欢迎使用 Semantica

# 基础安装；[all] 会安装全部可选依赖
pip install semantica
pip install semantica[all]

# 通过环境变量与 YAML 配置文件完成 Semantica 基础配置
import os
from pathlib import Path

# 设置 API 密钥与默认嵌入提供商
os.environ["SEMANTICA_API_KEY"] = "your_openai_key"
os.environ["SEMANTICA_EMBEDDING_PROVIDER"] = "openai"
os.environ["SEMANTICA_MODEL_NAME"] = "gpt-4"

# 定义 YAML 配置：嵌入模型、维度与知识图谱后端
config_text = """api_keys:
  openai: your_key_here
  anthropic: your_key_here
embedding:
  provider: openai
  model: text-embedding-3-large
  dimensions: 3072
knowledge_graph:
  backend: networkx
  temporal: true
"""
# 将配置写入磁盘并读回验证
Path("config.yaml").write_text(config_text, encoding="utf-8")
Path("config.yaml").read_text(encoding="utf-8")

# 创建一个小型示例文档，供后续流水线演示使用
from pathlib import Path

docs_dir = Path("welcome_docs")
docs_dir.mkdir(exist_ok=True)
text_path = docs_dir / "apple.txt"
# 写入一段关于 Apple 公司的示例文本
text_content = (
    "Apple Inc. was founded by Steve Jobs, Steve Wozniak and Ronald Wayne in"
    " Cupertino, California."
)
text_path.write_text(text_content, encoding="utf-8")
print(f"Created sample document at {text_path}")

# 最小端到端流水线：摄取 → 解析 → 规范化 → 抽取 → 建图 → 嵌入 → 混合检索
from semantica.ingest import FileIngestor
from semantica.parse import DocumentParser
from semantica.normalize import TextNormalizer
from semantica.semantic_extract import NERExtractor, RelationExtractor
from semantica.kg import GraphBuilder, GraphAnalyzer
from semantica.embeddings import EmbeddingGenerator
from semantica.vector_store import VectorStore, HybridSearch

# 1. 摄取文档目录
ingestor = FileIngestor()
documents = ingestor.ingest(str(docs_dir))

# 2. 解析为结构化文档
parser = DocumentParser()
parsed_docs = parser.parse(documents)

# 3. 规范化文本
normalizer = TextNormalizer()
normalized_docs = normalizer.normalize(parsed_docs)

# 4. 抽取实体与关系
ner = NERExtractor()
entities = ner.extract(normalized_docs)
rel_extractor = RelationExtractor()
relationships = rel_extractor.extract(normalized_docs, entities)

# 5. 构建并分析知识图谱
builder = GraphBuilder()
kg = builder.build(entities, relationships)
analyzer = GraphAnalyzer()
metrics = analyzer.analyze(kg)

# 6. 生成文本嵌入并存入向量库
emb_generator = EmbeddingGenerator()
embeddings = emb_generator.generate_embeddings(documents, data_type="text")

vec_store = VectorStore()
vec_store.store(embeddings, documents, metadata={})
# 7. 执行混合语义检索
hybrid = HybridSearch(vec_store)
search_results = hybrid.search("Apple founders", top_k=3)
len(search_results)

from semantica.visualization import KGVisualizer

# 创建一个可视化器实例
viz = KGVisualizer(layout="force", color_scheme="vibrant")

# 生成交互式网络可视化
# 这会返回一个在 notebook 中渲染的 Plotly figure 对象
fig = viz.visualize_network(kg, output="interactive")
fig.show()

from semantica.ontology import OntologyGenerator

generator = OntologyGenerator(base_uri="https://example.org/ontology/")

# 从已抽取的数据生成本体
ontology = generator.generate_ontology({
    "entities": entities,
    "relationships": relationships
})

# 查看推断出的类
[cls["name"] for cls in ontology.get("classes", [])[:5]]

# 使用 TextSplitter 将文档切分为适合 RAG 的小块
from semantica.split import TextSplitter

# 设置块大小与相邻块之间的重叠
splitter = TextSplitter(chunk_size=100, chunk_overlap=20)
chunks = splitter.split_documents(documents)

print(f"Original documents: {len(documents)}")
print(f"Generated chunks: {len(chunks)}")

from semantica.reasoning import Reasoner

# 简单规则：如果 X 创立了 Y，那么 X 为 Y 工作
rule = """
IF (?x founded ?y) THEN (?x works_for ?y)
"""

reasoner = Reasoner()
reasoner.add_rule(rule)
inferred_facts = reasoner.infer_facts(kg)

print(f"Inferred {len(inferred_facts)} new facts")

# 将知识图谱导出为 JSON 文件以便持久化
from semantica.export import GraphExporter

exporter = GraphExporter()
exporter.export(kg, format="json", output_path="knowledge_graph.json")
print("Graph exported to knowledge_graph.json")

# 使用核心 Semantica 类进行生命周期管理与知识库构建
from semantica.core import Semantica, ConfigManager

# 从配置文件加载配置
config_manager = ConfigManager()
config = config_manager.load_from_file("config.yaml")

# 初始化框架
framework = Semantica(config=config)
framework.initialize()

# 一键构建知识库（同时生成嵌入与图谱）
kb_result = framework.build_knowledge_base(
    sources=[str(docs_dir)],
    embeddings=True,
    graph=True,
)

# 关闭框架并查看构建结果
framework.shutdown()
sorted(kb_result.keys())