --dict my_dictionary.json
   

CLI Commands

• graph-sieve-scan: Extract terms from a directory or file.
• graph-sieve-lookup: Query terms and their graph context.
• graph-sieve-visualize: Generate an interactive HTML graph.
• graph-sieve-mcp: Launch the MCP server.
• graph-sieve-whois: Find the source document for a specific term.

Configuration (Environment Variables)

Variable	Description	Default
LLM_PROVIDER	openai, ollama, or vllm	ollama
OPENAI_API_KEY	Required if using OpenAI	None
OLLAMA_BASE_URL	URL for Ollama API	http://localhost:11434
MODEL_NAME	Model to use for extraction	gpt-4o-mini / llama3
STORAGE_DIR	Directory for graph data	Platform-specific

AI Agent Integration

Claude Desktop / Gemini CLI

To use Graph-Sieve as a tool, add it to your agent's config:

{
  "mcpServers": {
    "graph-sieve": {
      "command": "graph-sieve-mcp",
      "args": []
    }
  }
}

License

MIT License. See LICENSE for details.