Implement universal knowledge graph builder with LLM extraction and visual interface

[Copilot]

Implements a domain-agnostic knowledge graph construction tool that extracts entities and relationships from unstructured text using LLMs, with interactive visualization and a GUI interface.

Core Components

• Knowledge Graph (src/kg_builder/graph.py): NetworkX-based directed graph with entity/relationship management, JSON serialization, and graph statistics
• Extractor (src/kg_builder/extractor.py): LLM-powered extraction (OpenAI/Anthropic) with regex-based fallback for offline operation
• Visualizer (src/kg_builder/visualizer.py): PyVis interactive HTML graphs with color-coded entity types and draggable layouts
• Config (src/kg_builder/config.py): Environment-based LLM client initialization and settings management

Interfaces

• GUI (gui.py): Tkinter application with text input, domain selection, extraction controls, and visualization triggers
• CLI (cli.py): Command-line tool supporting batch processing, statistics display, and visualization generation
• Demo (demo.py): Interactive walkthrough demonstrating extraction → graph building → visualization pipeline

Usage

from kg_builder import KnowledgeGraph, KnowledgeGraphExtractor, KnowledgeGraphVisualizer

kg = KnowledgeGraph()
extractor = KnowledgeGraphExtractor(llm_client=openai_client)

extraction = extractor.extract_from_text(text, domain='technology')
kg.merge_from_extraction(extraction)

visualizer = KnowledgeGraphVisualizer(kg)
visualizer.visualize_interactive('output.html')

Or via GUI:

python gui.py  # Load text → Select domain → Extract → Visualize

Features

• Supports 7 domain contexts (medical, finance, technology, science, legal, education, general)
• Dual extraction modes: LLM-based (high accuracy) and pattern-based (no API required)
• Interactive visualizations with physics-based layouts and entity type coloring
• JSON import/export for graph persistence and incremental building
• UV project management with pip fallback

Testing

• test_complete.py: End-to-end validation of extraction → graph → visualization pipeline
• test_basic.py: Core data structure operations
• test_visualization.py: HTML output generation
• test_gui.py: Component initialization and workflow simulation

All tests passing with 0 CodeQL vulnerabilities detected.

Examples

Included examples demonstrate extraction from:

• examples/python_ecosystem.txt: Technology domain (programming languages, frameworks)
• examples/cardiovascular_system.txt: Medical domain (anatomy, diseases)

Original prompt

是一个现代化的、可视化的知识图谱构建工具。它不再局限于化学领域，而是提供了一套通用的流水线，帮助用户利用大语言模型 (LLM) 从非结构化或半结构化数据中提取信息，构建任意领域的知识图谱。我们以Python UV作为项目管理并主要通过运行gui.py运行，直接调用图形化界面进行操作。

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.