# Semantica

> Open-source, graph-native infrastructure that gives AI agents structured context, causal reasoning, and full decision provenance for explainable, traceable, and auditable AI systems.

Semantica is an open-source Python library (MIT licensed) that sits underneath LLMs, vector stores, and agent frameworks as a deterministic infrastructure layer for context graphs, decision intelligence, and full provenance tracking. It is installable via `pip install semantica` and is currently at v0.6.0, published July 2026. The project describes itself as "the glass-box alternative to black-box intelligence" and is positioned for high-stakes, regulated domains where AI outputs must be explainable and auditable.

## What It Is

Semantica is a graph-native AI infrastructure library that builds a structured, queryable context graph from enterprise data sources and records every AI decision as a first-class, auditable object. Unlike vector stores that retrieve similar text without understanding relationships, Semantica constructs a live knowledge graph of entities, relationships, and decisions with W3C PROV-O compliant lineage on every fact. It is fully self-hostable with zero vendor lock-in, supporting swappable backends for graph databases (Neo4j, FalkorDB, Apache AGE, AWS Neptune), RDF triple stores (Blazegraph, Apache Jena, Eclipse RDF4J), and vector stores (FAISS, Pinecone, Weaviate, Qdrant, Milvus, PgVector).

## Core Architecture and Modules

Semantica ships 13+ independently importable production modules spanning the full pipeline from ingestion to export:

- **Context Graphs** — Typed nodes and edges with temporal validity windows, SPARQL/Cypher queries, and point-in-time snapshots
- **Decision Intelligence** — `record_decision()`, `trace_decision_chain()`, `find_similar_decisions()`, `analyze_decision_impact()`, and `check_decision_rules()` turn every AI choice into a permanent, queryable graph node
- **Reasoning Engines** — Forward chaining, Rete network, Datalog, abductive reasoning, and SPARQL with fully explainable inference paths
- **Semantic Extraction** — Named entity recognition, relation extraction, event detection, and triplet generation from raw text
- **Provenance & Auditability** — W3C PROV-O lineage on every fact, exportable to JSON, CSV, RDF, and Parquet for regulator submission
- **Ontology Management** — Auto-generate OWL ontologies, validate with SHACL/HermiT/Pellet, manage SKOS vocabularies
- **Temporal Intelligence** — Allen Interval Algebra (13 relations), bi-temporal provenance, point-in-time graph snapshots, and `TemporalNormalizer`
- **Pipeline DSL** — Declarative, parallel stage chaining with validation gates and configurable retry policies

## Integration Ecosystem

Semantica connects to virtually every major AI tool category through native plugin bundles, an MCP server, and a REST API:

- **Native plugin bundles** for Claude Code, Cursor, and Codex CLI
- **MCP server + plugin** for Windsurf, Cline, Continue, VS Code, and OpenClaw
- **REST API** for GitHub Copilot, Roo Code, Goose, Aider, Amazon Q, and Zed
- **Agentic frameworks** — First-class Agno integration; LangChain, LangGraph, CrewAI, LlamaIndex, AutoGen, OpenAI Agents, and Google ADK supported via REST/MCP today, with dedicated native SDKs planned
- **Enterprise data platforms** — Native `DatabricksIngestor` (Unity Catalog + Delta Lake) and `SnowflakeIngestor` connectors
- **LLM providers** — All major providers via `semantica.llms` and LiteLLM: OpenAI, Anthropic, Gemini, Mistral, Llama, Groq, Cohere, Azure OpenAI, AWS Bedrock, Ollama, DeepSeek, and more

## Target Audience and Regulated Domains

The README explicitly targets AI/ML platform teams shipping consequential agents, data platform teams on Databricks or Snowflake, compliance and audit teams, and regulated enterprises in finance, healthcare, legal, government, and defense. The homepage lists vendor-published case studies including a marine natural capital finance deployment (Nereus) quantifying $1.62B in natural capital across Indo-Pacific sites, and an energy regulation deployment (Ausgrid) indexing 10 years of AER regulatory history with full provenance. The project states compliance alignment with GDPR, EU AI Act, HIPAA, and W3C PROV-O standards.

## Update: v0.6.0

Released July 21, 2026, v0.6.0 adds four notable capabilities: named-graph support for `JenaStore` completing cross-backend named-graph parity across Blazegraph, RDF4J, and Jena; parameterized SPARQL CONSTRUCT query templates extended to RDF4J and Jena with pipeline integration; a new `DatabricksIngestor` for Unity Catalog + Delta Lake ingestion with PAT/OAuth M2M auth; and a `SQLiteVecStore` disk-backed local vector store on `sqlite-vec`'s `vec0` virtual tables. The repository reports 3,100+ GitHub stars and 370+ forks as of the data snapshot.

## Features
- Context Graphs with typed nodes, edges, and temporal validity windows
- Decision Intelligence with record_decision(), trace_decision_chain(), find_similar_decisions()
- W3C PROV-O compliant provenance on every fact
- Forward chaining, Rete network, Datalog, and SPARQL reasoning engines with explainable paths
- Named entity recognition, relation extraction, event detection, and triplet generation
- OWL ontology generation, SHACL validation, and SKOS vocabulary management
- Temporal intelligence with Allen Interval Algebra, bi-temporal provenance, and point-in-time snapshots
- Polyglot graph storage: Neo4j, FalkorDB, Apache AGE, AWS Neptune, Blazegraph, Apache Jena, Eclipse RDF4J
- Vector store integrations: FAISS, Pinecone, Weaviate, Qdrant, Milvus, PgVector, SQLite
- Hybrid search combining vector similarity with graph traversal
- GraphRAG-native entity-aware, relation-aware, and ontology-aware document chunking
- Conflict detection and resolution across multiple data sources
- Entity resolution and semantic deduplication at scale
- Declarative parallel pipeline DSL with validation gates and retry policies
- Export to RDF, OWL, Parquet, Cypher, JSON-LD, and CSV
- Interactive Knowledge Explorer UI built on React 19 and Sigma.js
- MCP server for Claude Desktop, Windsurf, Cline, Continue, VS Code, and OpenClaw
- Native plugin bundles for Claude Code, Cursor, and Codex CLI
- REST API with endpoints for graph, decisions, reasoning, provenance, ontology, and export
- Databricks Unity Catalog + Delta Lake ingestion with PAT/OAuth M2M auth
- Snowflake ingestion with warehouse/database/schema and key-pair/OAuth auth
- Multi-agent shared context via first-class Agno integration
- CLI with 20+ command groups including ingest, reason, decision, provenance, and export
- Graph analytics: PageRank, betweenness centrality, Louvain community detection, Node2Vec, link prediction
- Audit trail export in W3C PROV-O format for GDPR, EU AI Act, and HIPAA compliance

## Integrations
Neo4j, FalkorDB, Apache AGE, AWS Neptune, Blazegraph, Apache Jena, Eclipse RDF4J, FAISS, Pinecone, Weaviate, Qdrant, Milvus, PgVector, Databricks, Snowflake, LangChain, LangGraph, CrewAI, LlamaIndex, AutoGen, OpenAI Agents SDK, Google ADK, Agno, Claude Code, Cursor, Codex CLI, Windsurf, Cline, Continue, VS Code, Claude Desktop, GitHub Copilot, Roo Code, Goose, Aider, Amazon Q, Zed, OpenAI, Anthropic Claude, Google Gemini, Mistral, Meta Llama, Groq, Cohere, Azure OpenAI, AWS Bedrock, Ollama, DeepSeek, LiteLLM, Kafka, Elasticsearch, Google Drive, MongoDB, DuckDB, Apache Arrow

## Platforms
WINDOWS, WEB, API, VSC_EXTENSION, DEVELOPER_SDK, CLI

## Pricing
Open Source, Free tier available

## Version
v0.6.0

## Links
- Website: https://getsemantica.ai
- Documentation: https://docs.getsemantica.ai/
- Repository: https://github.com/semantica-agi/semantica
- EveryDev.ai: https://www.everydev.ai/tools/semantica