Semantica

Overview#

Semantica is an open-source semantic layer & context graph framework designed to bridge the "semantic gap" and solve "black box" issues in AI systems. It provides a complete pipeline from unstructured data (PDF/DOCX/HTML/JSON/CSV/Excel/PPTX) to structured knowledge graphs, integrating NER, relation extraction, ontology generation, and vector retrieval.

Core Capabilities#

Semantics & Knowledge#

• Context Graphs — Structured knowledge representation with entity relationships and semantic context
• Decision Tracking — Full decision lifecycle management with precedent search and causal analysis
• KG Algorithms — Built-in centrality analysis, community detection, Node2Vec/DeepWalk embeddings
• Provenance Tracking — W3C PROV-O compliant data lineage tracking across 17 modules

Data Processing#

• Universal Ingestion — Support for PDF, DOCX, HTML, JSON, CSV, Excel, PPTX and database streaming
• Entity/Relation Extraction — NER, relation extraction, event detection with LLM enhancement
• Ontology Generation — 6-stage LLM pipeline for automatic OWL ontology generation with HermiT/Pellet validation
• Custom Ontology Import — Support for OWL, RDF, Turtle, JSON-LD formats

Governance & Quality#

• Conflict Detection — Semantic conflict detection and resolution
• Deduplication — Jaro-Winkler similarity-based entity deduplication
• Change Management — Enterprise-grade version control with SHA-256 integrity verification

AI Enhancement#

• GraphRAG — Knowledge graph-enhanced retrieval with multi-hop reasoning and semantic reranking
• Unified LLM Interface — Support for Groq, OpenAI, HuggingFace, LiteLLM (100+ LLMs)

Architecture#

Three-layer design:

1. Input Layer — Governed data ingestion with Docling, OCR, APIs
2. Semantic Layer — Trust & reasoning engine for NER, relation extraction, ontology induction, deduplication, conflict detection
3. Output Layer — Auditable knowledge assets: Knowledge Graphs, OWL Ontologies, Vector Embeddings

Storage Backend Support#

• Vector Store: FAISS, PostgreSQL/pgvector, Weaviate, Qdrant, Milvus, Pinecone, InMemory
• Graph Store: Neo4j, FalkorDB, Amazon Neptune, Apache AGE
• Triplet Store: Blazegraph, Jena, RDF4J

Installation#

# PyPI (Recommended)
pip install semantica

# With all optional dependencies
pip install semantica[all]

# Development install from source
git clone https://github.com/Hawksight-AI/semantica.git
cd semantica
pip install -e ".[all]"

Quick Start#

from semantica.context import AgentContext, ContextGraph
from semantica.vector_store import VectorStore

# Initialize
vs = VectorStore(backend="faiss", dimension=768)
kg = ContextGraph(advanced_analytics=True)
context = AgentContext(
    vector_store=vs,
    knowledge_graph=kg,
    decision_tracking=True,
    advanced_analytics=True,
    kg_algorithms=True,
)

# Store memory and auto-build context graph
memory_id = context.store(
    "User is working on a React project with FastAPI",
    conversation_id="session_1"
)

# Record decision
decision_id = context.graph_builder.add_decision(
    category="technology_choice",
    scenario="Framework selection for web API",
    reasoning="React ecosystem with FastAPI provides best performance",
    outcome="selected_fastapi",
    confidence=0.92
)

Typical Use Cases#

• 🏥 Healthcare: Clinical decision support, drug interaction analysis, medical literature reasoning
• 💰 Finance: Fraud detection, regulatory compliance (SOX, GDPR, MiFID II), credit risk assessment
• ⚖️ Legal: Evidence-based legal research, contract analysis, regulatory change tracking
• 🔒 Cybersecurity: Threat attribution, incident response, security audit trails
• 🏛️ Government & Defense: AI system governance, policy decisions, defense intelligence
• 🚗 Autonomous Systems: Vehicle decision logging, robotics safety

Design Principles#

• Opt-In Design — Provenance disabled by default, zero breaking changes
• Modular Architecture — Each module can be used independently
• Production Ready — Comprehensive error handling and scalability design